Smid

Credit where it’s due: Ladislav Smid had his best game as an Oiler yesterday.

He’s still lacking consistency and this season has displayed that ridiculous tendency to stop thinking when things are happening in front of the net, but he made a lot of big league plays in that game Sunday.

If Smid can continue to make smart passes, have an idea on his rushes, close gaps with authority, use his stick effectively to remove the puck from harm’s way, and do it consistently, then the Oilers may have something here.

We’ll need to see it again, but before yesterday we hadn’t seen that player at all. The light, as they say, appears to be “on.”

written by

The author didn‘t add any Information to his profile yet.
Related Posts

134 Responses to "Smid"

  1. B.C.B. says:

    Plus Smid hits! a lot.
    But LT you are right as usually: that was Smid’s best game (ever). I have always been cheering for him (no not because who we traded him for).

    The last thing I want to do is trade Smid: unless it brings back both a 3C and top 4D.

  2. Coach pb9617 says:

    We’ll need to see it again, but before yesterday we hadn’t seen that player at all.

    I disagree. It’s been coming this season. Slowly and sporadically, but it’s been there.

  3. kamus says:

    I have been waiting for a performance like that for a long time. Even as a Smid backer I started losing some faith when he couldn’t crack the opening day roster.
    On Sunday he just put it all together. He was knocking people off the puck on the boards, keeping good position, and getting the puck up ice smartly.
    Hopefully he builds on this and doesn’t look back.

  4. NotLeeFogolin says:

    Two seasons ago I was convinced he was the second coming of Jason Smith. In a single game they both fought, and they both grinned while doing it.

  5. Bruce says:

    He’s still lacking consistency … We’ll need to see it again, but before yesterday we hadn’t seen that player at all.

    LT: Hmmm. No doubt Smid’s confidence is growing as he starts to get a regular shift — after starting the season in the PB he’s played 6 out of the last 7, inc. all four games this past week — but so far he’s been pretty consistent in my view. I wrote on my blog a week ago that in limited ice time Smid had the best GA/60 while facing the highest QualComp of any Oiler defenceman. Four games and 0 GA later, he still holds both distinctions by a significant margin according to Behind the Net.

    Now I’m not sure how much stock we can put in QualComp at this early point in the season. It would seem odd for MacT to be giving the guy third pairing ice time (13-15 minutes a game, like clockwork) against top opposition, but as I understand Desjardins’ methods the guys he’s been playing are the guys who have actually been performing well, not necessarily the ones who you might expect to over the long haul. No matter how you slice it, 1 GA in 95 minutes TOI doesn’t exactly fit the description of “inconsistent”.

    Looking at that BtN link, it’s clear the consistency problems so far lie with Grebeshkov and Gilbert, who depite having faced the weakest QualComp have each been on the ice for more than half of the even strength goals the Oilers have allowed this season (9 of 16). Both are -4; no other Oiler defenceman is in red figures.

  6. godot10 says:

    I’d sort of like to see MacT keep Smid and Gilbert together for awhile, and play Grebeshkov with Staios.

  7. Scott says:

    Bruce,

    I think you have to throw out QUALCOMP/QUALTEAM for the first 30 games or so.

    That said, if he’s leading GAON/60 he’s excelling the most in his role on the defensive side of things, which is certainly a nice accomplishment.

    If you want to see his competition, just look it up for the past few games on Vic’s timeonice site. We know who the tough assignments are by seeing the names.

  8. misfit says:

    Smid has looked good by my eye as well, but it shoudln’t come as any surprise really. For one, he’s been playing on the 3rd pair this year, and this time around, he’s been paired with Staios (and Gilbert yesterday) instead of Greene. The other thing is, he’s been slowly but steadily improving as a player for some time now.

    Though I’ve never considered him a quality puck mover, but for the last few games he’s really seemed to pick up that part of his game. He hasn’t been taking the extra lap around the defensive zone, his passing has been better, and he hasn’t been trowing hand grenades to his defence partner near as much. Hopefully he can keep it up.

  9. MJT says:

    I agree that Smid has been coming along slowly this season, but that there have been a lot of positives over the last few games and he’s looking good out there. I’m guessing that when Grebs gets back Strudwick will be the odd man out, but he’s been solid so far as well.

  10. mjsh says:

    now if they could only kill penalties.

  11. Slipper says:

    I think it's practical to point out that Smid has the highest "on ice even strength save percentage" amongst ALL Oilers players. Only 1 goal for 51 shots on net (.981%). The goaltenders are due a good portion of the credit for Smid's team leading GA/60.

    http://timeonice.com/playershots.php?team=EDM&first=20001&last=21230

  12. Yeti says:

    Is it fair to say that keeping Smid and trading Greene was the preferable option for the club (presuming that there was an option)? I would suggest that if Smid continues on the current trajectory, he will be a better player than Greene by the end of the season.

  13. Sean says:

    Maybe he saw Sbisa playing his game as an 18 year old and finally decided enough was enough. kiddin. But he has looked better especially yesterday aft.

    I wonder if there is ways to use last years numbers to direct QUALCOMP/QUALTEAM the right way early in the season.

  14. Dennis says:

    This is the latest thread so I’ll put up what appears to be the latest news.

    Listening to Eric Cole doing an interview on NHL Live and he said that Steve Macintyre has been told that he can get his own place once he gets back to Edm so it looks like he’ll be staying here for more than the shorterm.

    So, let’s all take a minute to keep Bruce in our hearts as he chews on this news and wonders what it means for Stortini!!!

  15. Quain says:

    I would suggest that if Smid continues on the current trajectory, he will be a better player than Greene by the end of the season.

    It’s like you drew a line in the sand for Smid, but the line was a quarter mile behind him.

  16. Slipper says:

    It’s also important to note that Gilbert and Grebeshkov are the recipients of the worst EV goaltender save percentage amongs Oilers’ defenders.

    If Grebeshkov’s EV SvPct were adjust from .866% to the same as Smid’s (.981%), Grebeshkov would be +5/-1 for the season instead of +5/-9.

    Smid 0.981%(1)- 0.63 GAON/60(1)
    Staios 0.969%(2)- 0.96 GAON/60(2)
    Strudwick 0.957%(3)-1.36 GAON/60(3)Souray 0.951%(4)- 1.50 GAON/60(5)
    Visnovsky 0.949%(5)-1.49 GAON/60(4)
    Gilbert 0.893%(6)- 3.14 GAON/60(6)
    Grebeshkov 0.866%(7)- 3.67 GAON/60(7)

  17. Slipper says:

    As far as charts go that was a massive fail.

    What’s super neat is how goalie performance can influence the perception of defense.

  18. Yeti says:

    It’s like you drew a line in the sand for Smid, but the line was a quarter mile behind him.

    Great line!
    But do you really think Smid has already well surpassed Greene? This is the Smid that couldn’t hold down a place coming out of camp.

  19. Scott says:

    I know that this is why Dennis is tracking scoring chances, but, to an extent, a lower save percentage probably indicates higher percentage shots. When Dennis throws up his scoring chance numbers we’ll be able to get some idea of the context of some of these shots. If Smid is also making a killing by that metric, I think it would be more difficult to throw his strong GA/60 as simply save percentage. So I counted it up for the first eight games and the scoring chances break down as follows at EV:

    Smid +11 -15

    Grebeshkov +34 -37

    Gilbert +31 -40

    During the first eight games Smid only played in four, so and he’s playing fewer minutes, so he wasn’t even really a “safer” player. The conclusion that a good part of these results are just save percentage would then be strengthened, at least for the first eight games of the year.

  20. Schitzo says:

    Slipper: That cuts both ways, though. Considering that Grebs tends to get brain cramps that lead to odd-man rushes the other way, I’m not surprised that the SV% is lower behind him – no goalie is going to put up a .920 facing nothing but 2-on-1s

  21. Doogie2K says:

    What’s super neat is how goalie performance can influence the perception of defense.

    Well, yes, but there’s only so much you can disentangle the two. I mean, if a D-man makes a lot of boneheaded plays, he’s going to see a lot more chances and a lot more high-percentage chances go against the goalie, and more pucks will wind up in the back of the net. At the same time, if a goalie’s being a sieve, any minor mistake by a D can end in disaster when it really shouldn’t.

    As Scott suggests, that should show up in Dennis’s chance numbers: obviously, if Smid looks good by that metric, then he deserves a ton of credit for getting his shit together; if Smid looks average or bad by that metric, then the goalies are saving his bacon, and that’s not sustainable. I’m looking forward to seeing how the last three games shake out in the chance department, because they should confirm what we’re seeing with our eyes here.

  22. raventalon40 says:

    He’s Smidtacular.

    Dennis: I think Stortini is still the better overall player but he’s not what the team needs right now as a deterrent to other fighters.

    The “team toughness” aspect was causing a lot of injuries.

    But that’s the regular season. You would more likely be inclined to play a Stortini in the playoffs than a MacIntyre.

  23. Slipper says:

    Just because a player has a high save percentage behind them doesn’t mean they aren’t playing decent either. In the case of Smid, I think it’s fair to argue he’s made some huge strides with his game. On the other hand, he isn’t a world beater either, and barring really good fortune, it isn’t likely that the goaltenders will maintain a .981 save percentage when he’s on the ice.

    In the case of Tom Gilbert, without taking an in depth look at his competition for the first handful of games, it’s highly unlikely that they’ll continue to post a scoring percentage twice as high as Gilbert and his teammates.

    The greater point, regardless of who has performed well or not, is that after a very small smaple size of games played, the perception of the observer appears to be distorted in favour of those who get the saves, and against those who don’t.

  24. Slipper says:

    In regards to quality of opportunities, I don't neccessarily disagree with any of your posted theories, yet GAON/60 doesn't communicate quality of chances against. It doens't communicate quality of chances for, either!

    This early into the season, where a guy has been on the ice for 100 shots on goal or less, it is fairly useless to rely on most of these numbers (GA, save percentage, qualcomp, etc.). Everyone knows which numbers I prefer to use:

    http://timeonice.com/playershots.php?team=EDM&first=20001&last=21230

    People should begin to like these numbers this year because they paint the Oilers' as a half decent team. Ecspecially some of the fan faves!

  25. Doogie2K says:

    People should begin to like these numbers this year because they paint the Oilers’ as a half decent team. Ecspecially some of the fan faves!

    Holy shit! Horcoff has the best outshooting numbers of any forward on the team! Someone sign that guy, quick!

    (I also notice that Pisani is getting absolutely skullfucked by this metric. It makes sense, but it’s no less sad to see.)

  26. HBomb says:

    But that’s the regular season. You would more likely be inclined to play a Stortini in the playoffs than a MacIntyre.

    A team wanting to do good in the playoffs would acquire an extra forward at the deadline so they can nail both Stortini and MacIntyre’s asses to the bench once mid-April rolls around.

    d2k: On Pisani, I’d bet my left testicle that Fernando’s numbers will improve greatly if/WHEN he is returned to playing on the wing (i.e. when MacT figures out putting Pouliot in as the 3C and dropping someone to 4LW is the best thing for this team, hurt feelings be damned).

  27. Bruce says:

    The greater point, regardless of who has performed well or not, is that after a very small smaple size of games played, the perception of the observer appears to be distorted in favour of those who get the saves, and against those who don’t.

    Slipper: Those are good points, and I particularly agree with the cited one. Where I (and apparently others) differ from the Statzis is what that Sv% ON actually means. To me the goaltenders are bound to have lower Sv%s behind poor defenders, whereas the position of you, Tyler, Vic and others seems to be — and PLEASE correct me if I’m wrong — that it’s all a matter of clattering dice, all will normalize toward the mean with time. Certainly .981 cannot continue, and some normalization must occur, but at the end of the year Randy Gregg will have a better Sv%ON than Paul Coffey, I’d bet your life on it! :)

    At face value, I think we can already conclude that Smid has limited the number of 50-cent opportunities from guys who beat him to the edge of the crease which was his huge weakness in the early part of last season.

    The shots data from the link Slipper posted are interesting to say the least. Below I have normalized the raw numbers to ES minutes played. The numbers shown in brackets are from Slipper’s rankings of Sv% ON and GA ON/60, and the juxtaposition is startling:

    ESSA ON/60
    ——————–
    1.Grebeshkov 26.5 (7,7)
    2.Visnovsky 26.9 (5,4)
    3.Gilbert 28.3 (6,6)
    4.Souray 30.1 (4,5)
    5.Staios 30.3 (2,2)
    6.Strudwick 31.2 (3,3)
    7.Smid 32.9 (1,1)
    ——————–

    So the shots allowed numbers are pretty much polar fucking opposite the GA data. It’s almost stunning.

    What can we draw from this? Firstly, it’s small number statistics in both cases, so caveat emptor. Neither measure is perfect at the best of times; the truth no doubt lies somewhere between the two.

    I also note the separation between defensive defenders, 5, 24, and 43, vs. the puck movers 37, 71, and 77 with 44 lodged pretty squarely in the middle. The former all have high shots allowed, low goals against, and by extension, great Sv% behind them. The puck movers are good at preventing shots, but not so good at limiting the quality of those shots. Is it reasonable to suppose they spend less time in their own zone cuz of their mobility and puckmoving prowess, but allow higher percentage shots through some combination of more high risk plays (a.k.a. turnovers), and simply poorer defensive skills without the puck?

    Again I can’t stress enough that all these data are probably too small to be reliable; however the trend of not just individual defencemen but clusters of stay-at-home vs. puck-movers is at the least, worth following. Perhaps it will all disappear with time, but early returns from 2008-09 are suggestive of an inverse relationship between shot quantity and shot quality.

  28. Schitzo says:

    A team wanting to do good in the playoffs would acquire an extra forward at the deadline so they can nail both Stortini and MacIntyre’s asses to the bench once mid-April rolls around.

    Fuck, how good of a team would this be if we could acquire one more bottom-6 LW? Someone like Kyle Calder could probably had for the price of his plane ticket.

    If your bottom 6 is
    Moreau-Brodziak-Cole
    Calder-Pouliot-Pisani

    You’ve got to imagine you’re in pretty good shape.

  29. Doogie2K says:

    On Pisani, I’d bet my left testicle that Fernando’s numbers will improve greatly if/WHEN he is returned to playing on the wing (i.e. when MacT figures out putting Pouliot in as the 3C and dropping someone to 4LW is the best thing for this team, hurt feelings be damned).

    I hope so. Vic’s tool only goes back to last year, which is probably not the best benchmark for what we know about Pisani’s established level of ability, so it’s hard to say what his numbers should look like. You’d think they’d be at least somewhat in the black, but then if he’s facing tough competition, maybe not.

  30. Asiaoil says:

    There will be more a lot more lessons for Smid to learn this season – but this is the point in terms of games played where the light starts to go on for quality defensive dmen. He’s still only 22 so he’s actually way ahead of the curve for this type of player who typically doesn’t really settle in until 25. I still think he has top pair “shutdown” upside and would like to see him matched with a guy who has a bit of offense but is still solid in this own end (Gilbert?).

    He’s big, mean, coachable, can skate and pass so all he needs is time and experience. Patience is the key word here…

  31. Quain says:

    Fuck, how good of a team would this be if we could acquire one more bottom-6 LW? Someone like Kyle Calder could probably had for the price of his plane ticket.

    I can’t imagine that even a healthy Brule wouldn’t make the fourth line hum. What MacT (and to a lesser extent, the GM) is doing to the bottom six is criminal.

  32. kamus says:

    Let me get this right, are some of the “stat guys” claiming that Grebeshkov is playing well and is just unlucky that the goalies aren’t stoping enough pucks for him?

  33. HBomb says:

    Schitzo: I’d try for a bottom six RW myself.

    Pisani-Pouliot-Cole
    Moreau-Brodziak-????

    A LW would also work, because if Moreau-Brodziak-Pouliot was your FOURTH line, well hell, you’d be icing one damn good hockey team.

    As someone else suggested, maybe even Brule is the solution once they get down to only two goaltenders.

  34. Jonathan says:

    Fuck, how good of a team would this be if we could acquire one more bottom-6 LW? Someone like Kyle Calder could probably had for the price of his plane ticket.

    Michel Ouellet has been put on re-entry waivers. Take him and run.

  35. Fingort says:

    Speaking of young defencemen that seem to be ‘getting’ it: Have any of you noticed Danny Syvret’s numbers so far this season with the Phantom’s of the AHL? In 10gp, Syvret has 11 points, and must be considered for a recall soon, which is a vast improvement on his previous minor league numbers. Granted, the Flyers are icing a fairly inexperienced top 6 right now, with the 18 year old Sbisa playing while Syvret toils in the minor leagues, but, interesting start to the season anyways.

  36. Dennis says:

    I was thinking just last night how the 4th line might look with Brule fleshing out the 51/78 combo and I think it might work. But, Schremp’s probably gonna get the first call and maybe that might work too.

    One of the things I’d like to see happen soon is 26 playing with 12-89. I like 12-89 the most of the kids and I think 26 might mesh nicely with them as the go-to-guy. I’d leave 27 with 10-83 because I think that’s the best way to try and get some worth from him and it’s obvious that 78 or 51 should be the third line pivot. That would put 13 on the 4th line but that’s not the end of the world, either.

    Back to Smid for a second, if him and 24 are gonna be the third pairing then they Should put up awesome numbers; especially on home ice when MacT has the chance to run the show.

  37. mc79hockey says:

    To me the goaltenders are bound to have lower Sv%s behind poor defenders, whereas the position of you, Tyler, Vic and others seems to be — and PLEASE correct me if I’m wrong — that it’s all a matter of clattering dice, all will normalize toward the mean with time. Certainly .981 cannot continue, and some normalization must occur, but at the end of the year Randy Gregg will have a better Sv%ON than Paul Coffey, I’d bet your life on it! :)

    That sounds right to me but really, it’s an issue of how big the difference is going to be. All other things being equal Bruce, in a league with .920 ES goaltending, a million rolls of the dice, what do you think that the save percentage is going to be behind Coffey versus the save percentage behind Gregg?

    What can we draw from this? Firstly, it’s small number statistics in both cases, so caveat emptor.

    I woulda stopped there.

  38. Oilmaniac says:

    I think you guys are also leaving out a key component…

    (just looking at the last three games superstats) Smid was only on the pk the game strudwick was out (vs. Carolina).. vs. Nash & Phi he was barely on the ice for a kill…

    So, doesnt that automatically point towards a higher expected save%… grebs n gilbert are leaned on sig. more…

    I like how Smid has been throwing the body.. hes become reliable in that regard…

    cheers,

  39. Bruce says:

    So, let’s all take a minute to keep Bruce in our hearts as he chews on this news and wonders what it means for Stortini!!!

    Dennis: thanks for your kind thoughts. You may recall I wrote a lengthy comment on the BofA site last spring — can’t seem to find it now, but’s it’s in one of the threads there — in which I compared Stortini to Calgary goon Eric Godard. I listed all sorts of statistics, from minutes played to GF:GA ON/60 to points, hits and giveaways per 60 to penalties drawn vs. taken to instigator penalties (not) taken to you name it, and statistically Zack was superior in every way, and by a large margin. Now as I see it, MacIntyre = Godard, which is a major step back for this club competitively.

    So far this year MacIntyre has had about half an hour of ice time, during which time the Oilers have generated all of 5 shots on goal. The guy brings no offence to the table, none. His Corsi number is already -15. That’s a guy you just can’t afford to put out there very often, witness his sub-4:00 per game to date.

    Now Zack’s start isn’t a whole lot better, but Zack’s track record is a lot more extensive at the NHL level despite his being 5 years younger. And Zack showed last year that when given regular ice time — something which he has not been afforded to date this season — he could deliver not badly. I’ve cited his +/- splits before, but here are his half season totals from the Time on Ice charts, split at Game #615:

    Stat: 1st half | 2nd half | Total
    ————————————–
    GF/GA: +3/-10 | +18/-8 | +21/-18
    SF/SA: +63/-121| +155/-143| +218/-264
    Corsi +/-: +116/-207| +273/-284| +389/-491
    Sh% ON: 4.8% | 11.6% | 9.6%
    Sv% ON: .917 | .944 | .932

    … which showed across-the-board improvement that surely is off-the-charts in terms of degree. Not only did the attempted shots greatly even out in the second half, the percent that got through to on goal rose from 54% to 57%, while the opposition’s was reduced from 58% to 50%.

    So I’m a little baffled as to how little rope has been given this guy who led our team in +/- in the second half of last season. MacT made some comments about how Zack’s attitude and work ethic have been great but at a certain point he has to get things done on the ice, which is fair, but he has already shown he can get things done given a regular, well-defined role. It’s a bit of a Catch 22, but to me it’s hardly worth the difference of SMac throwing punches with Riley Cote vs. Zack throwing shorter range punches at the same guy. Those who enjoy the sight of blood might vote otherwise, but to me the most important part is showing up at the dance, and Zack has never let Oilers down in that dept. While the clock is running, he’s a far superior option.

    My hope is it’s just a speed bump, like Smid starting the season in the PB, that once SMac delivers a couple of messages he can hang around the PB and we can have a fourth liner that can move his legs at least a little and play defence a little and move the puck in the right direction a little and pound the boards a lot.

  40. Jonathan says:

    I was thinking just last night how the 4th line might look with Brule fleshing out the 51/78 combo and I think it might work. But, Schremp’s probably gonna get the first call and maybe that might work too.

    Or Potulny. 8 goals in 9 games and pro experience to boot. Toss in his versatility, and he really ought to have a leg up on Schremp. Brule’s injury should rule him out for a couple of weeks too – the last thing he needs is to come to the NHL at less than 100%. The other thing about Brule is there’s an awfully small window where he can play NHL w/o becoming waiver eligible.

  41. Bruce says:

    //What can we draw from this? Firstly, it’s small number statistics in both cases, so caveat emptor.//

    I woulda stopped there.

    Point taken, MC, and if I was asking you to referee a paper for Nature magazine I would heed your advice. Since, however, I am just a lowly hockey fan throwing out stuff at a blog, if I see an interesting pattern I’m apt to mention it. And I’ll follow my little hypothesis until it proves meaningless, which it very likely will soon enough. But on the off-chance it doesn’t, I’m on it. And if in the process I risk making a fool out of myself, well, been there, done that. :)

    Oilmaniac: the cited Sv% ON stats are for even strength situations only.

  42. Lowetide says:

    The next thing to watch and see re: Smid is if he can move up the depth chart. There are still a lot of young men who are going to push for roster spots. Right now they’re in the AHL and NCAA but in the words of Neil Young “comes a time.”

    As for Stortini, he’s exactly the kind of player who will always play right on the edge of a roster. When things go well for him there’s no doubt he’s going to play, but the longer the 4line does little when Zack’s out there the more likely he’s the odd man out.

    He should be good to spend the entire season here in 08-09 though, especially considering JDD is basically not in use.

  43. Bruce says:

    TSN just reported that the 7 points put up by Gilbert, Smid and Souray yesterday represented the first time in 15 years that three Oiler defencemen had multiple-point games.

    And taking it all the OT, did you notice LT that Dwight Helminen scored his first NHL goal against the Leafs yesterday? His goal tied the score 3-3; later Helminen got an assist on the eventual game winner. That’s a nice way to break the ice.

  44. mc79hockey says:

    …statistically Zack was superior in every way, and by a large margin. Now as I see it, MacIntyre = Godard, which is a major step back for this club competitively.

    So far this year MacIntyre has had about half an hour of ice time, during which time the Oilers have generated all of 5 shots on goal. The guy brings no offence to the table, none. His Corsi number is already -15. That’s a guy you just can’t afford to put out there very often, witness his sub-4:00 per game to date.

    The problem I see with all of this is that you shouldn’t be comparing Stortini to MacIntyre – you should be comparing Stortini to MacIntyre + wherever the extra minutes go. The one thing about this team, at least so far, with everyone healthy, is that even the third line has a group of guys who the coach is going to be inclined to find minutes more. As you shrink that fourth line role, the table starts to tilt in MacIntyre’s direction.

    As for the SF/SA thing…I’m still not sold that the credit for the mega improvement rests entirely with Zack. Brodziak and GlenX were legitimate hockey players, good fourth liners, IMO. I know it’s wrong of me, but I have a hard time getting over the fact that the guy looks like he’s wearing snowshoes when he skates. I believe he improved as the year went on, I believe he works hard and has a great attitude…personally though, I’d rather that the coach burned 3:52 a night on MacIntyre and gave those six extra minutes to guys like Moreau, Cole and Nilsson than burned ten on Stortini.

    This goes, in a way, to the prospect thing, which is an issue that dates way back to HF, long before you ventured into this debate Bruce. I don’t think that you’re going to make the argument that Zack is ever going to be anything more than a fourth liner. You might think he’ll be a good fourth liner, I have a hard time envisioning him as anything more than a bad fourth liner. Either way, I have a hard time getting worked about his usage – how much time should be poured down the Stortini hole in order to find out if he can be a fourth liner?

  45. Bruce says:

    how much time should be poured down the Stortini hole in order to find out if he can be a fourth liner?

    Answer: Ignoring for a moment the inference that we didn’t already find that out last season, I would still say “lots”, because every team needs three good fourth-liners. If we can grow them from within, that works for me.

    Zack is himself a prospect, something which seems to be downplayed by you and others. He just turned 23. His improvement during the course of last season was in large part due to better linemates, sure, but he carried his weight. And he fought above his weight class a few times.

    Looked at another way, Stortini’s line — including 9 games with Stoll and Torres, recall — compiled just 52% of the shots of his opposition and 30% of their goals in the first half; to 108% and 225% respectively in the second half … well surely some of that has to be a fluke, but holy shit. I say give the young man a chance.

    I see this guy with third-line upside, btw, but that’s probably two years away. For now he brings lots of positives to the table, including all the physical play that he will deliver that Ales Hemsky and Robert Nilsson won’t in a month of double shifts.

    As for those who want to see Potulny or Schremp in the 12th forward spot, that makes for a pretty soft lineup in my view. I think MacT will dress at least one of the big fellas pretty much every game. ‘course, I been wrong before.

  46. Bank Shot says:

    Those who enjoy the sight of blood might vote otherwise, but to me the most important part is showing up at the dance, and Zack has never let Oilers down in that dept.

    Hemsky doesn’t agree, and has said so on record before this season started.

    Who knows how much influence he has with coach, but its just another thing stacked against Stortini.

  47. Vic Ferrari says:

    I don’t know if Dennis has summed up his EV scoring chances yet, but I’ve been manually doing a tally as he posts them.

    At evens:

    Smid is +19/-20
    Gilbert is +42/-46

    And that’s scoring chances at evens only. And I may have made a mistake in summing, it would be cool if somebody else invested 20 minutes to double check.

    Of course Gilbert has played tougher ice time as well.

    Smid’s playing better, for sure, but slipper and scott are absolutely right imo.

    Sounds like slipper is skimming through gauging the relative levels of Brucelexia amongst the crowd here. That’s odd.

    If Smid really is doing something to cause the goalies to become three times more likely to stop the puck than Hasek in his best year … well I haven’t noticed it. And Dennis’ scoring chance against numbers don’t support the argument either, not even a bit.

  48. Scott says:

    Vic,

    I did a count and I came up with a slightly different number for Gilbert, but it’s near as. Damn counting.

    Gilbert +42 -47

    Grebeshkov +41 -44

    Smid +19 -20

    But yeah, Smid’s not really doing anything to make me think that he’s on fire in terms of scoring chances although games 9 and 10 were better for him (+8 -4) so it seems those that think he’s played better the last few games are right on. Still, the flattering GA/60 is a foundation built on sinking sand.

  49. Oilman says:

    What’s super neat is how goalie performance can influence the perception of defense.

    Late to the party here I know, but what’s also super neat is how good defense can influence the perception of goaltending.

  50. Bruce says:

    Very well said, Oilman. I think the key word here was raised by Doogie, namely “disentanglement” of the goalie’s performance from the defencemen’s, so it’s gonna be a grey area no matter what.

    Scott: Rather obviously, 0.63 is unsustainable. Differences in Sv% have to normalize somewhat from small-number extremes, as I have stated throughout. However, if the assumption is that those differences “should” balance all the way out to league or team average Sv%, well, I just don’t accept that. In my view some defencemen/players will allow a higher quality of shot than others and therefore the expected Sv% behind them should be different.

    That said, your suggestion and follow-up of parsing the differences between GA and SA by measuring actual scoring opportunitiers is reasonable, and would be even better if somehow the quality of those chances was measured (what Vic call 10- and 25-cent chances).

    I’m not sure why challenging of assumptions always seems to meet with such a hostile reaction from some quarters. Especially given said luminaries have made significant advances in our understanding of the game by similarly challenging older assumptions. It’s a crucial part of the scientific method.

    That said, “Brucelexia” made me laugh out loud. Do you mind if I use it?

  51. mc79hockey says:

    In my view some defencemen/players will allow a higher quality of shot than others and therefore the expected Sv% behind them should be different.

    I’m open to agreement on this. I asked this above but, over the course of a million rolls of the dice, what’s the difference between Paul Coffey and Randy Gregg going to shake out to?

  52. mc79hockey says:

    (I’m willing to bet that it’s smaller than the difference between the odds of Randy Gregg successfully performing surgery on you and Kevin Lowe doing so.)

  53. mc79hockey says:

    Fuck, meant to say Coffey the second time. What can I say, I think Lowe is probably a bad surgeon in addition to being a mediocre GM.

  54. Vic Ferrari says:

    Scott,

    Thanks. I can live with missing one here or there, at least it looks like I didn’t miss a whole game or anything.

  55. Oilman says:

    If Smid really is doing something to cause the goalies to become three times more likely to stop the puck than Hasek in his best year … well I haven’t noticed it. And Dennis’ scoring chance against numbers don’t support the argument either, not even a bit.
    That’s not a fair comment right there. I’m sure Hasek also had better numbers with certain D-men on the ice than with others. As I’m sure that Osgood’s SP is better with Lidstrom on the ice. Every goalie in the league has a variation amongst shutdown and high risk defenders I would guess.
    Also, Dennis’ scoring chance numbers don’t give a quality assignment to the chances – which I’m sure you’d agree has an impact on the overall value of the stat. Just counting the chances doesn’t tell you anything about the goaltending.

  56. mc79hockey says:

    I’m sure Hasek also had better numbers with certain D-men on the ice than with others. As I’m sure that Osgood’s SP is better with Lidstrom on the ice. Every goalie in the league has a variation amongst shutdown and high risk defenders I would guess.

    I believe this. I have my doubts that it was .980 with some guys and .820 with others though. What’s a realistic spread?

  57. Vic Ferrari says:

    Oilman:

    I’m with MC79 here. I’m sure there is an effect, I just don’t think it’s very much.

    I mean if you and I played D for the Oilers in the next game I think it would be a gong show of great chances for the Blue Jackets. But the difference between NHL D men just isn’t that much to my eye, not in this regard. A big difference between them in affecting zone time though.

    And the guys that are less risky (say Nick Schultz) are probably going to play more against better shooters and playmakers than the bigger risk takers (say Bergeron) are going to. So that probably levels it out as well.

    We’ll see with Dennis’ scoring chances, though we’re going to have to wait for the sample to grow I suspect. If some guys are consistently on the ice for more scoring chances per 100 shots against … then they’re affecting shot quality, and we’ll have a feel for how much.

    Around Christmas time we can take a look at it.

  58. Dennis says:

    Bruce: You’re welcome;)

    My reasoning for wanting 46 out of sight is the same as Ty’s: when 33′s in the lineup, this means that the 12th forward spot will be open for a real player for the vast majority of the time. 33 can come out for a few skates to let everyone know he’s there and then he can sit down and “rah rah” and we can let 83, 12, 34 etc pick up a few shifts.

    That works fine by me.

  59. Bruce says:

    what’s the difference between Paul Coffey and Randy Gregg going to shake out to?

    MC: Good question, worth asking twice. i meant to reply earlier but missed it.

    If the team Sv% was .890, I would guess it might range from .875 with Coffey on the ice to .900 with Gregg. Maybe more. Coffey took a lot of brutal risks up ice and made some horrendous own-zone plays. His +/- figures indicate that the reward was well worth the risk, but some of it was just silly. Whereas Gregg excelled at eliminating risk and keeping attackers to the outside.

    I would be shocked if the style differences didn’t show up in on-ice Sv% if such data were available. Best we can do though is compare GA data and estimate ice time. In 1985-86 for example, Coffey went +179/-118 excluding PPGF/A, and Gregg went +93/-63. That’s 87% more GA with “Coff” on the ice, and I can’t imagine he got as much as 50% more ES TOI, probably less. (Coffey got a ton of PP time of course, while both were on the PK unit. I chose that particular year since 4v4 play, a huge Coffey factor, was virtually legislated out of the game by then, courtesy Cliff F. Fletcher) And my educated guess would be that the difference between the defensive components of the two would show up much more in OppSh% than any signifncat difference in the rate of actual shots.

    But rather than speculate about old-timers (as I am prone to do when searching for examples), we’d likely be better served by choosing modern teammates with major style differences, like say Mike Green and Shaone Morrison. Or even Denis Grebeshkov and Ladi Smid. :)

    the guys that are less risky (say Nick Schultz) are probably going to play more against better shooters and playmakers than the bigger risk takers (say Bergeron) are going to. So that probably levels it out as well.

    A very legitimate assumption.

  60. Oilman says:

    I’d guess that a realistic spread for a goaltender who posts a .920SP would be plus/minus .03 to .04….but that’s a guess, nothing more.

  61. Bruce says:

    Forgot to add, the corollary being that the Oilers shooting percentage, which was a sick 16.1% in 1985-86, would have been its absolute sickest with Coffey on the ice. He was the ultimate high-event player, lots of ten-bell chances at both ends when he was out there, but with Fuhr and Moog behind him and Gretzky and Kurri, Messier and Anderson in front of him, the percentages were very definitely in paul’s favour either way. Trading opporunities was a winning proposition for that gang, and Coffey embodied that perhaps even more than Gretzky.

  62. Vic Ferrari says:

    Oilman:

    Did you mean .003 to .004? So a better defender should expect his goalie to stop an extra shot every 300 or so, due to his effect on limiting the quality of the chances?

    That feels about right to me, I was just under the impression that you and Bruce were thinking bigger numbers.

  63. Sean says:

    I think for the extreme case Coffey vs Greg, that Bruces numbers are accurate of if anything would be an under estimate. If you face 1000 shots with Coffey on the ice, how many of those would be 2 on 1s alone?

    For a typical case, say Staois vs Grebeshkov, my guess is about .04 to .05 difference.

  64. Alice says:

    Couple things: the Dmen come in pairs, so that would (assuming the coaches try to balance strengths and weaknesses) tend to dilute the per-defender swing effect. Unless the GM wants two 5-alarm guys out together making highlight reels to sell some seats. Of course that would never happen.

    Re SMac, Dennis said “and we can let 83, 12, 34 etc pick up a few shifts.”
    Right idea, but a monster like Cole is going to be harder on the opposition – if softening them up is the goal. Hemsky might make you look silly, but he won’t wear you the hell out. The 4L guys should leave some bruising impressions on the opp D.

  65. Bruce says:

    Did you mean .003 to .004? So a better defender should expect his goalie to stop an extra shot every 300 or so, due to his effect on limiting the quality of the chances? That feels about right to me, I was just under the impression that you and Bruce were thinking bigger numbers.

    I’m definitely thinking bigger numbers, way more than 1 in 300. I’m thinking a range of a couple to a few percent, not tenths of a percent. I think Oilman didn’t say “.030 to .040″ becuase the extra significant digit implies a degree of accuracy in the estimate which is absolutely unwarranted. My own guess would have been closer to +/- .02, but with outliers.

    And yes, Sean, my speculative Coffey vs. Gregg Sv% differential was very conservative. I didn’t want to overstate it, but it would be pretty large. Not as large as the difference in Sh% while the two were out there, but large. Surely a few percent.

  66. PerformanceOil says:

    Bruce,

    I too am puzzled by MacT’s lack of love for Stortini this season. Last season, MacT showed an incredible amount of patience with Storts, which seemed to pay off. This season, Stortini has had almost no rope, and we seem big Mac drawing in game after game. It’s pretty clear a paradigm shift has been made where more value is being put on having a legitimate heavyweight, versus a middle-weight agitator.

    This seems a bit off to me, since regardless of how good of a player Stortini really is (and like you mention, he probably hasn’t peaked as a player yet), he was a quality agitator. He filled a role, and that role certainly had value last season, not sure what changed since then. Then again, who knows what is going on behind the scenes. Maybe Stortini is suffering from an acute case of Big George syndrome, and forgot what got him to the show. Maybe (and I think someone else suggested this, couldn’t find it though, apologies) big Mac is being used to fire a few warning shots across the NW bow to start the season off. Once the message is sent, we might see less of the enforcer and more of the pest.

    Who knows, I for one hoped to see Stortini pick up where he left off, but for whatever reason, it wasn’t meant to be. If he was for real though, he’ll find a way back in sooner or later.

  67. PerformanceOil says:

    Speaking of weird MAcT coaching decisions, I don’t think the Pisani at centre experiment will end until the actual results start resembling the underlying results that apparently look so bad. Fact is, on paper, Pisani + Moreau + whoever have been our most consistent line this season (according to the conventional numbers). Whether the results are sustainable or not remains to be seen, but so far they
    have done more to win games for the Oilers than lose them. ‘Course, you could easily go Moreau-X-Pisani, but then where do you put Cogliano/Cole?

    Probably what makes the most sense would be to trade Pisani the RW for some guy just like Pisani who plays Centre. Assuming that isn’t available at this time, I guess making Pisani a C makes sense within a limited view.

  68. PerformanceOil says:

    The great thing about Smid this season is he actually seems to be using some of his plus skills more consistently (in a very short stretch, granted).

    He’s always been a good skater and above average stick handler (and I’ve always though he was a good passer, at least technically, even if his decision making was suspect).

    So far, he is moving the puck with more authority, either via his skating or some solid passing, and I like the way he has started adding an offensive dimension to his game. His shot actually appears credible this year, and I think he has been supporting the offense a bit better than last year.

    Overall, he looks someone who has finally started figuring out that he can be a pretty good player, if let’s himself be. Hope it lasts.

    As far as SV% – obviously it won’t sustain at its current level, however that said (and as Slipper implied), that doesn’t mean his ‘real’ EV SV% should be at the level of the Oilers’ team average or below (though that certainly might be true). The current level may well be a bit of puck luck combined with a lot of doing good things. Even if it comes crashing to Earth, it doesn’t prove that he wasn’t playing great hockey, it could simply be that he couldn’t sustain the great hockey.

  69. Dennis says:

    There have been games where the 34 line has been absolutely fucking owned at evens. It’s true they had two goals apiece in wins over Col and Cgy but I just don’t understand how MacT was pissed that they were letting chances develop from losing own zone faceoffs in the EX season and now he’s just gonna let it pass in the reg season???

    And, yes, 26 could get some time in the 33 slot as well. I remember someone, I think it was Bruce, saying how 26 was one of the top guys at drawing penalties last year and he’s lived up to that rep thus far in ’08.

    BTW, what are the odds that Reasoner winds up back in Edm before the season’s over? I’m thinking it HAS to be better than average. MacT loves the guy and he was super well-liked by the boys so he could come back and be like Rem Murray in ’06. There’s no way 33′s gonna play if we make the playoffs and it would be nice to have Marty back taking draws and shoring up the thus far awful PK.

    Also, in looking at last year’s PK numbers, why in the holy hell isn’t Cogs killing penalties?? If Mact’s trotting out 89 because he feels it will increase the chances of SH opps, allow me to remind him Cogs blew everyone away last year in a small segment sample; 4/5 was his actual number with a 3.68 GFON and a 4.60 GAON. Granted, that wouldn’t hold up over a season but it’s a fine fucking start, no?

  70. PerformanceOil says:

    Dennis,

    I often wanted someone (other than me!) to look at scoring chances for/against, so good job there. Do you have your methodology detailed anywhere? I.e. how do you decide what a scoring chance is?

    I know shot distance has been used previously, but I don’t think scoring chances can be broken down that way very well (though certainly better than nothing). Screens, tips, passing plays etc. all could heavily influence the probability of a given shot going in against an ‘average’ goaltender. And this neglects the fact that real-life goaltenders are probably quite variable in what constitutes a difficult save and an easy one.

  71. PerformanceOil says:

    Also, in looking at last year’s PK numbers, why in the holy hell isn’t Cogs killing penalties?

    Yeah that one is a real head scratcher. Seems like MacT is still in development mode a little bit, and is trying to accelerate the ‘Gagner becomes complete player’ storyline.

    To me, it seems like Coach does these kind of weird things every year. I think he likes to challenge the players and take them out of their comfort zone a bit. Eventually, he comes back to what everyone thinks is obvious and the Oilers start putting up results. It’s tough to watch in the early going, but maybe it’ll pay dividends when it really matters (in the playoffs a year or two from now).

  72. Bruce says:

    I know it’s wrong of me, but I have a hard time getting over the fact that the guy looks like he’s wearing snowshoes when he skates.

    It’s not wrong of you, Tyler, you’re certainly entitled to your opinion, but it’s pretty inconsistent. You’re the numbers guy, but when I throw some pretty impressive numbers at you about a guy you don’t happen to like, it gets reduced to “the guy looks like he’s wearing snowshoes when he skates”. Not too scientific.

    Obviously skating is hugely important and you have to trust what your eyes see to a large extent. I will however, point out that guys like Harold Snepsts and Brad Marsh succeeded in this league despite apparently forgetting to take their skate guards off. George Armstrong used to get around on willpower alone or so it seemed. Moreover, there are many big men who skate faster than they look; Phil Esposito looked like a plodding skater for goodness sake but he got from A to B a split second faster than it appeared he “should” and he had a way of making that split second count.

    Zack is not that kind of player of course, but if you want to judge his skating watch him on the forecheck esp. after a shoot-in, he wins more races than one might have any reason to expect based on appearances, sometimes cuz the other guy is throwing a little snow and sometimes cuz he’s got momentum on his side and just wins the race. And watch him bust his ass back to his own zone and pick up the trailer on the 3-on-3. He may look like a sasquatch in snowshoes, but he gets there. His biggest issue is first step and acceleration, and that has improved by leaps and bounds in the past two years. Hard to imagine he won’t keep improving for years to come, esp. given his work ethic.

  73. Bruce says:

    To me, it seems like Coach does these kind of weird things every year. I think he likes to challenge the players and take them out of their comfort zone a bit.

    PO: That has the ring of truth to it. They actually teach that technique in adult education courses, that folks outside their comfort zone are better positioned to learn. (Within reason.)

    Personally I’d have Cogs on the PK in a New York minute. I love speed on the PK, it’s a huge weapon in all three zones, with and without the puck. Cogs is more likely to get a wide-open one-on-one on the PK than he is at even strength, and he might even get it against a forward.

    As an old-school Oilers fan, I like absolutely nothing better than shorthanded offence, so I don’t oppose the Gagner experiment either, but it’s going to come at a cost. Hell, I’d throw Hemsky out there on the PK occasionally just for fun.

  74. PerformanceOil says:

    PO: That has the ring of truth to it. They actually teach that technique in adult education courses, that folks outside their comfort zone are better positioned to learn. (Within reason.)

    Makes sense Bruce. I’ve often wondered if another aspect is that once the players get used to hauling big ass anchors around the ice, you cut them loose and watch them soar. Combine that with the first and you would get the whole ‘see, listen to me and look what happens’ synergistic effect between the learning and the potential morale boost.

    I hope MacT writes about all this crazy stuff in a book someday, ’cause I really want to know what goes on in that guy’s mind.

    I think your point about speed is a great one. It is something most good PKers have (by my perception at least). That is why I can’t see Gagner being a SH point producer. He doesn’t have the speed, and great passing/vision probably isn’t what will get the job done (offensively) most times. Hemsky though, would be extremely dangerous, on paper at least.

  75. Dennis says:

    I could buy the whole “developing for later” thing if he were using Both 89-13 on the PK but the latter was on for 4 SHGF in a short segment sample in ’08 so if it’s offense you’re looking for, That’s the guy who we should be ‘wasting’ time on.

    And it’s frustrating for me because we’ve missed the playoffs three of the last four seasons and I think that playing 89 on the PK instead of 13 is a short-sighted move that shouldn’t be tried when you’re not exactly playing with house money. And the trickledown is that it’s better to play guys like 51-78-36 to round out the 34-18-10 troika and then we save the kids for the soft first-shift-after-the-PK action.

    In that regard it doesn’t make sense to play either of the kids but if you’re going to, the numbers say 13 is the one who gets the push. But MacT’s got it in his head that 89 will be developed at the expense of the ’09 season — at least early on — and recent history shows he’ll keep fucking around with this for a little while yet at the very least.

    Regarding what constitutes a scoring chance, I guess I’ll have to get to it sometime but unless things look really out of whack, I don’t think there’s an urgency to wholly explain it. I guess the best way to blanket it is I look at plays that appear to be threatening and I’m always wary of awarding players and punishing players; if there’s a screen on the Oilers goalie and a shot gets through then I’m counting a scoring chance for the opp to reflect the fact that our D isn’t doing their job covering the front of the net.

    The entire how-and-why isn’t probably something I’ll ever get to; in the end the hardcores will take a look and see if it passes the smell test and they can decide from there.

  76. Oilman says:

    Oilman:

    Did you mean .003 to .004? So a better defender should expect his goalie to stop an extra shot every 300 or so, due to his effect on limiting the quality of the chances?

    That feels about right to me, I was just under the impression that you and Bruce were thinking bigger numbers.

    Vic, I think you underestimate the bigness of my thinking! I think that you’re off by a factor of 10 – or more likely, we’re both way off and the number is somewhere in the middle. I figure by your numbers, 1 stop in 300 shots means 1 extra goal against (for most defenders) every 20 to 25 games and that would equate to a SP difference of say .919 with one D man on and .922 with another – that’s negligible really. I’m thinking more along the lines of 1 extra GA every 3 to 5 games – so instead of stopping 55 of 60 shots, the goalie stops 54 and the SP with this defender X on the ice becomes .900 instead of .920. Then Lidstrom jumps over the boards and the goalie stops 56 of 60 shots while he’s on and the SP is .933 – giving the .03 swing. I don’t know if that’s realistic or not – it’s a swing of about 30 goals against over the course of a season – but I think it sounds fair to think that elite d-men can reduce GA by 15 over the course of a season.

  77. HBomb says:

    BTW, what are the odds that Reasoner winds up back in Edm before the season’s over? I’m thinking it HAS to be better than average. MacT loves the guy and he was super well-liked by the boys so he could come back and be like Rem Murray in ’06. There’s no way 33′s gonna play if we make the playoffs and it would be nice to have Marty back taking draws and shoring up the thus far awful PK.

    I sure to hell hope this doesn’t play out. If they’re going to go get a center, he needs to be capable of 3rd line minutes.

    Marty Reasoner is a 4th line center who can win some draws. If they’re going to add a guy for the 3rd line, they need to either do better than this or give the minutes to either Pouliot or Brodziak.

    Reasoner’s time here has come and gone. Time to move on, people.

  78. Quain says:

    I thought Vic was understating it a bit too, but I buy that a defender doesn’t influence SV% that much (we’re talking about average versus above/below average, not 16 year old kid versus Bobby Orr), what he influences is how many shots get fired at your guy.

    Obviously a guy like Bergeron is going to decrease SV% a bit more, given his tendancy to whiff on passes and give up breakaways, but the difference between an average defender (Steve Staios, say) and Nik Lidstrom is that Lidstrom is going to get the puck and get it out of the zone whereas Staios is going to limit the Grade A scoring chances decently enough, but when he gets the puck his clearing attempt is going to fail because all he knows how to do is ring it around the boards.

  79. Vic Ferrari says:

    Well, Lidstrom just has a .898 EVsave% behind him so far this year. Good underlying numbers though.

    I’m sure that if you go back over the seasons where Lidstrom wasn’t a serious contender for the Norris, that he had subpar on-ice EVsave%’s behind him. Because really, the guy should have won ten in a row by now.

    The thing to do, Oilman, is to pick 15 defensemen (one off of each team in the west) that you think are the best guys at limiting shot quality. Or 30 guys if you watch the east enough to have strong opinions.

    Then we can bookmark this thread, and starting from this point forward (game 20170 or whatever we are at) we can see if your list of 15 or 30 guys outperforms the team EVsave% over the remainder of the season.

    I’m guessing that your list of defenders will average about a .002 better EVsave% behind them than the western conference as a whole. We’ll see.

  80. Oilman says:

    What if I pick the wrong 15?:o)

  81. Scott says:

    Vic,

    Where do you get the EV SV% numbers from?

  82. PerformanceOil says:

    I could buy the whole “developing for later” thing if he were using Both 89-13 on the PK but the latter was on for 4 SHGF in a short segment sample in ’08 so if it’s offense you’re looking for, That’s the guy who we should be ‘wasting’ time on.

    Unless MAcT thinks Cogliano already ‘gets it’ but isn’t good enough to supplant the better options. Maybe having two developmental slots is all we can afford (Gagner + Penner), and of course some would argue that’s two slots too many.

    I agree with you re: the playoffs, though what has happened the last few seasons has no impact on what is the best moving forward. If putting Gagner on the PK means a cup in a couple of years, at the expense of the playoffs this year, that’s a decent enough yield, no?

    That said, I don’t know that I support Gagner on the PK as such (or believe that it will have such a dramatic effect), I’m simply trying to put reason to MacT’s seemingly erratic behavior. Could be the guy takes a month or so to remember how to coach each season, but that seems unlikely. He has a purpose here, but like Picasso, it might not be obvious upon casual inspection. Whether he is genius or a kid painting randomly, remains to be seen (and is subject to constant re-evaluation).

    The entire how-and-why isn’t probably something I’ll ever get to;

    I applaud the effort (I want to emphasize that, I really think it is a worth-while thing to do), but there seems to be a double-standard here. Staples’ error stat is probably no more subjective and he has been rigorous and transparent in its application. Yet, it is (near) universally derided here and elsewhere. On the other hand, your scoring chances which are vague and murky seem to be equally widely accepted and are used to support arguments in this thread. Something doesn’t compute here.

    in the end the hardcores will take a look and see if it passes the smell test and they can decide from there.

    So if it says what the consensus thinks it should say, it is good? I think I understand the disconnect now.

  83. Doogie2K says:

    Marty Reasoner is a 4th line center who can win some draws. If they’re going to add a guy for the 3rd line, they need to either do better than this or give the minutes to either Pouliot or Brodziak.

    Isn’t that the idea, though? Put Marty Sakic on the 4th line and bump Poo or Brodziak to the 3rd? Still means we have too many top-niners without putting half the team out of position, but that’s not a horrible problem to have.

    Where do you get the EV SV% numbers from?

    http://www.timeonice.com/playershots.php?team=EDM&first=20001&last=21230

    Replace the team with whichever one you like; "first" and "last" games are NHL game numbers (2 = regular season; 0001-1230 = games 1-1230, by the NHL schedule).

    @PerformanceOil: I'm not sure if you really want to open this can of worms at just this point, but you're welcome to it, with my sincerest sympathy.

  84. PerformanceOil says:

    @PerformanceOil: I’m not sure if you really want to open this can of worms at just this point?, but you’re welcome to it, with my sincerest sympathy.

    I’m not sure exactly what you mean by this (did something happen? are you just trying to tell me to shut the hell up?), but asking questions and providing criticism is part of the process. I think doing this is valuable, though others may disagree *shrug*.

  85. Dennis says:

    If the whole “can of worms” thing was directed towards me, I really don’t think there’s a need to frame it in that respect.

    I’m glad that some guys haven taken my “stats” to heart and are quoting them and I’ll say that maybe it’s because they’ve read enough of my opinions over the years to know I’m a hard and fair marker; and in Vic’s case, he has some other numbers that he says are jiving well with mine.

    And, hey, no one has been more dismissive towards the error stat than myself and the only thing that I’ll say about the difference, at least how I see it, is that I’m logging info and players on ice for Every scoring chance and not just the ones that end up in the net.

  86. Bruce says:

    Replace the team with whichever one you like

    Also, you have to know the secret handshake. The code for Tampa Bay for instance, isn’t TB or TAM or TBL, it’s T.B (one period, in the middle). I still don’t know the code for a few teams. I know there’s a ton of good info on that site, but between having to remember all the statistical categories and all the team codes, I know I’m missing out, only get fragments here and there. It can be frustrating.

    Hopefully some day there will be an interface on the site so that it’s a simple matter of point and click. It’s a gold mine of great data that Vic went to a lot of work to compile, and I don’t imagine anybody’s getting rich off it, so I don’t want to come across as a whiner. Suffice to say accessibility is an issue.

    asking questions and providing criticism is part of the process. I think doing this is valuable

    PO: So do I. Just be prepared for the inevitable slings and arrows.

  87. mc79hockey says:

    I applaud the effort (I want to emphasize that, I really think it is a worth-while thing to do), but there seems to be a double-standard here. Staples’ error stat is probably no more subjective and he has been rigorous and transparent in its application.

    Just a comment here – I think that there’s a significant difference between the two things. The problem that I have with what Staples is doing is that an error tm only happens when a goal happens. If David was tracking all of the errors made in the course of a hockey game (likely an impossible task), I don’t think that there’d be nearly as much as objection to the stat, even if it wasn’t precisely defined. You watch enough hockey, you get a sense of what the errors are. David’s obviously watched a ton of it and I don’t doubt that he can pick them out; the problem (as I see it) is with only picking them out on goals and taking that to be signficant. I don’t think that that’s analogous to what Dennis is doing, even if it’s not precisely defined. Dennis is essentially acting as a judge at a boxing match, scoring the bout. David is counting the mistakes that each boxer makes, but only when those mistakes allow the other fellow knocks him down. The difference seems obvious to me.

    Also, you have to know the secret handshake. The code for Tampa Bay for instance, isn’t TB or TAM or TBL, it’s T.B (one period, in the middle). I still don’t know the code for a few teams. I know there’s a ton of good info on that site, but between having to remember all the statistical categories and all the team codes, I know I’m missing out, only get fragments here and there. It can be frustrating.

    Maybe this is just because I’ve spent a ton of time looking at the various stat outputs from the NHL and I know the various abbreviations that they use, but “secret handshake” seems like a bit much. It’s not like the code for T.B was S7Z or something. The harder thing is knowing the various php scripts that he’s got there and what’s available.

    In any event that’s an easy enough problem to fix. If Vic’s got no objection, I’ll take a stab at throwing together an interface page at my place that people can use.

  88. PerformanceOil says:

    Just a comment here – I think that there’s a significant difference between the two things. The problem that I have with what Staples is doing is that an error tm only happens when a goal happens.

    True enough, but is it better to do a mediocre stat well, or a good stat poorly (not saying Dennis is doing a bad job, just that I have no idea, and neither does anyone else)?

    That said, scoring chances don’t encapsulate everything good/bad a player does either. A player could be doing lots of things right but not get any scoring chances for. The assumption is that these things even out, but that assumption should be equally valid for the error as well (since if a player is making tonnes of mistakes, over time, a fair share of them should end up in the back of the net, no?). The only difference is that you’re working with smaller numbers with the error, so it will take longer to generate a significant sample.

    Is that enough to summarily reject one, while embracing the other? I say no.

  89. Bruce says:

    “secret handshake” seems like a bit much.

    MC: Figure of speech. I’m not a code guy, so it causes me no little grief. e.g. I am trying to compare Mike Green v. Shaone Morrisonn so I enter WAS and WSH and get nothing, and now what? A technoflub like me is apt to just throw up his hands and miss out.

    I spend a lot of time looking at stats breakdowns and event summaries at NHL.com, but not much of it looking at the URL strings. When I do I always see game numbers and the same two teams (Home and Away), and the team stats pages seem nondescript in terms of identifying teams. Obviously I’m not looking in the right places. All I know for sure is that they do not use the same three-letter team codes that are listed in stats output (e.g. N.J not NJD, and ??? not WSH)

    The harder thing is knowing the various php scripts that he’s got there and what’s available.

    Agreed.

    In any event that’s an easy enough problem to fix. If Vic’s got no objection, I’ll take a stab at throwing together an interface page at my place that people can use.

    That would be terrific, and I hope Vic agrees. As I said it’s a gold mine of data, it would be great to be able to get at it more directly.

  90. PerformanceOil says:

    PO: So do I. Just be prepared for the inevitable slings and arrows.

    When I roll on down to this end of town, I always put my helmet and cup on first :)

    Agreed about Vic’s site, and if you were to improve it’s accessibility, you would be doing a great service to statisticians, Brucelexists and lunatics everywhere MC.

  91. Bruce says:

    The assumption is that these things even out, but that assumption should be equally valid for the error as well (since if a player is making tonnes of mistakes, over time, a fair share of them should end up in the back of the net, no?). The only difference is that you’re working with smaller numbers with the error, so it will take longer to generate a significant sample.

    PO: Entirely agreed, esp. with the parenthetical remark. Error stats for an individual game are almost anecdotal, but over the course of a season they should start to be meaningful. As Staples himself has commented (frequently), it’s like points, a guy can play a great offensive game and have fuck-all to show for it (I think that’s what David said :), and another night he can make a few routine plays in his own zone and wind up with three assists. So points too need a significant sample size to generate meaning. Which is a very long way from saying they are meaningless cuz they only measure those scoring opportunities that wind up as actual goals.

    That said I am all in favour of tracking scoring chances, it’s the missing link. While the ideal method would have rigorous and transparent methodology, esp. of differentiating quality of chances, what Dennis is doing is an important step and I look forward to following his work.

  92. Vic Ferrari says:

    Oilman,

    That’s the thing though, the fan boards are chock-a-block with folks that have mastered hindsight. I’m looking for the guy that’s making a decent guesses at what’s going to happen next.

    And I think this is fair, no? Nobody is going to ridicule, here, Oilman. I think you will do as well as anybody at picking these guys, I just don’t think that there is much in it.

    And back on topic:
    Smid and Gilbert, by scoring chances as shown above, up to and including the CAR game:

    Smid
    SC+ 19, SC- 20
    Corsi+ 69, Corsi- 74
    EVsave% behind him: .977

    Gilbert
    SC+ 42 SC- 47
    Corsi+ 144 Corsi- 151
    EVsave% behind him: .883

    And I know that that the tandem of 77/37 has had some adventures. I’m rolling my eyes at Grebeshkov, other people pin most of the blame on Gilbert, time will tell which is the better player.

    In any case, if the saves had happened behind Gilbert, and not Smid, the conversation in this thread doesn’t happen.

    The scoring chance +’s and -s are aligning themselves really well with the shots metrics at this point. Gagner seems to be the guy having the largest impact on shot quality against the Oiler goalies … which doesn’t make sense to me, I’m sure it’s just coincidence. And besides, they’re all bunched closely together already.

    Grebs is the worst though, the ratio of SC- to shots- is 0.50 so far for the Oilers if my math is right (feel free to check yourself). And Grebeshkov, is at 0.66.

    The underlying numbers are giving you a good idea of the territorial advantage, how much you’ve outplayed the other guys, and that leads to good scoring chance numbers eventually. And that in turn, leads to goals eventually.

    I mean right now the territorial numbers for MacIntyre are damning, but he’s +3, -3 in terms of scoring chances. Me, I think he’s looked pretty bad out there, and that the scoring chances are going to gravitate towards the way the team has played when he’s on the ice.

    At some point, he’ll have a stretch where every mistake he makes turns into a scoring chance, and if a couple or three of those go in during that span … the AHL is probably the next stop. Tough gig for that guy, I’m cheering for him though, I hope the hockey gods stay on his side.

  93. mc79hockey says:

    I’m not a code guy, so it causes me no little grief. e.g. I am trying to compare Mike Green v. Shaone Morrisonn so I enter WAS and WSH and get nothing, and now what? A technoflub like me is apt to just throw up his hands and miss out.

    WSH works for me.

  94. PerformanceOil says:

    That said I am all in favour of tracking scoring chances, it’s the missing link. While the ideal method would have rigorous and transparent methodology, esp. of differentiating quality of chances, what Dennis is doing is an important step and I look forward to following his work.

    Agreed entirely, it’s just when I perceive hypocrisy, it bugs me a bit (you’ve probably noticed). Dennis (and Staples!) should be commended for doing what others are too busy (and/or lazy) to do.

    Which is a very long way from saying they are meaningless cuz they only measure those scoring opportunities that wind up as actual goals.

    Typical Brucelexist thinking! That line of thought isn’t gonna win you any money in Vegas.

  95. Vic Ferrari says:

    bruce

    http://timeonice.com/teamlist.html

    I put that up there and gave you the link last time you complained, no? Bookmark it, dude.

    And WSH is the official NHL abbreviation, you must have something else typed in wrong, Bruce.

  96. Vic Ferrari says:

    mc79hockey:

    Yeah, go ahead.

    Right now it’s kind of nice, just 300 visitors a day or so there, so when I check the links, I generally find great some stuff written about hockey. There aren’t many wacky kids capable of figuring out how to use it, apparently.

    But I don’t think that an easier platform will make much difference to traffic, typically if timeonice.com stuff is linked in a post at HFboards, then if there are about 1000 reads and 40 comments, there will probably be only 3 or 4 people who clicked through to the information.

    So I’ve solved the mystery of why so many comments on HF threads seem detached from the subject being advanced and unrelated to the information being used to support the argument. :D

  97. Doogie2K says:

    If the whole “can of worms” thing was directed towards me, I really don’t think there’s a need to frame it in that respect.

    It’s more about the general response to criticism and/or lack of understanding in the ‘sphere. We’ve had this argument two or three times before, and I even suggested last time pointing to the guy who fucked up on any given scoring chance, but was met with hostility. Dunno if that was emotions talking, or some further flaw with the concept that wasn’t adequately discussed, but I’m not really married to the idea one way or the other, so…yeah.

    Gagner seems to be the guy having the largest impact on shot quality against the Oiler goalies … which doesn’t make sense to me, I’m sure it’s just coincidence.

    Small sample size, but he certainly hasn’t looked like a liability in the games I’ve seen so far. How much of it is a function of the ice he’s getting? He and Cogs are tied with 23 more O-zone faceoffs than D-zone, while there’s not much dropoff in where they end (19 more O than D).

  98. Bruce says:

    WSH works for me.

    … and for me, now. Earlier I twice typed WSH instead of EDM on a working link and got an instant fail. This time it worked. Go figure … such are the trials of being a technoflub.

    I put that up there and gave you the link last time you complained, no?

    If you did, I must have missed it.

    Bookmark it, dude.

    Done.

  99. PerformanceOil says:

    and I even suggested last time pointing to the guy who fucked up on any given scoring chance, but was met with hostility.

    Oh, you’re another of the crazies then! I understand your comment now, all too well indeed.

  100. Vic Ferrari says:

    P.O.:

    It seems intuitively obvious that Dennis is measuring quality of play (granted, context of ice time has a huge impact on that). And Staples is mostly measuring luck, because he’s just working backwards from when they happened to go in the net.

    If we embrace the fact that hockey, at 5v5 at least, is a simple game … then that’s pretty clear I think.

    So we can look at Xmas, should be enough goals by then, and Staples stuff, if he compiles the totals somewhere, will mesh really well with the on-ice EVsave%’s. And Dennis’ scoring chances with the shots metrics and zone stuff.

    And Dennis’ stuff should be giving us a really good idea of what to expect from the team rolling forward.

  101. Vic Ferrari says:

    Doogie:

    I haven’t seen anything to suggest that Gagner is limiting the quality of shots against. Sometimes you outplay the bad guys and end up getting outchanced as well (like Game 2 of the SCF in Carolina). We all know it happens. Sometimes your line struggles but you get away with it, the other team just doesn’t quite convert good play into scoring chances on the night. And the shooters miss the chances that they do get.

    Same with MacIntyre, there have been just four shots on goal by the Oilers when he’s out there, and Dennis has him at +3 for scoring chances. Strudwick’s up there too. And I don’t think either is doing much to turn possession into chances Gretzky-style, they haven’t played much, and it’s gone their way so far. Good for them, but I don’t think that even their parents are expecting that to carry on.

    As you say, it’s early. For the most part the dust seems to be settling on that stuff already though.

  102. PerformanceOil says:

    It seems intuitively obvious that Dennis is measuring quality of play (granted, context of ice time has a huge impact on that). And Staples is mostly measuring luck, because he’s just working backwards from when they happened to go in the net.

    I prefer objectively obvious, rather than intuitively, myself. They are both defining an event (scoring chance or goal) and then determining who was responsible (everyone on the ice for scoring chance I guess? vs. whomever the eye of Staplon casts its un-lidded gaze upon). You just happen to like the fact that Dennis’ scoring chances are tracking shot differential. Staples’ error happens to track EV GD very well. Both make sense to a degree, but neither is necessarily true based on the methodologies (though it really depends on what constitutes a scoring chance). In neither case does that mean the information gained is redundant either.

    I agree with you that if the Oilers are truly out-chancing their opposition (and the chances are of equal quality), the points will start coming (both individually and for the team), assuming the goalies hold up their end of things. That said, looking at errors can give us an idea of players defensive worth, at least partially independently from their offensive ability. This does matter of course, because the NHL is a league where roles are important, and will probably become more so as the cap world continues (and especially if it shrinks).

    Besides that, the error (again theoretically) is a stat at the level of the player, while scoring chances measure whole lines (or you could do individual players as well, but that would probably be unfair). This means changes in error rates can determine if a player is actually improving, without having to factor in quality of teammates (at least to quite the same degree).

    Bottom line is they both have potential worth, just as they both have potential pitfalls, and picking one because it tracks (insert your favorite stat or conventional meme here), is biased to the extreme.

  103. Oilman says:

    Vic,

    I think this is where we differ on the argument. You’re looking for repeatability, predictability, and I’m not. I’m just saying that for some reason, in this stretch of games, the goalies with Smid on the ice are stopping 98 of 100. The question is, is it because of Smid’s play or in spite of it. Stat guys will say that Smid is the beneficiary of good SP without allowing for the fact the SP is actually the result of Smid’s good play. I’m sure there will be stretches this season where he’s playing poorly and the SP behind him will go down.

    And we’re also talking different comparisons as well. You are wanting to show that a D-mans supporting SP throughout the season would be within a certain variance of the avg SP throughout the league. I’d rather compare the SP’s top to bottom on the team roster – I think that’s where the .03 difference would be.

    What if I go through rosters and give you which D I’d predict would have the best supporting SP on his team and who would have the worst – based on who I feel is better at limiting high quality chances?

  104. Vic Ferrari says:

    Oilman:

    That’s fine, sounds fair to me.

  105. PerformanceOil says:

    stat guys will say that Smid is the beneficiary of good SP without allowing for the fact the SP is actually the result of Smid’s good play.

    Just the sane ones. BTW, you’re talking so crazy you are actually starting to make sense to me!

  106. Doogie2K says:

    I haven’t seen anything to suggest that Gagner is limiting the quality of shots against. Sometimes you outplay the bad guys and end up getting outchanced as well (like Game 2 of the SCF in Carolina). We all know it happens. Sometimes your line struggles but you get away with it, the other team just doesn’t quite convert good play into scoring chances on the night. And the shooters miss the chances that they do get.

    Fair enough. Still, he’s mostly looked good in the games I’ve seen, which granted hasn’t been all of them (pay per view, rec soccer, and Hitmen games have cost me four of the eleven games already). I’m happy with the kid so far.

  107. Oilman says:

    Vic – I’ll work on it when I get some time and post it over at 2mins.

    PO – happens alot. In fact, once I start making sense to others, it all sounds like jibberish to me;o)

  108. Vic Ferrari says:

    Doogie:

    I think that Gagner has looked terrific so far. I’m just saying that the reason he’s got such gaudy scoring chance +/- numbers is mostly because his line has had the puck, much of the time, in the good end of the rink.

    And he’s flattered a bit as well because it looks like, on aggregate, the other guys haven’t been able to turn the possession that have had into scoring chances, not as much as we should expect. Stuff happens.

    The real oddity on the team is Cole; Corsi +/- of +1 and scoring chance +/- of -11. Go figure.

    I think that Cole has been good though, he was unfortunate to be playing with Hemmer and Horcoff when they were searching for their games a bit early. And the winger on Pisani’s line is a tough gig. Plus it seems like every time he is going to get a chance to score he gets hauled down lately. I’m not worried about him, anyways.

  109. Vic Ferrari says:

    Oilman:

    So 20 scoring chances against when Smid is on the ice, and one goes in, and that’s all because of his play?

    And when the worm turns on this (and it will) and if the other team buries three chances, and garon whiffs on one that wasn’t even a scoring chance … and only 7 scoring chances against have past, on 15 shots.

    Will Smid have become the worst defenseman on the team over this stretch? Man, I just don’t get that. For the life of me, I don’t see what Smid has been doing extra special to have been on the ice for only 1 goal against at evens.

  110. Bruce says:

    So we can look at Xmas, should be enough goals by then, and Staples stuff, if he compiles the totals somewhere, will mesh really well with the on-ice EVsave%’s. And Dennis’ scoring chances with the shots metrics and zone stuff.

    So bookmarked. :) It’s still apples and oranges, of course: like Shots or Corsi numbers or any counting stat, all Scoring Opportunities are created equal. Whereas Errors will largely isolate high-quality scoring opportunities allowed for the very reason that the plays all result in goals. As a group, scoring plays are surely not “average scoring opportunities”.

    I think this is where we differ on the argument. You’re looking for repeatability, predictability, and I’m not.

    As becoming an old fart living in the past, I am looking for analysis of the games that have been played rather than the ones that haven't. As for crystal ball stuff, I don't play $port$elect, so maybe it doesn't matter as much to me as to some. As the old adage goes, "that's why they play the games".

    Of course it's fun to speculate on who might develop into what type of player, and statistically that can be done to some effect by projecting development curves on historical examples (per Desjardins) or finding comparables. The trouble with the New Statistics — no matter how advanced, or inspired — is that the history is short, often garbled by evolutionary processes, and comparables are mighty thin on the ground. What was Nicklas Lidstrom's Sv% ON the last time he didn't win the Norris? I'd be shocked if anybody can tell me.

    Let alone the fact that we can't agree on how to interpret that data even if we had it. Say Lidstrom really did have a low Sv% ON that season. Would it mean that his defensive record was let down by poor goaltending, or that he had more defensive lapses leading to higher quality scoring opportunities? No doubt such a discussion would be split along "party lines" with Tyler, Vic, Slipper and Dennis on one side of the debate and Oilman, PerformanceOil, Doogie2K and me on the other … because ultimately it comes down to interpretation and opinion, and there is a fundamental philosophical divide betweeen those groups. Which maybe explains why it sounds like Question Period around here some days. :)

    A simplified expression of this difference is that one side greatly values Quantity >>>>> Quality, and the other sees things the other way round or at least a different balance between the two. This difference shows up in Vic's misinterpreting Oilman's Sv% ON estimates by an order of magnitude, which is about how far apart we are on such discussions.

  111. Scott says:

    The thing is, if the system that Vic, Tyler, Dennis et al. are using is better at predicting what will happen, I think that probably makes it a better explanation of what has happened.

    I mean if I roll a dice three times and get three sixes there could be all kinds of interpretations. Maybe one group says the dice is weighted and another says that what happened only has a 1/216 chance of happening, but the dice is fine, and over a long stretch the pattern will come to look like a regular dice. So I roll ten more times and things start to normalize, 2,3,1,5,4,3,6,1,1,2. I think at that stage, the group that made the more accurate prediction starts looking better in their explanation of the past.

    It’s the same thing here. If the scoring chances can predict future results (in terms of correlation with winning), well then they’re probably doing a pretty good job of describing what’s happened in the past too. If errors do that they’re equally valid. If they don’t do that, then they’re less valid. I guess I just find the suggestion that I secretly changed dice improbable.

  112. PerformanceOil says:

    I agree with what you are saying Scott. The only problem is that differences in results in hockey aren’t necessarily due to chance. Saying there is variation, doesn’t mean that the variation is due to chance.

    To be really technical, a dice roll isn’t pure chance either, though the physics behind the roll are so complicated no one can predict the result.

    Furthermore, for all you might hear otherwise, the predictive power of many of the shooting stats being thrown around is pretty poor, especially at the level of the player.

    Part of the divide comes into explaining the deviations from expectation that are seen. If the data doesn’t fit the model, to you write it off as chance? Well, if you do, then you can justify any difference in this manner (which is often done, with little justification in my opinion). If you are willing to do this, it becomes easy to build a strong case, because you build on the concepts you’ve already ‘proven’ and keep harping on the reproducibility (not predictive power) of data like it means something.

    I guess I just find the suggestion that I secretly changed dice improbable.

    There have been pretty major changes made to the game over the years. So in effect it is possible someone has been swapping dice every so often.

  113. Scott says:

    There have been major changes to the game, but the fundamental goal of putting the puck in the net more times than the opponent in order to win remains static. That’s what I’m talking about with the dice.

    I take your point about dice-rolling not being technically random. If you can come up with a better way of describing it that helps to predict future dice rolls, so much the better. The problem, as you said, is that randomness, for now, seems to explain things better. In the example I gave, I was merely trying to illustrate that randomness was a better explanation than the proposition of a weighted dice.

    It would seem to me that goal differential correlates well with winning. If the goal differential is being driven by a save percentage well outside of historical norms (like Smid’s for example), it’s been demonstrated that the success achieved will not be repeatable if underlying numbers like Corsi (and by the sounds of things, scoring chances) don’t line up. Corsi numbers tend to be more repeatable, i.e. they are a better reflection of what’s actually happening. At that point, it seems quitle likely to describe the save percentage as mostly luck.

    This isn’t to say that Corsi does a perfect job of predicting the future. It certainly doesn’t. The argument to this point from a lot of guys smarter than I has been that the biggest factor is actually luck. So far, that has been the most compelling argument. It doesn’t take a genius to say that the team with the higher save percentage is going to win. What needs to be discovered is why. So far anyway, why looks to be mostly luck. Perhaps someone will do a convincing study arguing otherwise.

  114. Vic Ferrari says:

    Scott:

    Yeah, that’s it. I posted a couple of files on timeonice a while ago, maybe late September, season splits. I was working towards something, trivia craps and the posts around it were part of the progression.

    I was hoping to lead to a groupthink on the whole predictive issue, I just needed to weed out the dummies first, and then I got sidetracked on other things. The trivia craps post was meant to separate the sensible from the teenage gamers that have become common around here. I’m not a big believer in hiveminds, but I do believe that 10 sensible people are better than one.

    Anyhow, even using just six randomly chosen games (almost random, 3 home and 3 away, and just games with the same amount of rest for both teams) … it turns out that the repeatability of the underlying numbers is huge, of course. r=.8

    What is much more impressive is that the predictive value, from just six games, of corsi+/- to future EV+/- is r=.6, which is bumping right up against the trivia craps model

    Timeonice lists out the raw data. This isn’t rocket science, takes maybe 30 minutes to reproduce that.

    Turns out that if you are playing well, eventually it turns into results. Whoda thunk it?

    But as much as I believe in the simple truths of hockey, obviously there are other things at play. Not to the extent that the wacky kids around here imagine, but obviously there.

    I mean we all know that PIT are likely to have a better rate of finish that CBJ, and that VAN are likely to have a better rate of saves than COL. Yet we’re brushing up on trivia dice predictive values here, and we’re ignoring this stuff still. Odd.

    We are close to having this shit figured, and while I imagine that Bruce and Performance Oil are nice people, and surely terrified to bet against me (I’ll extend either of you 20K with a verified home address btw) and are flattered that I acknowledge them [which sounds horrible to say I know, but it's true] … it’s time to move on.

    I’m open to any new format. If anyone is going there, let me know.

  115. PerformanceOil says:

    In the example I gave, I was merely trying to illustrate that randomness was a better explanation than the proposition of a weighted dice.

    Yes, this I agree with. However the key points are seem random, and best explanation.

    If we acknowledge that there is some fuzz in the data, we should be seeking out more data, not discarding that which doesn’t fit our model (or conversely endorsing only that which does).

    It would seem to me that goal differential correlates well with winning.

    First, why say ‘it would seem to me’? Why not present some data to support your assertion? Anyway, if you did, you would see it correlates quite well at the level of the team, but the numbers that most usually talk about are at the level of the player, and these do not predict EV +/- nearly so well. So, for example, I haven’t been hearing a lot about how the Oilers as a whole are being outchanced/outchancing, but rather how player X is playing better/worse than their counting numbers indicate (for instance in the case of Smid, and his unsustainable SV% ON is held up as further evidence).

    The first is generally a fair extrapolation to make, the second is not.

    So far, that has been the most compelling argument. It doesn’t take a genius to say that the team with the higher save percentage is going to win. What needs to be discovered is why. So far anyway, why looks to be mostly luck.

    What is the evidence that it is luck? Saying there is no evidence that it is not luck does not prove the reverse.

    Also, while luck may indeed factor into it, that doesn’t mean all variance should be discounted.

    For instance, as Oilman aptly pointed out, it is pretty clear that Smid’s ungodly SV% ON is unsustainable. However, that does not prove it was ‘luck’. There are basically four variables that go into this result (simplified). 1) # of chances generated. 2) Quality of chances generated. 3) Execution of said chances (i.e. having the puck in a prime scoring with the goalie down and out won’t help if you bury it in his crest. 4) Goalie’s ability to save the puck. This would also imply the goalie needs to see the puck and be in position, so a defensemen’s ability to clear screens/rebounds would fit in here.

    Of those things, three things could be directly impacted by a given skater (only the opposing player’s execution, or lack thereof is out of control, although you could argue that quick pressure on the puck might impact this). Additionally, in the end the goalie is responsible for making the save, and beyond making it easier/more difficult a skater can’t generally control that.

    Only one we have an easy objective measure of. So the solution is then to declare that that is all that is relevant, and any variance in the rest (which would basically aggregate into SV%) can be tossed out as luck?

    Foolish.

    You acknowledge the search must continue for the answer to ‘why?’. That is all us wacky kids are doing. ‘Luck’ is the easy way out.

  116. PerformanceOil says:

    (I’ll extend either of you 20K with a verified home address btw)

    So basically, I tell you where I live and you give me 20K?

    My mom warned me about the internet..

  117. Oilman says:

    So 20 scoring chances against when Smid is on the ice, and one goes in, and that’s all because of his play?

    I’m sure there’s a lot of “shit happens” happening there too Vic. I don’t recall too much of what he’s been doing either to be honest but I don’t recall many goalposts or 10 bell saves with him on either – he’s just been efficient. Does that make him the best D-man on the Oilers? no. But he’s been doing more right than wrong down this stretch it seems. What’s the quality of those chances anyways Vic? I’d take unscreened slapshots from the high slot to my glove side all night long if given the choice – they look like great chances and big saves but truth is they’re easy stops….shhhh, don’t tell anyone.

  118. Scott says:

    First, why say ‘it would seem to me’? Why not present some data to support your assertion? Anyway, if you did, you would see it correlates quite well at the level of the team, but the numbers that most usually talk about are at the level of the player, and these do not predict EV +/- nearly so well.

    Well, I said seems because I didn’t look into it. Goal differential correlating well with winning has just been conventional wisdom. Same reason for the lack of data to be totally honest.

    I’m surprised that you don’t think these underlying numbers eventually translate into goal differential which will eventually translate into winning. Last year for example, here is the leading Corsi (with goal differential) per team and the worst Corsi (with goal differential) per team (minimum sixty games, all taken from BtN):

    Anaheim:
    Getzlaf, 10.2, +29
    Pahlsson, -8.3, -6

    Atlanta:
    Kozlov, -3.5, -6
    Boulton, -17.4, -10

    Boston:
    Sturm, 6.8, +8
    Schaeffer, -3.5, +4

    Buffalo:
    Roy, 9.4, +12
    Peters, -11.1, -3

    Carolina:
    Whitney, 11.3, +2
    Letowski, -3.4, -11

    Columbus:
    Russell, 10.2, -10
    Novotny, 0.2, -10

    Calgary:
    Langkow, 9.7, +18
    Godard, -11.3, -9

    Chicago:
    Lang, 8.1, +15
    Burish, -8.1, -17

    Colorado:
    Arnason, 6.1, -1
    Guite, -5.1, -1

    Dallas:
    Lehtinen, 11.7, +10
    Barnes, -5.7, -9

    Detroit:
    Zetterberg, 24.0, +34
    Stuart, -5.5, -10

    Edmonton:
    Glencross, 5.4, +7
    Cogliano, -11.7, -1

    Florida:
    Booth, 1.7, +15
    Campbell, -5.7, -15

    Los Angleles:
    Armstrong, 4.6, +5
    Ivanans, -19.3, -10

    Minnesota:
    Bouchard, 2.8, +14
    Fedoruk, -6.3, +2

    Montreal:
    Kostitsyn, -2.0, +16
    Dandenault, -13.1, -12

    New Jersey:
    Elias, 16.4, +3
    Madden, -6.9, +5

    Nashville:
    Hlavac, 7.4, -4
    Smithson, -12.0, -15

    Long Island:
    Bergenheim, 4.5, -2
    Guerin, -3.7, -8

    New York:
    Jagr, 14.8, +11
    Hollweg, -3.1, -12

    Ottawa:
    Redden, 6.9, +9
    Phillips, -4.6, +6

    Philadelphia:
    Upshall, 7.5, +3
    Dowd, -15.6, -2

    Phoenix:
    Jovanovski, 4.6, -6
    Tjarnqvist, -7.3, -2

    Pittsburgh:
    Gill, 1.0, +3
    Laraque, -12.6, 0

    San Jose:
    Carle, 11.7, -5
    Mitchell, 0.4, -4

    St. Louis:
    Backes, 3.4, -6
    Mayers, -8.1, -20

    Tampa Bay:
    Ouellet, 8.6, +12
    MacDonald, -5.2, -11

    Toronto:
    Ponikarovsky, 11.3, +7
    Moore, -3.1, -4

    Vancouver:
    Sedin, 9.8, +5
    Ritchie, -16.5, -11

    Washington:
    Backstrom, 14.6, +15
    Cooke, -2.5, -1

    Looking at these numbers then we have the outshooting guy coming out ahead 25 times, with 3 losses and 2 ties. So it’s not a perfect predictor by any means, but any fool can see that outshooting helps. Now, I’m not saying that Corsi is the only thing going on, so we agree there, but the reason Corsi is helpful is because, from what I understand, it’s more repeatable than things like EV save percentage which tends to fluctuate more. People have been calling that luck, and so far no one has been able to describe it in a way that’s more helpful using data. If you think you can, I’m all ears. That would be awesome actually! Thing is, with erros, they don’t. So people don’t care about them so much. Now with scoring chances, they might. And what if they do it better? Wouldn’t that be something. And what if they repeat? These seem to be the questions that the stats community is asking without being too worried about explaining why they’re using the methodology. They’re using it because it might work. They’re not using Staples’ because it doesn’t. Or at least that’s my understanding.

    As for evidence that it’s luck, go over to IOF and read Vic’s post about trivia craps. To me, that was a pretty good argument that it’s luck. Sorry you haven’t read it, but I don’t know that’s wise to run around telling people they’re fools just because you don’t know the rationale.

    So yeah, we’re looking for “why.” I’m mostly an observer frankly, but this discussion is helping me to sort things through, so thanks for engaging.

  119. PerformanceOil says:

    Scott:

    You've shown that the extreme cases fit the model (generally). This is to be expected. The larger the discrepancy between shot differential, the more likely there will be a discrepancy between goal differential. This isn't because shot quality is meaningless, but rather that it will probably matter less for those at the extreme ends. As a very extreme example, someone who generates an infinite number of shots will likewise generate an infinite number of goals, assuming SH%>0. It won't matter how much greater than 0 it is.

    That's well and good, but there are a hell of a lot of players you didn't cover who lie in between the extremes. Also, while the relative difference between the best and worst you have presented is obvious, not all the best/worst GD guys were the best/worst shot differential guys on their respective teams.

    So, in general, yes if you have a positive SD you are more likely to have a positive GD. However, 30 players with the same SD'es could (and likely would) have vastly different GD'es.

    So again, looking at the extremes is one thing, but what about the other 400+ players that lie between the two extremes and don't fit on a nice straight line between the previously mentioned data points? I mean, do we need the underlying numbers to tell us that Lidstrom is a hell of a hockey player? No, the conventional numbers say the same thing, just as strongly. The question is what to do with the players whose GD is crap but SD is good, or vice versa.

    Anyway, no one has ever argued that shot differential doesn't matter. I certainly think underlying numbers 'add colour' (as those wacky Brucelexists are fond of saying), but I don't think they tell the whole story. Nor do I think they should be considered more believable then actual results, without some justification (which would be on a case-by-case basis).

    As for evidence that it’s luck, go over to IOF and read Vic’s post about trivia craps. To me, that was a pretty good argument that it’s luck.

    I did read it (well browsed it actually, usually Vic starts out pretty clearly and then goes all sideways, which is my cue to find something better to do). All I saw was the typical oversimplification of and condescension towards opposing views. Nothing to analyze or debate, since it was just him trying to lay out his opinion in a colourful way.

    Besides that, I’m not debating with Vic, I’m debating with you. If all you’ve got is ‘Vic’s right and you’re wrong’, there’s not much point. Using his arguments as a basis is fine, but at least paraphrase them to show your own thoughts as well. And I don’t think you are a fool, though I do think summing up all those things as luck without hard evidence is foolish.

    Anyway, if intuition was all we needed, the scientific method would never have gained traction. Instead, someone recognized that as wonderful as our ability to reason was, we were inherently unreasonable, and needed some rules to keep us in check. That way we could start moving past ‘and on the first day, God created..’, and get to the real meat of things.

    You show me where someone provides some hard evidence to support the idea of blind luck being the answer and I’ll take it seriously. Showing variability isn’t the same, by the way.

    An opinion piece by the prophet himself won’t cut it though.

  120. Bruce says:

    here is the leading Corsi (with goal differential) per team and the worst Corsi (with goal differential) per team (minimum sixty games, all taken from BtN)

    This is good work, Scott, and the sort of information that I would find more convincing if — and you knew that was coming! — the bottom guy on many of these teams wasn’t Eric Boulton, Andrew Peters, Eric Godard, Adam Burish, Raitis Ivanans, Todd Fedoruk, Ryan Hollweg and Georges Laraque. That’s a veritable goon show. These guys can’t play, they are going to get outshot and outscored, period. They’re going to look bad compared to anybody.

    Still, a good start, it gave me some ideas of a methodology when I have a little more time. Now that I’ve been formally dismissed, though, maybe I needn’t bother.

  121. Scott says:

    Bruce,

    It’s sad that you feel like you can’t put your ideas out there. If you change your mind, there’s a good chance many here will engage them, and more still if they hold water under scrutiny. So I guess that’s in your court.

    On to your point:

    That’s a veritable goon show. These guys can’t play

    That’s the point sir. That’s the point.

    PO,

    This isn’t because shot quality is meaningless

    I’m not saying that shot quality is meaningless. I think that it’s rather intuitive that not all shots are created equal. A guy generally has a better chance on a breakaway then a wrister from the point with no screen. We agree here.

    Not all the best/worst GD guys were the best/worst shot differential guys on their respective teams.

    Here is the rank of each player that finished 1st in Corsi on their team in terms of +/-:

    Getzlaf – 1st of 15
    Kozlov – 5th of 16
    Sturm – 4th of 14
    Roy – 2nd of 13
    Whitney – 8th of 12
    Russell – 12th of 14
    Langkow – 2nd of 15
    Lang – 2nd of 12
    Arnason – 10th of 14
    Eriksson – 6th of 14
    Zetterberg – 3rd of 18
    Glencross – 2nd of 15
    Booth – 2nd of 13
    Armstrong – 1st of 13
    Bouchard – 2nd of 15
    Kostitsyn – 1st of 16
    Elias – 8th of 16
    Hlavac – 10th of 16
    Bergenheim – 3rd of 11
    Jagr – 2nd of 18
    Redden – 3rd of 19
    Upshall – 5th of 16
    Jovanovski – 10th of 13
    Gill – 8th of 16
    Carle – 13th of 17
    Backes – 11th of 15
    Ouellet – 1st of 16
    Ponikarovsky – 3rd of 12
    Sedin – 4th of 11
    Backstrom – 3rd of 17

    And now a list of where the leader in terms of +/- finished in terms of rank in Corsi:

    Getzlaf – 1st of 15
    Armstrong – 4th of 16
    Chara – 7th of 14
    Pominville – 6th of 13
    Corvo – 3rd of 12
    Hejda – 11th of 14
    Iginla – 3rd of 15
    Keith – 5th of 12
    Stastny – 2nd of 14
    Morrow – 3rd of 14
    Datsyuk – 2nd of 18
    Nilsson – 5th of 15
    Cullimore – 7th of 13
    Armstrong – 1st of 13
    Gaborik – 10th of 15
    Kostitsyn – 1st of 16
    Oduya – 6th of 16
    Arnott – 2nd of 16
    Satan – 5th of 11
    Dawes – 4th of 18
    Heatley – 8th of 19
    Coburn – 4th of 16
    Malkin – 2nd of 16
    Thornton – 2nd of 17
    Perron – 6th of 15
    Ouellet – 1st of 16
    Sundin – 2nd of 12
    Pyatt – 8th of 11
    Kozlov – 2nd of 17

    First off, I recognize that I’m still dealing with leaders here and not every player, but I think that the leaders can tell us something:

    Players that led their team in +/- were in the top half in Corsi 25 out of 30 times, in the top third 18 out of 30 times and in the top three overall 14 out of 30 times.

    Players that led their team in Corsi were in the top half in +/- 23 out of 30 times, in the top third 19 out of 30 times and in the top three overall 16 out of 30 times.

    That looks like something to me. I know it’s not perfect and I know it’s not linear but it’s something and it’s something that looks to repeat. Guys finishing in the bottom third in one category probably need a correction in the other. But guys in the top three in each are probably good players in their given role and it’s something I’d be willing to bet on since Corsi looks to repeat: Kozlov, Sundin, Ouellet, Thornton, Malkin, Arnott, Kostitsyn, Armstrong, Datsyuk, Morrow, Stastny, Corvo, Iginla, Getzlaf, Backstrom, Ponikarovsky, Redden, Jagr, Bergenheim, Bouchard, Booth, Glencross, Zetterberg, Lang, Langkow and Roy.

    Will people that have different SD’s have different GD’s? Of course. That’s puck luck, and shooting accuracy and home-ice advantage and a bunch of other things that we’re having a hard time defining. There’s a lot more going on than Corsi and it isn’t ALL luck. I know that. But this correlation is real. This is where I see Corsi applying on the individual level re:Smid. His save percentage says he’s awesome but the Corsi don’t bear it out, and from what I can tell, there’s enough correlation between individual Corsi and individual goal differential that he’s bound to fall back to earth somewhat even though he might not be doing anything different.

    A question for you, PO, do you think it’s more likely for Pyatt and Hejda to outshoot this year than last, or do you think it’s more likely for their goal differential to be worse? Basically, I’m hoping to show with this that if you have a good SD you probably have a good GD and the same is true of GD to SD. But of course it’s a matter of probability and not certainty, and maybe that’s the part you’d prefer to bold. Which of course leads us to why.

    If all you’ve got is ‘Vic’s right and you’re wrong’, there’s not much point. Using his arguments as a basis is fine, but at least paraphrase them to show your own thoughts as well.

    Why would you talk about Vic being condescending and that write this? You had said no one laid out an argument. I honestly thought that was an argument, and he actually helped out a lot when people asked questions in the comments. I’m not trying to be disingenuous with you and Vic is not my “prophet,” just an intelligent guy that’s done some good thinking and that, in my view, is worth listening to. What can I say, I’m convinced of the argument he presented.

    What evidence do you have for a contrary view? I thank you for helping me to refine my own views but you haven’t provided much in the way of hard evidence for your own views either.

    Anyway, my understanding of the luck premise is as follows, and yes it is on the team level:

    You look at a game and see what percentage of even-strength shots are put on goal. Vic compared the first half of the season with the second half as it pertains to Corsi and shooting percentage differential. If you look at first half shooting percentage differential at the end of the first half to explain first half results it explains winning more accurately than Corsi does. But then if you use first half shooting percentage and first half Corsi to predict second half results, shooting percentage differential sucks something fierce and Corsi turns out to be a much better predictor.

    Or to quote from the “prophet” himself in case you’d like it explained more clearly (since this is really Vic’s view to begin with):

    “Not saying that shooting % and save% aren’t important, they obviously are. Just that so much of it is pure shithouse luck, that sensible people who are trying to analyze this stuff are well advised to treat them as pure luck to start, and add an element in for them after. Of course that’s a Bayesian approach, we’re taking what we know about hockey and trying to build a model. We learn more, we change the model.

    If you don’t do that, if you take a frequentist approach, then I think you’ll lead yourself into the woods and you may never come out.”

    What part of this do you think is wacky or not a real argument?

  122. Bruce says:

    It’s sad that you feel like you can’t put your ideas out there. If you change your mind, there’s a good chance many here will engage them, and more still if they hold water under scrutiny. So I guess that’s in your court.

    Scott: Actually I’m not too shy about getting my ideas out there. I do find the dismissive tone from some quarters to be a little off-putting, is all.

    My compliments for your work above, btw. There is clearly a positive relationship between shot differential and goal differential on an individual basis. My own approach towards this issue has largely been team-oriented rather than on an individual level, holistic if you will. It indicates a postive, but weak, relationship between outshooting and winning; in the Bettman Point era, outshooting teams take about 53% of the available points, outshot teams 47%. Some teams are exceptions to the trend, including the Oilers over the long term (throughout the dynasty years, and throughout the MacT years). Another notable exception was the 2007-08 Montreal Canadiens who won the East without a single player with a positive Corsi number. Meanwhile, the Columbus Blue Jackets missed the playoffs depsite the fact that every player with 25+ GP had a positive Corsi number. So it is clear (to me at least) that team strategies may differ w.r.t. the value of shot quantity vs. quality. (Special teams is obviously an important factor too.) OTOH, in other teams (Detroit) the relationship between outshooting and winning is much stronger.

    As mentioned above, Scott, you’ve given me a couple of ideas of methodologies to pursue. It’ll take time, though. Eventually I will post something on this subject on my own blog Oil Droppings.

  123. Scott says:

    Thanks for replying Bruce. I think special teams and percentages do play a big factor, but even on the teams not dependent on Corsi, the ones that do the best at it, still tend to do better at 5v5. Take Montreal last year for example:

    (Corsi rk) Player (+/- Rk)

    (1) Kostitsyn (1)
    (2) Kostopoulos (14)
    (3) Hamrlik (6)
    (4) Kovalev (2)
    (5) Plekanec (3)
    (6) Streit (13)
    (7) Koivu (11)
    (8) Ryder (10)
    (9) Smolinski (14)
    (10) Higgins (7)
    (11) Bouillon (4)
    (12) Gorges (8)
    (13) Komisarek (5)
    (14) Latendresse (9)
    (15) Markov (11)
    (16) Dandenault (16)

    The other interesting thing is taking out the dmen:

    (1) Kostitsyn (1)
    (2) Kostopoulos (8)
    (3) Kovalev (2)
    (4) Plekanec (3)
    (5) Koivu (7)
    (6) Ryder (6)
    (7) Smolinski (8)
    (8) Higgins (4)
    (9) Latendresse (5)
    (10) Dandenault (10)

    It’s quite a bit more jumbled than I’d expect, but with just forward you have seven out of ten guys within two spots and including all sixteen you only have seven out of sixteen guys within three spots. So I don’t know if this is representative of players in general, but looking at this whole team it seems that Corsi does the better job of projecting:

    1. Forwards
    2. Guys landing at the top.

    Anyway, this doesn’t really prove anything, more data collection than anything, we can see what we can see.

  124. PerformanceOil says:

    Scott,

    I’m going to go point by point, so it will be long:

    I’m not saying that shot quality is meaningless. I think that it’s rather intuitive that not all shots are created equal. A guy generally has a better chance on a breakaway then a wrister from the point with no screen. We agree here.

    First, while intuition may lead us to asking interesting questions, when it comes to determining where the truth lies, I believe facts, rather than intuition, should guide us. Intuition changes with experience and perception and is often flawed.

    Ok, that aside: the question is, do different players tend to generate different quality of shots? An extension would also be, do different players tend to have more success with the same quality of shot?

    If you accept the answer to both these question is yes or at least possibly, then you can’t boil things down to shot # and luck, can you? Or at least you have to provide evidence that shot # is the large majority of the effect.

    That looks like something to me. I know it’s not perfect and I know it’s not linear but it’s something and it’s something that looks to repeat. Guys finishing in the bottom third in one category probably need a correction in the other. But guys in the top three in each are probably good players in their given role

    Yes this analysis I agree with, I think I already more or less conceded as much. In fact, you could take it a step further and analyze thresholds (and maybe that was the analysis Bruce alluded to wanting to do). So let’s if you have a SD of 20, you might have an 80% chance of having a GD of 10+, or something along that line. That certainly could work. Problem is, there might be a lot of year-to-year and team-to-team fluctuations, so it is hard to say how much you could generalize these thresholds.

    Anyway, assuming we just want to use discrete values, the problem again is, what about the middle-of-the-pack guys?

    What about the guy that is say +10 GD and -20 SD? Is he a good player?

    What about -10GD, + 40 SD? +10 +10? etc. etc. We don’t really need SD to tell who is elite or super-crappy, though it is nice to confirm things in multiple ways. We need to make conclusions about the guys in the ‘fuzzy’ zone. I mean, what about a guy who generates many chances, but is horrible at executing? Does such a player exist? If so, is he more/less valuable then someone who generates few chances, but has a higher turnover rate? You’d need a hell of a lot more than just SD to really judge, in my opinion.

    I should note here, though I implied it, the correlation between SD and GD league wide (@ player level) is very poor. I think it cleans up a bit if you go team-by-team (been a while since I did the analysis), but it is still quite poor.

    What can I say, I’m convinced of the argument he presented.

    Fair enough, but I’ve already had this discussion with Vic many, many times. I don’t mind if you think he is right, but just telling me to ‘go read Vic’ doesn’t really progress things from where I’m sitting. I’ve already read much of what he says, and I think much of what he says is unproven at best and wrong at worst. Tell me what in particular you think is valid and we can have a discussion based upon that (as you did in the rest of your post).

    You look at a game and see what percentage of even-strength shots are put on goal. Vic compared the first half of the season with the second half as it pertains to Corsi and shooting percentage differential. If you look at first half shooting percentage differential at the end of the first half to explain first half results it explains winning more accurately than Corsi does. But then if you use first half shooting percentage and first half Corsi to predict second half results, shooting percentage differential sucks something fierce and Corsi turns out to be a much better predictor.

    A few things. First, I have found flaws in several of Vic’s analyses previously. As such, I don’t blindly accept his math, without seeing it first. Generally he does not present his data, nor describe his methods in great detail, so I would recommend you use caution when he (or anyone!) gives you data you haven’t verified yourself, and I certainly include myself in this.

    That said, for the sake of discussion, and since I never had the time to double-check that one, I’ll concede the point.

    However, yet again, this is team-based stuff. I’ve already said I am fine with using SD as a means of evaluating teams, since it works pretty well at that level (though still nowhere near perfection).

    What part of this do you think is wacky or not a real argument?

    I don’t find any of it wacky. In fact it is a great hypothesis. However, in the absence of proof, that is all it is. Moreover, I agree completely SD may be the best metric we have to model success in the long term. However, just because it is the best, doesn’t mean it is any good. I think it is quite poor actually (@ player level), and so when people start coming up with hard conclusions based on it, that is where I have a problem. Using it is a starting point for a discussion on a player, not so much.

    A question for you, PO, do you think it’s more likely for Pyatt and Hejda to outshoot this year than last, or do you think it’s more likely for their goal differential to be worse?

    I have no way to objectively measure whether their Corsi will improve or not. It would depend on many things, which are difficult to quantify (team, growth as a player, changing roles, etc.). Additionally, what is a significant change? Do you judge it relative to the team? Year-over-year (in which case, what is %’age difference is significant)?

    Anyway, I don’t think that is really your point, so I’ll answer a different question. Given last year’s Corsi numbers, Hejda had a predicted GD of +0.07/60 (versus his actual of +0.68), while Pyatt had a predicted GD of -0.0049/60 (versus his actual of +0.79). So, assuming their Corsi remained static, I would predict that their GD would fall. Using last year as a guide, there would be a very good chance I would be right with Hejda (~85%) and even better with Pyatt (~97%). Note, this assumes this year’s data would be quite similar to last year; this may or may not be true.

    So that is really good, no doubt. A few things though:

    First, I somehow doubt these were random examples. I’ll reiterate, I have no problem using SD as a way to critically look at a player, but more analysis needs to be done. You obviously picked these guys as extreme cases, and yes, these extreme cases are unlikely to repeat. This also gets back to assigning probabilities and thresholds, which is essentially what I did here.

    Second, and I think this is more important, how far should we expect them to fall? The mean for Pyatt’s group is about -2 GD/60, so somewhere towards there. But how close would he be expected to go? Also, what if he manages to maintain an atypically high GD/60? Does that mean something? Is he just consistently lucky, or something else? For what it’s worth, Pyatt’s GD is still tracking above what his Corsi would predict, while Hejda has flip-flopped and is tracking below what his Corsi would predict, based on last season’s overall results (still early this season though).

    So to emphasize, this is based on one year. It would be interesting to see how the prediction would hold up using similar Corsi’s from this year. If it does hold up, then thresholds would be a great way to go about things. But that would be pretty labour intensive (though a good script could probably do it is well). Additionally, just generating the data does nothing to verify the assumptions made, so some sort of check woul dneed to be done. I’m not sure what would b ebest in this regard, but I’m sure some things could be thought up. Anyway, until such a model is built, case-by-case is what we have to work with, in my opinion.

    But of course it’s a matter of probability and not certainty, and maybe that’s the part you’d prefer to bold. Which of course leads us to why.

    I have no problem if someone wants to say “X player has Y% chance to regress down to the mean of his Corsi group” (assuming we could verify that the thresholds we assigned were generally accurate). The statements I have trouble with are “X player had Y Corsi, therefore he is the bomb/suck”.

    Hope that gives you a bit more insight; I don’t think we are particularly far apart in our reasoning.

  125. Vic Ferrari says:

    Scott

    Good stuff. Of course at the player level we need a real feel for the context of their ice time, or those numbers can change in a hurry.

    So if the Oiler’s third line starts playing better (either because Pisani finds his groove at centre or because someone else gets the pivot spot), the underlying numbers for those guys won’t change much, they’ll just take on more responsibility. And the 4th line will soar, because much of that will be lifted from their shoulders.

    At the team level Corsi is the absolute bomb, though. And at the player level equally, though as I say, repeatability depends on consistent context, and the bounces have a larger say with a smaller sample. But all in the right measure.

    The best predictor for teams and players, and I’ve tried a bunch of things, is Corsi^1.3.

    In the trivia craps model, over a six game stretch, “thinking positively while you roll the dice” is the overwhelming driver of immediate results, of course. And being able to win at trivia is the only predictor of future results, basically it’s telling you how good you are at the game, and that the bounces will even out eventually.

    By grabbing the results of six games of trivia craps, the predictive value of trivia ability over the next 76 games (trivia win rate in six correlated to future earnings) r=.611.

    By grabbing the results of six random games of NHL hockey (limited to games with equal rest for both teams, and 3 at home and 3 away), the predictive value of corsi ability over the next 76 games (corsi in six correlated to future EV+/-) r=.577.

    Corsi^1.3 (plus a smidge of zone diff, if I’m to be frank, but nothing else) and r=.595. That’s as close as I can get though. I’m struggling to gain that last bit of real estate.

    Corsi^1.3 outperforms straight Corsi as a predictor 87% of the time, btw, Scott. Don’t let the kids here sway you from your instincts, it’s just hockey.

    The other numbers all mesh. In the short term, the percentages drive the results almost entirely. They are just are terrible measure of how good your team actually is, for reasons that seem obvious to you already.

    If you send me your email address I can send you the raw data for last year, or a link that will generate it. If you generate the same independently, I can be sure that I didn’t make a scripting error.

  126. Scott says:

    PO,

    Just wanted to let you know that I did read what you had to say. It would be frustrating to write all that and then need to wonder. At any rate, I don’t think that we’re far apart in reasoning either. What it comes down to for me is that, even if the correlation isn’t great, it’s still the best we’ve got and so it’s probably worth working with as we try to push onwards. For me, Corsi provides some correlation at the individual level, so SD leads to GD and not the other way around. So when I see a big discrepancy like Hejda’s I think I have every reason to expect his GD to be worse this year than last. If he manages to consistently buck the trend, that may be significant. We can look more closely at his game to see what other factors may be at work, maybe find some other outliers to go with him and see what they have in common and go from there.

    Anyway, fun talks.

    Vic,

    I sent you an email at the address provided on clicking your name.

  127. PerformanceOil says:

    Scott,

    Thanks for that – good discussion, agreed.

    Corsi^1.3 outperforms straight Corsi as a predictor 87% of the time

    Fascinating Vic. Out of curiosity, what would the corrected value of a Corsi of -1 be? (I.e. what is -1^1.3 = to?).

  128. Vic Ferrari says:

    PO

    I took the absolute value and applied the exponent, then reapplied the sign. Is that what you asked? So for example a corsi of -23 would become -58.9. And one of +23 becomes +58.9. So -1 would remain -1.

    Intuitively I thought that toning down the shots metric’s power, from the sample of six games, and adding a bit in for zone diff (to credit those teams that work the cycle well) and shots a bit more heavily (thinking that teams that are a bit less shot happy probably have more of the play at evens than the corsi measure credits them).

    But it turns out that Corsi almost alone, taken from 6 games and exaggerated a touch with a smallish exponent, works a shade better in relating to results in the other 72 games. At least by correlation (and I doubt that any of this data is distributed normally).

    Which doesn’t make a great deal sense to me intuitively, especially from a math point of view. And granted, it’s a marginal improvement to go with the exponent.

  129. Vic Ferrari says:

    Scott (and anyone else interested):

    The raw data for EV results by game is at timeonice.com in scott1.html and scott2.html

    I think I have the ‘days rest since last game’ right. Though it’s been so long since I wrote the cron script to take that stuff off of NHL.com that I’m not sure where I was pulling the game dates from.

    Let me know if there are any errors in there, I haven’t noticed any.

    Props to anyone who has a go. I’ve tried goading other people into doing a trivia craps model for a while, without success. But it turns out that it wasn’t as much work as I thought it would be.

    And just EV shots +/- alone has terrific value if you start with a big enough sample of games. So we should be able to get that from the PL sheets for several seasons back, or from mc79 if he has them handy.

    It’s a good start, anyways.

    Some other things should obviously be considered, like the expected save% of the goalie tandems on the teams, based on the historical ability of the goalies to stop pucks. But that’s a lot of manual work, so it will be a long time before I get around to it.

  130. PerformanceOil says:

    Vic,

    What exactly are you correlating the Corsi from your random 6-game sample to? The team’s seasonal Corsi?

  131. Vic Ferrari says:

    No, to the results in the other 76 games. If you have a different methodology, feel free though. That’s the point really.

    The thinking being, in a sport and a league where the balance of play and scoring chances is close to even most games (the last two Oiler games haven’t been, but most are) and where the better team often loses, that even by sacrificing a bit by using a small sample (6 games), you’re grabbing some simple truth from it my think, a reasonable indication of whether or not they outplayed the bad guys at EV in those 6.

    And that the luck, ever present in hockey and in the trivia craps model, while still strong and important, would have faded in the remaining 76 games.

    Inspired by something Lowetide said this summer, and Matt referenced to the Flames after just 4 or 5 games. (they had a poor record but strong underlying numbers … i.e. he felt they had deserved better and the ship would right itself.

    And MacTavish, commenting on just two regular season games iirc, plus the preseason, and at least the last couple had nearly a full roster. He said, more or less (the quote was in one of the papers) that getting outshot was a sign of not having the territorial advantage, and that good teams get the better of the play. Something along those lines. Seemed a little early to be worrying, but it was the final nudge that moved me to have a look. And at first blush it looks like MacTavish has a point.

    So trivia ability to winnings in the next 6 matches correlation about r=.6. And a touch under that for Corsi in 6 games correlated to EV+/- in the other 76 hockey games.

    Makes sense, no?

  132. PerformanceOil says:

    So, by results you mean GD then?

  133. Vic Ferrari says:

    Yup. We’re just talking even strength play, here, of course.

    It might make more sense to look at the relationship of goals in the other 76 games to corsi+ (or some other measure of team quality at EV). A lot of ways to skin the cat, of course.

    And as I said, I used 3 away and 3 home, and games with equal rest for both teams for the 6 game sample. I didn’t look at how tough the competition was for each 6 game sample though, not beyond that.

    Again, feel free do do your own thing.

  134. Vic Ferrari says:

    edit for above

    that should read trivia ability in six matches correlates to winnings in the other 76 with r=.6

    $ +/- from the 6 games correlates much weaker to future winnings in the next 76.

    So the dice are modelling the percentages, and the trivia ability is modelling the Corsi (or other rational measure of team ability).

    in terms of correlations r=
    Trivia repeats: .81
    Corsi repeats: .77

    dice repeat: .00
    percentages (PDO#) repeat: -.10

    trivia predicts future earnings ($+/-): .61
    corsi predicts future EV+/-: .58

    trivia craps current results in 6 games, correlation(dice: .93, trivia ability: .15)

    NHL EV+/- for 6 games, correlation (PDO#: .87, corsi: .20)

    So as you can see, the model works a lot better than I thought it would when I was trying to prod others into doing it. I mean obviously it was always going to be close, I didn’t anticipate this close though.

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

© Copyright - Lowetide.ca