A HAZY SHADE OF WINTER

Go home, 2017 draft, you’re drunk. As the final lists are rolling in, vast differences are being seen in rankings. Craig Button, whose list I respect a great deal, released his top 31 today and it gives us a very good look into the gap in evaluation. Let’s start with Button’s list compared to my list and that of Future Considerations (who have published their final top 100).

BUTTON’S LIST 2017

The players in red (and white) are the outliers and the Tippett item is a smack. That said, what I like most about Button’s scouting lists is they have the courage of his convictions. If you’re going to evaluate these players, having a strong opinion on each of them is vital. I don’t know what reason(s) exist for Button to have him there, but am eager to find out. Same with Joseph and Chytil on the other end. My top ranked player not on Button’s list? Robert Thomas, who I had at No. 13 (Button No. 46). FC’s top player not on Button’s list? Nick Hague, who they have at No. 20 (Button No. 43).

WHAT DOES IT ALL MEAN?

If Owen Tippett or Kailer Yamamoto are available at No. 22, one imagines the Oilers grab the player and consider themselves lucky. I’ll tell you though, this thing seems weird. If all of the top flight offensive players are gone at No. 22, the play is probably trading down and grabbing two picks in the second round. You can see this for miles, because these kids are all different shades of equal.

 

written by

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

32 Responses to "A HAZY SHADE OF WINTER"

  1. John Chambers says:

    Strange draft. Holy moly.

    Usually that means though that the quality is a tad lower than usual, as it was in 2012.

  2. monsterbater says:

    LT,

    Off topic but i wanted to share with the group what i got out of a conference i attended today. One of the learning sessions was put on by Dan Haight, president and cofounder of DarkHorse Analytic who i am sure most of us are aware of. He was a really good speaker and provided some great information some of which we have discussed on here, and some that may be interesting tidbits from the work they did with the Oilers. To prevent a wall of text i will split this up into multiple posts. Thanks for the opportunity to share this.

    He opened with talking about his previous work with the Oilers and how the draft was their “Stanley cup”. 4 Time champions – Dynasty

    He talked about how they built models to help the team with drafting efficiency. They identified that teams are built with 10 “Core Skaters” basically top 6 forwards and top 4 D. He then mentioned that they didn’t spend a lot of effort on goalies because they are “voodoo”. He also mentioned that they avoided looking at players that projected to be “gritensity” because “they’re not important”. Both of these made me chuckle as i associate those comments with LT and Woodguy respectively.

    He went on to say they also looked at peak production ranges and said most of your core skaters are good for 10 years and therefore you can expect to draft 1 core skater per draft. Anything more is gravy, but not finding one can kill you. He then talked about how after the 2nd round it doesn’t really matter what round it is the likelihood of drafting a core skater is basically the same. Therefore it makes sense to trade down a bunch and stockpile lottery picks. Things we’ve talked about and have been published, but still good to hear this was all part of their work with the Oilers. Makes me wonder if 2013 and that crazy trade was the testing ground for this.

    He then talked about how the point of analytics is to assist in decision making, it isn’t the be all end all, and should be used in conjunction with scouting and “your gut” in order to make the most informed decisions. Analytics should supplement scouting and if they are applied properly, the current expectation would be a 10% improvement in efficiency with drafting. He then gave a bunch of examples of how 80% of the time analytics aren’t applied correctly and lead to a train wreck, even in their own case where they had a really good model that provided the client (an ambulance service) what they asked for, but they solved the wrong problem and ultimately their solution wasn’t adopted.

    Part 1 of 3

  3. monsterbater says:

    Continuing my last post, Dan then went on to discuss how to best develop an effective analytics team, often these team members take on multiple hats, but the flow of communication and decision making needs to be clear and appropriate.

    1. Data Steward – Basically the programming geek who can take in all the data and filter it into usable information that can be deciphered by others and quality checks the outputs to ensure it doesn’t have errors in it.

    2. Analytic Explorer – The people that take that data and cast about to find out what it all means, searching for that pot of gold so to speak, often spending tons of time to find out something that we already knew or something that is useless. I would say our own WheatNOil and Woodguy have done some of this with their Woodmoney work.

    3. Information Artist – This person takes the useful answers that the explorers find and makes it consumable to decision makers. Basically the person who can take a regression model with a certain R value, and tell a less computer/math savvy scout what it means and why its important.

    4. The Automator – This is the person that allows you to mass produce the results from all of this information and provide it to decision makers quickly and readily update it as more information comes in.

    5. The champion – This is the non-analytics or programming background person who has to be able to communicate between both groups and bridge the gap between “saw him good/bad” and “good corsi/bad corsi”. Without a champion, the two sides scouting/management and analytics will never make an informed decision.

    Part 2 of 3

  4. monsterbater says:

    Final bit on this, Dan ended the hockey portion of his talk by saying that the key thing they learned through this is that most scouts are open to the information, but it has to be presented to them in a bite size nugget that they can easily understand. Getting rid of the noise and multitude of data/work that they don’t need to see or understand other than the results is imperative. Ultimately he said they provided rules of thumb to scouts such as “don’t draft a player from the BCHL unless he has scored more than 40 goals because no one has ever made it from the BCHL who has scored less”. The scout can then go fact check this and see that yes this is true and this helps show them the value analytics can provide when they are making decisions about multiple prospects, especially in the later rounds.

    I will end this with the question I was able to ask and I won’t share much of the answer because I was too busy listening to him and didn’t want to be writing it all down as he was speaking to me. I asked “ You spoke about providing rules of thumb to scouts to help bridge the gap between their scouting and what analytics are telling us, how do you do this with a stubborn person, let’s call him steve simmons or marc spector, that despite the mountains of evidence refuse to see value in analytics”?

    He got a good chuckle out of it and said that you can’t win over everyone, you just have to win over the right people by showing the value and not discounting the value that gut decision can provide.

    Thanks LT for letting me share this. Hopefully at least 1 person reads this and finds it as interesting as i did.

    Part 3 of 3

  5. Lowetide says:

    Monster: Awesome stuff, and thanks for sharing. I would add that the bite sized nugget may have to be repeated over and over until the end of time. Adult learning and all. 🙂

  6. Kinger_Oil.redux says:

    monsterbater,

    – Great post and info, and Great Post LT!

    -Monster: I wonder what kind of list we could create that stats have demonstrated, along the line you mention: i.e. don’t draft BCHL if goals < 40, etc.

    – What are the "conventions" that are true, like the example you presented? (like for instance all things being equal the earlier born in the year, the better for draft considerations)

  7. Nuclear leak says:

    If this is the draft, trade all the picks. Stock up for next years draft, and find some forwards other teams have turned on.

  8. digger50 says:

    monsterbater,

    Great info thanks.

  9. theres oil in virginia says:

    monsterbater: Makes me wonder if 2013 and that crazy trade was the testing ground for this.

    Time for another update on “How’s that crazy trade?”

  10. Scungilli Slushy says:

    Monster great info cheers.

    I think the strategy trading down also needs the other direction. If a third is essentially the same as a lower round it is logical to trade down for more chances.

    But it is also logical to trade up to particularly the first round or early second if it is possible on its own or as part of a deal because the odds of an impact player are higher.

    You might find a player in a later round but because there are probably no unscouted markets left to exploit in a Datsyuk way you are probably not going to find an impact player which is what you want to find in the draft. Bottom roster players aren’t the type hard to find.

  11. Professor Q says:

    So, if we see Calgary’s owner in Hamilton and Seattle, we might expect a new arena?

    The NHL did recently buy the rights to Seattle Metropolitans, after all…

  12. stush18 says:

    Nuclear leak:
    If this is the draft, trade all the picks. Stock up for next years draft, and find some forwards other teams have turned on.

    I think this is the year to trade down and grabs all the picks,like 2013.

    Unless someone falls to us late in the first, I trade down with New Jersey. They have something like 11 picks this draft

  13. VOR says:

    scungilli slushy,

    At a rational level, I agree with you that there are no unscouted markets left to exploit. However, there is this little niggly fact that gets in the way. Or rather two related facts.

    The number of overage draft choices is increasing. The number of free agent college signings is increasing. So by definition the scouts are missing a lot of possible NHL players each year. These aren’t unscouted kids, they were hiding in plain sight.

    Plus, returning to LT’s post, I wonder if teams are influenced by these lists. Do any of these lists effect subsequent drafting behavior? What limited research there is on the subject suggests they do. Which would make the question should they?

    Thinking about that go me thinking about Michael Schuckers. In 2014 Schuckers wrote a paper about beating Central Scouting Service and the return on scouting investment.

    https://arxiv.org/ftp/arxiv/papers/1411/1411.5754.pdf

    In it he quotes Serge Savard:

    “We were wrong on the first round maybe 50 percent of the time. That’s mainly because of Central Scouting. When Central Scouting comes out with their first-round list, all the scouts think, “Oh, Christ, I better get this player in my list or I’ll look bad.” [All the scouts’] lists are similar because of Central Scouting. I only had one guy, Rick Taylor, who didn’t care about Central Scouting’s list and his list was so different than the others…. How come we missed Luc Robitaille? One of my
    scouts, Rick Taylor, had Luc Robitaille [rated to be drafted] in the first round and nobody else had him in the top fiverounds. The other scouts down-played Taylor. They said, “You only see Quebec. You don’t see Ontario. You don’t see the West. You don’t see college. You don’t see Europe.” So scouting is a tough thing to do.”

    At least we can say Button isn’t following Central Scouting.

    But once I got thinking about Schuckers and all his work I couldn’t help wanting to tell you about his latest work in data mining, the stuff they presented at this year’s Ottawa Hockey Analytics conference. I know a lot of people who post here are of the opinion that scouting adds no value. Schuckers, Seppa and Rovito agreed with you, up until they did the following experiment:

    http://statsportsconsulting.com/main/wp-content/uploads/Text_Mining_of_Scouting_Reports_for_Improving_NHL_Draft_Analytics.pdf

    It is hard to read their results and not conclude that we may be being a bit hasty in suggesting scouting can be beaten by performance based stats.

  14. LMHF#1 says:

    I knew that joker would have Cal Foote up way too high.

    Sure hope the Oilers don’t have some scout who was Adam’s good buddy at some point. Avoid the overrated WHL defencemen at al costs!

  15. rickithebear says:

    monsterbater: He talked about how they built models to help the team with drafting efficiency. They identified that teams are built with 10 “Core Skaters”

    Cup Core you say.

    1. Top 10 HD goalie:
    Goalies are not voodoo.
    They are quite simple to Identify.
    You just have to eliminate the useless data! (noise)

    2. #+ top 60 HD Dmen.

    Teams vary in their HDSH60 rates.
    Using an average goaltender:
    .965 LD Save%
    .825 HD Save%

    Looking at the best , avg and worst teams something becomes vary evident.
    Holding the Shot/gm rate constant (30 sh/gm)

    Worst team will give up
    13.5 HDSH/60 (see Calgary)
    16.5 LDSH X .965 = .5775 GA
    13.5 HDSH x .825 = 2.3625 GA
    2.3625 + .5775 = 2.94 GA
    (30-2.94)/30 = .902 Save%
    This is the mean established by the teams HD defence that the goalies performance (+/-) XXX Save%

    An average team:
    19.5 LDSH/60 X .965 = .6825 GA
    10.5 HDSH X .825 = 1.8375
    1.8375 + .6825 = 2.52 GA
    (30 – 2.52)/30 = .916 Save%

    The best Teams:
    22.5 LDSH X .965 = .7875 GA
    7.5 HDSH X .825 = 1.3125 GA
    1.3125 + .7875 = 2.1
    (30 – 2.1)/30 = .930 Save%

    By taking the Avg goaltender we eliminate Goalies influence on the GA results.
    Strictly HD defence can establish a mean range a goalie is performing around of .902 to .930
    In the last 4 seasons on average58 Goalies started 15gm.
    231 goalies.

    in that time 10 goalies performed At or above .930
    1. Talbot 13-14 NYR .941
    1. Hammond 14-15 OTT .941
    3. M. Jones 13-14 LAK .934
    4. Price 14-15 MTL .933
    4. Harding 13-14 MIN .933
    6. Stalock 13-14 SJS .932
    7. Bobrovsky 16-17 CBJ .931
    7. Dell 16-17 SJS .931
    9. Elliott 15-16 STL .930
    9. Rask 13-14 BOS .930

    On the bottom end 32 Goalies have had .902 or worse
    a lot of teams repeat giving an idea of Bad HD years.

    Edmonton 13-14 to 15/16
    Dubnyl; Scrivens; Fasth; Nilsson

    CGY 13/14 to 15/16
    Hiller; Berra; Ortio

    Dallas 13-14 to 16/17
    Ellis; Niemi; Lehtonen

    COL 13-14 to 16-17
    Varlamov; Berra

    CAR 13-14 to 16-17
    Lack; Ward; Khodobin;

    NYI 13-14 to 14-15

    I use to LOL thinking of the report that said Dmen do not influence GA.

  16. Rondo says:

    So hard to figure who Oilers will take at #22. Maybe one of these

    Lias Andersson,
    Kailer Yamamoto,
    Erik Brannstrom
    Ryan Poehling,
    Klim Kostin,
    Isaac Ratcliffe,
    Robert Thomas,
    Jason Robertson,
    Filip Chytil,
    Josh Norris

  17. N64 says:

    VOR: It is hard to read their results and not conclude that we may be being a bit hasty in suggesting scouting can be beaten by performance based stats.

    Wow. From worst to best performance:

    Actual NHL team picks
    Performance Metrics Only
    Scouting Report Only
    Performance & Scouting

    The study only had access to half a million words of public scouting reports. A wise team would commssion a private run with their own internal reports compared to metrics and external reports. Is the severe drop off in actual result coming from bad internal scouting, herd mentality of scouting org, final ranking process, procurement philosophy., draft for need, manager interference​, owner interference? Doubt that basic scouting is the issue. Suspect​ that some teams aren’t​ being beat.

  18. Barcs says:

    If Tippet or Yamamoto are available at 22, that would be fantastic.

    Personally, I’m hoping for Liljegren to drop. Surprised FC had him so low, highly doubt he makes it to the Oil at 22.

    If none of those 3 are still around? Trading down seems like a smart move to me.

    Probably important to note that if the Oil wish to trade down because of how close the prospects are, there may not be as much incentive the other way for teams to trade up.

    I’m sure they could find a partner, but my point is that the picks offered back in that trade might be less than we think.

  19. PunjabiOil says:

    monsterbater: Thanks for the summary. Do you know if they are employed currently for the Oilers?

    Have a soft spot for Dan – he was a co-professor for one of the courses I took.

  20. digger50 says:

    If we are contemplating trade down to increase the likelihood of success, then Trading out should be the move. Make certain of success by trading #22 for a real NHL player, one that can fill a present need.

    The take your chances on rounds two and onward. Seems a better strategy than trading down and crossing your fingers.

  21. --hudson-- says:

    monsterbater,

    Great summary!

    If you are interested in watching more, he gave a similar talk earlier this year: https://youtu.be/Uj1WzYxUEV0?t=38m11s

    G-Money is also in the first part of the video with his presentation.

  22. digger50 says:

    Could Sekera’s spot be covered by Brandon Davidson next year?

    Davey may end up in Vegas, not many know him. Would a third round pick bring him back to Edmonton? I wonder if He could do the job.

  23. Ryan says:

    digger50:
    If we are contemplating trade down to increase the likelihood of success, then Trading out should be the move. Make certain of success by trading #22 fora real NHL player, one that can fill a present need.

    The take your chances on rounds two and onward. Seems a better strategy than trading down and crossing your fingers.

    We don’t have a second round pick.

  24. Scungilli Slushy says:

    VOR:
    scungilli slushy,

    At a rational level, I agree with you that there are no unscouted markets left to exploit. However, there is this little niggly fact that gets in the way. Or rather two related facts.

    The number of overage draft choices is increasing. The number of free agent college signings is increasing. So by definition the scouts are missing a lot of possible NHL players each year. These aren’t unscouted kids, they were hiding in plain sight.

    Plus, returning to LT’s post, I wonder if teams are influenced by these lists. Do any of these lists effect subsequent drafting behavior? What limited research there is on the subject suggests they do. Which would make the question should they?

    Thinking about that go me thinking about Michael Schuckers. In 2014 Schuckers wrote a paper about beating Central Scouting Service and the return on scouting investment.

    https://arxiv.org/ftp/arxiv/papers/1411/1411.5754.pdf

    In it he quotes Serge Savard:

    “We were wrong on the first round maybe 50 percent of the time. That’s mainly because of Central Scouting. When Central Scouting comes out with their first-round list, all the scouts think, “Oh, Christ, I better get this player in my list or I’ll look bad.” [All the scouts’] lists are similar because of Central Scouting. I only had one guy, Rick Taylor, who didn’t care about Central Scouting’s list and his list was so different than the others…. How come we missed Luc Robitaille? One of my
    scouts, Rick Taylor, had Luc Robitaille [rated to be drafted] in the first round and nobody else had him in the top fiverounds. The other scouts down-played Taylor. They said, “You only see Quebec. You don’t see Ontario. You don’t see the West. You don’t see college. You don’t see Europe.” So scouting is a tough thing to do.”

    At least we can say Button isn’t following Central Scouting.

    But once I got thinking about Schuckers and all his work I couldn’t help wanting to tell you about his latest work in data mining, the stuff they presented at this year’s Ottawa Hockey Analytics conference. I know a lot of people who post here are of the opinion that scouting adds no value. Schuckers, Seppa and Rovito agreed with you, up until they did the following experiment:

    http://statsportsconsulting.com/main/wp-content/uploads/Text_Mining_of_Scouting_Reports_for_Improving_NHL_Draft_Analytics.pdf

    It is hard to read their results and not conclude that we may be being a bit hasty in suggesting scouting can be beaten by performance based stats.

    Vor I appreciate your work and love of stats, I wonder were you around when the seminal works on many things about the draft were studied? The only part of hockey stats that have endured without much challenge relate to that.

    Most of the seminal works weren’t by people for hire at the time or for sale. Almost all were by people who used math in their work life that also loved sports. The good ones offered mathematical proofs of their ideas and work. Most are hired now and the blogs dark, which leads to a lot of questions we used to simply post links for that people could explore their questions at with evidence.

  25. dustrock says:

    Dumb question: if it’s such a weak draft, which team is going to trade us two 2nds for our 22nd?

  26. Ryan says:

    Okay this thread is nearly in the books.

    My idea of trading Eberle for Kovalchuk has been met with comments of derision at best.

    Ladt year Kovalchuk had an NHLe of 79 points while Jordan Eberle had 51 points in 82 games.

    Clearly, my trade idea is absurd.

  27. Barcs says:

    Ryan,

    I don’t think it’s absurd. I feel like Eberle is worth more due to his age than Kovalchuck, even if K scores more in the next season or two.

    Especially if it costs around 6M to sign him. No cap space, gained, and an asset that will depreciate much faster. I would rather look to a D-man.

    For those reasons, I don’t think it happens. But Chiarelli has surprised me before.

    K would look great T-ing up passes from McD and Drai on the PP though

  28. VOR says:

    scungilli slushy,

    I am older than sports analytics.

    So yes, I was around. I was all ready working as a modeler when Bill James first Abstract came out. My initial thought was, wow I am not the world’s only sports nerd. I soon learned there are a lot of us out there.

    I lived the glory days of the amateur sabermetrician. But I think analytics in all its forms was always destined to turn into a serious field of inquiry dominated by professionals and academics. The early bloggers getting picked off by professional teams is just part of that trend.

    So while I hate how many of the great blogs I used to read have gone black I am delighted to see more and more hockey related material showing up in scientific journals. I seldom get through a week without stumbling on a new paper or thesis. What I love is there is still real creativity at work and the field is largely free of the tyranny of least publishable units.

    I think many of the pioneers you refer to will soon be passed by the technology in any case. Their early work will be rendered irrelevant. I had a conversation this week with a young researcher who tried very hard to explain to me this radical idea he has for using a very specific kind of neural network called a perceptron with data mining, Markovian Monte Carlo Simulations, and linguistic meta mapping to improve NHL drafting. I could keep up, barely. The next generation of Masters and PhD students will leave me behind. I don’t think I am going to be alone either.

    For example, and the concept is easy to understand. at Ottawa a lot of the excitement was about drafting based on fit. This the idea you can data mine public sources (as in the Seppa paper above) and work out whether or not the player will be a good fit for a particular NHL team, both on the ice and in the locker room. Speculation is running high that it may be possible, with large enough data sets and good enough mining and scraping of data to match prospective draft choices not just to team but even to a line mate or line mates. It is all just a little too Orwellian for me. Not to mention that to do it takes serious programming skills, large amounts of computer resources, and months of free time.

  29. Professor Q says:

    Ryan:
    Okay this thread is nearly in the books.

    My idea of trading Eberle for Kovalchuk has beenmet with comments of derision at best.

    Ladt year Kovalchuk had an NHLe of 79 points while Jordan Eberle had 51 points in 82 games.

    Clearly, my trade idea is absurd.

    No derision from me. I suggested it a few weeks ago but I usually get crickets anyway. Don’t feel bad, haha.

  30. monsterbater says:

    PunjabiOil,

    He never specifically addressed whether they still do work with the Oilers, so i am unsure. I suspect they still do though, but that is more of a gut feeling than anything

  31. Mariusz Czerkawski says:

    Thanks again for letting me post the info on the analytics talk LT.

    I have changed my screen name to something less immature and to honour the Polish Assassin

  32. Scungilli Slushy says:

    VOR:
    scungilli slushy,

    I am older than sports analytics.

    So yes, I was around. I was all ready working as a modeler when Bill James first Abstract came out. My initial thought was, wow I am not the world’s only sports nerd. I soon learned there are a lot of us out there.

    I lived the glory days of the amateur sabermetrician. But I think analytics in all its forms was always destined to turn into a serious field of inquiry dominated by professionals and academics. The early bloggers getting picked off by professional teams is just part of that trend.

    So while I hate how many of the great blogs I used to read have gone black I am delighted to see more and more hockey related material showing up in scientific journals. I seldom get through a week without stumbling on a new paper or thesis. What I love is there is still real creativity at work and the field is largely free of the tyranny of least publishable units.

    I think many of the pioneers you refer to will soon be passed by the technology in any case. Their early work will be rendered irrelevant. I had a conversation this week with a young researcher who tried very hard to explain to me this radical idea he has for using a very specific kind of neural network called a perceptron with data mining, Markovian Monte Carlo Simulations, and linguistic meta mapping to improve NHL drafting. I could keep up, barely. The next generation of Masters and PhD students will leave me behind. I don’t think I am going to be alone either.

    For example, and the concept is easy to understand. at Ottawa a lot of the excitement was about drafting based on fit. This the idea you can data mine public sources (as in the Seppa paper above) and work out whether or not the player will be a good fit for a particular NHL team, both on the ice and in the locker room. Speculation is running high that it may be possible, with large enough data sets and good enough mining and scraping of data to match prospective draft choices not just to team but even to a line mate or line mates. It is all just a little too Orwellian for me. Not to mention that to do it takes serious programming skills, large amounts of computer resources, and months of free time.

    I believe that once a math system becomes so complex and large that it can’t be understood it becomes basically useless because nobody knows if the output is good or not. Until the results are observed over time, and then adjustments can be made, again nobody really knowing if it will work.
    Or you could just hire a scout who looks at NHLE and draft success odds and reads a bit about the kid, talks to his coaches. I never said I thought stats should replace scouts.

    Draft success odds have been categorized. You must know that a undrafted player or overager has very little chance of becoming an impact player. There are some decent college players, but again not a lot of them become impact players.

    There are some good ones yes. But not a lot of them, and they are relatively expensive, take a roster spot, and are not under control for long even if closer to the NHL and don’t cost a pick.

    I take that your point was that the draft eligible players aren’t being discovered and that is the untapped market. Again, looking at success odds which Cullen set at a meagre 100 NHL games, by mid second round the odds of those 100 games is at 33.3%. That is the odds for a fringe player which 100 games is to me.

    Basically there are not many players with an NHL skill set in the world. Which is why I think trading up has more value than trading down and lowering already low odds. Looking for outliers seems to me not a good use of resources beyond a scout’s opinion because it is essentially lucky to find one. The Oilers had to go out and get the real Lucic for example.

Leave a Reply

Want to join the discussion?
Feel free to contribute!
© Copyright - Lowetide.ca