Really enjoyed your PDO article! I was wondering about players who sustain high shooting percentages in defiance of regression - at what point does it become possible to say with confidence that this outperformance is due to skill rather than luck, and that it's ~repeatable in following years? Like with McDavid shooting above 15% in four of his five seasons in the NHL
That's something I tend to play by ear. Some guys are just super selective or super accurate with their shooting. When I look for a regression candidate I'm looking for a big difference in results - like Zibanejad going from 12.7% to 12.7% to 19.7%.
Not amazing but I like Hornqvist at least. Bobrovsky could be a bounceback candidate as well. And they didn't do the stupid thing and trade Mackenzie Weegar, which is good.
Is there a way to calculate the error / confidence interval with Patrick Bacon's model? I know it was impossible with EvolvingWild because of how that's calculated. Just would be interesting to look at 2 players, one worth 2.0±0.2 WAR and another worth 2.0±2.0 WAR.
Toronto, significantly. Possibly another few Canadian teams as well. And especially teams like Arizona, Minnesota, and the California teams in the Pacific who have a playoff spot up for grabs that might not be otherwise. Who's hurt is probbaly one of CBJ, DAL, NSH, or CAR.
Hello. I see that most analytical models project the capitals to take a significant step back from last year, even though it feels like they've improved in the areas that they struggled last year. Is it because of the age? Or is it something else im missing about the offseason
Age is probably a big part. Goaltending is also a big question mark since Samsonov just okay last year. I have the Capitals as a playoff team but I could also see them take a big step back. Adding Schultz is bewildering to me and playing him the top six is questionable. So is pushing Jonas Siegenthaler out of the lineup.
I wanna know what the fuck the Preds are doing. We didn't add any big FAs for a playoff push but we didn't trade anyone away for draft picks either. Hell, we traded AWAY draft picks for a bottom sixer. I liked the Askarov pick but where is this franchise going??
If I had to guess, I think they're rolling the dice on this roster to figure it out one more year. They've committed to being competitive by signing Duchene, and with Josi, Ellis, and Forsberg in their primes it makes sense. Their roster really isn't bad - I have them projected in the playoffs. The question is if they can put it together. I liked Benning and Borowiecki as the bottom pair, I like bringing in Haula and Granlund. The x-factor for this team is Ryan Johansen figuring out his crap.
Could Mcdavid or MacKinnon average over 2 pts per game this season? Weaker opponents, strong supporting teammates. Seems a bit crazy, but perhaps worth some consideration.
I would be surprised but it's not impossible. The Wild did well to buy low on him in 18-19 but his individual shooting was nuts this year. Seems more likely to me that he'll settle as a 55-65 point player but maybe he properly broke out as a sniper.
Unfortunately I don't see any reason to believe that Zacha is on track to be anything but a bottom six PK specialist. His underlying numbers have been pretty appalling so far. Maybe he'll change that this season but there's no foreshadowing yet.
One more (series of) question(s) - what are your thoughts on analytics in prospect evaluation? I think there is obvious value, but only when hedged with not just the eye test, but projection moving forward, which is something I haven't really seen from the big names. Your analysis of Drew Doughty, while not re: a prospect, was a masterclass in combining all three elements.
How much weight would you give analytics in prospect evaluation? Are you aware of anyone who you feel does this particularly well? Do you believe there is a way to build projection into a model with any degree of sophistication beyond simple production translation rates?
Amateur scouting with analytics basically takes what I do but tries to do something much much harder. When I write about a player (like Doughty), all I want to figure out is how good he is in his current league and what makes him good or bad. An amateur scout has to do that but then estimate how that will translate to playing against much better players, after development, oftentimes in a completely different country. It doesn't matter if player X is a 95th percentile QMJHL player, it matters how his skillset will do and evolve at the next level. Doing all that stuff within one league is hard enough without having to project outwards.
From what I can tell, prospect analytics are all about quantifying the eye test part of scouting. So the report doesn't just have to say "Player X is strong in transition" but they have actual evidence for it. You still have all the challenges of projection and translation and all that, but at least you have something concrete to point to. I am not confident that you could build a proper and predictive model for prospects using this data because you'd need to be able to translate microstats/corsi/etc. from league to league and the amount of data you'd need to get even part of the way there would be massive. But at least you can improve and sharpen your scouting reports.
In summary: i will never be a prospects guy and this is part of the reason why.
Hey there, I'm just a college student English major who quite likes statistics so I wanted to say the work you do is really educational in terms of the nature of predictive statistics. I spend an ungodly amount of time looking at the resources you put out and I feel like my understanding of probability and hockey have both dramatically I proved because of your hard work. So thank you for that! Also, I was wonder who are players that, in your experience, are the best at creating doubt in most predictive models. Players like Patrick Kane and Ovi, from what I've seen, generally don't fair too well in analytics circles relative to their perceived legacy. Do you think they actually bring more value than the models show or are they overhyped by the public? And if they are just overhyped, who do you feel is commonly undervalued by your data and that of your peers?
I really appreciate it! Predictiveness is really the killer for hockey stats and the trick is to try to be as "less wrong" as possible.
Kane and Ovi are tricky. The model I use, Patrick Bacon's WAR model, is much higher on Ovechkin and other snipers like Laine and Hoffman than the EvolvingWild model or others because it values individual goal-scoring ability more. Kane is also a great sniper but he's one of a small coterie of players who you could argue elevates the finishing of his linemates with passing. Those are the players I would say are undervalued offensively due to fundamental and unavoidable issues with xG models. This could also apply to players who are very good at generating offence off the rush or screening the goalie.
I'm not overly confident in any of them. I have serious reservations about Kapanen, Matheson, and Ceci and I really hope that coaching can address those issues (esp. with Matheson. On the other hand Jankowski, Rodrigues, and Sceviour are all competent bottom sixers so hopefully they'll live up to that.
Pretty clear that the Leafs, Avalanche, and Hurricanes put a lot of emphasis on it, other than that it's not super easy to tell and I'm not even close to enough of an insider in this industry to have a sense for it.
May you please rank the Sens cast of misfit impending UFAs (Coburn, Gudbranson, Paquette, Stepan, Galchenyuk, Anisimov, Reilly) in terms of how much interest you believe there will be at the deadline, and in terms of how much interest you would have (in a vacuum)?
I would have 0 interest in any of those players except for maybe Reilly for a late round pick. Only Stepan strikes me as a guy who might get some attention at the deadline and even then I see him as a 3rd-4th kind of player. Really do not like Ottawa's player evaluation this summer.
For the most part I agree. I liked the idea of a Stepan acquisition, but only contingent on a) them flipping Tierney elsewhere when he was unsigned and b) the price being very low. I think Paquette will provide some value, and should generate interest. Reilly I suppose can play a 7/8 role and plug in as an ok, but heavily sheltered, puck-mover. The rest of them are the epitome of "taxi squad", imo, but I have a feeling (perhaps a hope) that GMs around the league may not treat them as such at the deadline.
Depends on what kind of team I have. If I'm trying to win and have a bit of roster/cap space I like Derek Ryan, Nick Holden, and Mathieu Perreault, and Jordan Weal. If I have a young team I'd look at Ho-Sang and Balcers.
It's hard to bet against Colorado or Tampa Bay which means it will be somebody totally left field. I really think certain elements of the media have been overrating Ottawa's improvements this offseason - I think they are going to be worse than they were last season, maybe even by a lot. I also am not even close to sold on the Rangers. The only reason they even came close to a playoff spot last season was ridiculous puck luck for the Panarin line and Zibanejad and a borderline Norris-calibre rookie season from Adam Fox. I could easily see them take a step back.
In terms of underrating, my model likes the Jets and Ducks more than most others and definitely more than most of the media. The Jets were a 99 point team in 2019 and a legit contender before that. Their defence is rough but not as bad as last year, and adding a legit 2C is huge. I think they'll be competitive at least. I'm more skeptical about the Ducks but they're in a soft division and I like how their top 4 looks now. If Zegras or Steel can be a legit offensive threat they could fight for a spot.
As a Rangers fan this was tough to hear, and I kinda agree with you, but despite that I am still more excited for this year than I have been in a while.
Pettersson is a very interesting player and this season will be a good test of how good he actually is. Some WAR models rank him as legit a top 3 player already, while the one I use is a bit more hesitant to give him full credit for his on-ice finishing. If he can once again outperform on-ice expected goals to such a huge extent while having very strong playdriving results that would be a good sign for him. But if Miller or Boeser take a step back it could look like a decline in his own results when it's actually just the model correcting itself if that makes sense.
I wrote a piece a few days ago about players whose individual and on-ice shooting percentages dropped, and that's usually the best bet. Dallas players like Seguin/Benn/Radulov for one. Also I'd check out goalies with decent pedigree who struggled last season - Bobrovsky could be good value since his results really weren't that bad after October.
Yes I do. The question for me is whether it's good enough for them to contend for a Cup. Analytically speaking the pieces are all solid enough, it's just a matter of whether the puzzle pieces can fit together properly. Can Jankowski and McCann work well together despite being shoot-first defensive forwards? Can Lafferty bring a new element? Can Blueger and Tanev work well without having Aston-Reese as the forechecker? Will be interesting to see, especially if a guy like Kapanen finds himself down there eventually.
in honor of the penguins playing AGAINST jack johnson again instead of WITH jack johnson, which players would make your 2021 worst rated all-NHL team?
What do you think of Alexander volkov? Does he take kucherovs RW spot in the top 6?
Really enjoyed your PDO article! I was wondering about players who sustain high shooting percentages in defiance of regression - at what point does it become possible to say with confidence that this outperformance is due to skill rather than luck, and that it's ~repeatable in following years? Like with McDavid shooting above 15% in four of his five seasons in the NHL
That's something I tend to play by ear. Some guys are just super selective or super accurate with their shooting. When I look for a regression candidate I'm looking for a big difference in results - like Zibanejad going from 12.7% to 12.7% to 19.7%.
How do the Florida panthers look after losing Dadanov and Hoffman while making a handful of big depth tweaks
Not amazing but I like Hornqvist at least. Bobrovsky could be a bounceback candidate as well. And they didn't do the stupid thing and trade Mackenzie Weegar, which is good.
Is there a way to calculate the error / confidence interval with Patrick Bacon's model? I know it was impossible with EvolvingWild because of how that's calculated. Just would be interesting to look at 2 players, one worth 2.0±0.2 WAR and another worth 2.0±2.0 WAR.
You would have to ask Patrick about that
What couple teams do you think will benefit from the new divisions? What cpl teams does it hurt the most?
Toronto, significantly. Possibly another few Canadian teams as well. And especially teams like Arizona, Minnesota, and the California teams in the Pacific who have a playoff spot up for grabs that might not be otherwise. Who's hurt is probbaly one of CBJ, DAL, NSH, or CAR.
Hello. I see that most analytical models project the capitals to take a significant step back from last year, even though it feels like they've improved in the areas that they struggled last year. Is it because of the age? Or is it something else im missing about the offseason
Age is probably a big part. Goaltending is also a big question mark since Samsonov just okay last year. I have the Capitals as a playoff team but I could also see them take a big step back. Adding Schultz is bewildering to me and playing him the top six is questionable. So is pushing Jonas Siegenthaler out of the lineup.
I wanna know what the fuck the Preds are doing. We didn't add any big FAs for a playoff push but we didn't trade anyone away for draft picks either. Hell, we traded AWAY draft picks for a bottom sixer. I liked the Askarov pick but where is this franchise going??
If I had to guess, I think they're rolling the dice on this roster to figure it out one more year. They've committed to being competitive by signing Duchene, and with Josi, Ellis, and Forsberg in their primes it makes sense. Their roster really isn't bad - I have them projected in the playoffs. The question is if they can put it together. I liked Benning and Borowiecki as the bottom pair, I like bringing in Haula and Granlund. The x-factor for this team is Ryan Johansen figuring out his crap.
Any idea what kind of banner I should look forward to seeing raised?
Could Mcdavid or MacKinnon average over 2 pts per game this season? Weaker opponents, strong supporting teammates. Seems a bit crazy, but perhaps worth some consideration.
I mean, it would be fun. It would take a pretty nuts scoring explosion but I wouldn't be upset to see it happen haha
What is Fiala's ceiling? Will he keep producing like late last year?
I would be surprised but it's not impossible. The Wild did well to buy low on him in 18-19 but his individual shooting was nuts this year. Seems more likely to me that he'll settle as a 55-65 point player but maybe he properly broke out as a sniper.
Is there anything under the hood that suggests Pavel Zacha might have a breakout year? Or is he what he is at this point? Thanks.
Unfortunately I don't see any reason to believe that Zacha is on track to be anything but a bottom six PK specialist. His underlying numbers have been pretty appalling so far. Maybe he'll change that this season but there's no foreshadowing yet.
One more (series of) question(s) - what are your thoughts on analytics in prospect evaluation? I think there is obvious value, but only when hedged with not just the eye test, but projection moving forward, which is something I haven't really seen from the big names. Your analysis of Drew Doughty, while not re: a prospect, was a masterclass in combining all three elements.
How much weight would you give analytics in prospect evaluation? Are you aware of anyone who you feel does this particularly well? Do you believe there is a way to build projection into a model with any degree of sophistication beyond simple production translation rates?
Amateur scouting with analytics basically takes what I do but tries to do something much much harder. When I write about a player (like Doughty), all I want to figure out is how good he is in his current league and what makes him good or bad. An amateur scout has to do that but then estimate how that will translate to playing against much better players, after development, oftentimes in a completely different country. It doesn't matter if player X is a 95th percentile QMJHL player, it matters how his skillset will do and evolve at the next level. Doing all that stuff within one league is hard enough without having to project outwards.
From what I can tell, prospect analytics are all about quantifying the eye test part of scouting. So the report doesn't just have to say "Player X is strong in transition" but they have actual evidence for it. You still have all the challenges of projection and translation and all that, but at least you have something concrete to point to. I am not confident that you could build a proper and predictive model for prospects using this data because you'd need to be able to translate microstats/corsi/etc. from league to league and the amount of data you'd need to get even part of the way there would be massive. But at least you can improve and sharpen your scouting reports.
In summary: i will never be a prospects guy and this is part of the reason why.
Thanks! Really appreciate it.
Hey there, I'm just a college student English major who quite likes statistics so I wanted to say the work you do is really educational in terms of the nature of predictive statistics. I spend an ungodly amount of time looking at the resources you put out and I feel like my understanding of probability and hockey have both dramatically I proved because of your hard work. So thank you for that! Also, I was wonder who are players that, in your experience, are the best at creating doubt in most predictive models. Players like Patrick Kane and Ovi, from what I've seen, generally don't fair too well in analytics circles relative to their perceived legacy. Do you think they actually bring more value than the models show or are they overhyped by the public? And if they are just overhyped, who do you feel is commonly undervalued by your data and that of your peers?
I really appreciate it! Predictiveness is really the killer for hockey stats and the trick is to try to be as "less wrong" as possible.
Kane and Ovi are tricky. The model I use, Patrick Bacon's WAR model, is much higher on Ovechkin and other snipers like Laine and Hoffman than the EvolvingWild model or others because it values individual goal-scoring ability more. Kane is also a great sniper but he's one of a small coterie of players who you could argue elevates the finishing of his linemates with passing. Those are the players I would say are undervalued offensively due to fundamental and unavoidable issues with xG models. This could also apply to players who are very good at generating offence off the rush or screening the goalie.
Of the Penguins' offseason additions, which are you most confident in having positive results?
I'm not overly confident in any of them. I have serious reservations about Kapanen, Matheson, and Ceci and I really hope that coaching can address those issues (esp. with Matheson. On the other hand Jankowski, Rodrigues, and Sceviour are all competent bottom sixers so hopefully they'll live up to that.
I have one more. Which front offices in the NHL do you think are the most analytically driven?
Pretty clear that the Leafs, Avalanche, and Hurricanes put a lot of emphasis on it, other than that it's not super easy to tell and I'm not even close to enough of an insider in this industry to have a sense for it.
Do you see sorokin overtaking varlamov as the isles starter? Also, what is sorokin ceiling in the NHL?
Don't know enough about Russian hockey but as a goalie you couldn't really ask for a better situation to step into the league than Korn and the Isles.
May you please rank the Sens cast of misfit impending UFAs (Coburn, Gudbranson, Paquette, Stepan, Galchenyuk, Anisimov, Reilly) in terms of how much interest you believe there will be at the deadline, and in terms of how much interest you would have (in a vacuum)?
I would have 0 interest in any of those players except for maybe Reilly for a late round pick. Only Stepan strikes me as a guy who might get some attention at the deadline and even then I see him as a 3rd-4th kind of player. Really do not like Ottawa's player evaluation this summer.
For the most part I agree. I liked the idea of a Stepan acquisition, but only contingent on a) them flipping Tierney elsewhere when he was unsigned and b) the price being very low. I think Paquette will provide some value, and should generate interest. Reilly I suppose can play a 7/8 role and plug in as an ok, but heavily sheltered, puck-mover. The rest of them are the epitome of "taxi squad", imo, but I have a feeling (perhaps a hope) that GMs around the league may not treat them as such at the deadline.
PS, thanks JFresh! You're the shit.
Top three players that you would have claimed in waivers
Depends on what kind of team I have. If I'm trying to win and have a bit of roster/cap space I like Derek Ryan, Nick Holden, and Mathieu Perreault, and Jordan Weal. If I have a young team I'd look at Ho-Sang and Balcers.
Who’s ur pick to go all the way, and who’s the media underrating/overrating
It's hard to bet against Colorado or Tampa Bay which means it will be somebody totally left field. I really think certain elements of the media have been overrating Ottawa's improvements this offseason - I think they are going to be worse than they were last season, maybe even by a lot. I also am not even close to sold on the Rangers. The only reason they even came close to a playoff spot last season was ridiculous puck luck for the Panarin line and Zibanejad and a borderline Norris-calibre rookie season from Adam Fox. I could easily see them take a step back.
In terms of underrating, my model likes the Jets and Ducks more than most others and definitely more than most of the media. The Jets were a 99 point team in 2019 and a legit contender before that. Their defence is rough but not as bad as last year, and adding a legit 2C is huge. I think they'll be competitive at least. I'm more skeptical about the Ducks but they're in a soft division and I like how their top 4 looks now. If Zegras or Steel can be a legit offensive threat they could fight for a spot.
As a Rangers fan this was tough to hear, and I kinda agree with you, but despite that I am still more excited for this year than I have been in a while.
Does Pettersson firmly establish himself as a top 5 Forward/Player in the game
Pettersson is a very interesting player and this season will be a good test of how good he actually is. Some WAR models rank him as legit a top 3 player already, while the one I use is a bit more hesitant to give him full credit for his on-ice finishing. If he can once again outperform on-ice expected goals to such a huge extent while having very strong playdriving results that would be a good sign for him. But if Miller or Boeser take a step back it could look like a decline in his own results when it's actually just the model correcting itself if that makes sense.
who are good sleepers for fantasy
I wrote a piece a few days ago about players whose individual and on-ice shooting percentages dropped, and that's usually the best bet. Dallas players like Seguin/Benn/Radulov for one. Also I'd check out goalies with decent pedigree who struggled last season - Bobrovsky could be good value since his results really weren't that bad after October.
Do you think that the Penguins' bottom-six is good enough to make the playoffs?
Yes I do. The question for me is whether it's good enough for them to contend for a Cup. Analytically speaking the pieces are all solid enough, it's just a matter of whether the puzzle pieces can fit together properly. Can Jankowski and McCann work well together despite being shoot-first defensive forwards? Can Lafferty bring a new element? Can Blueger and Tanev work well without having Aston-Reese as the forechecker? Will be interesting to see, especially if a guy like Kapanen finds himself down there eventually.