hockey analytics
-
- Posts: 529
- Joined: Thu Jun 18, 2015 10:51 am
hockey analytics
Are you pro or con analytics? or both? Do you still think the eyeball test is better? I really think the future of hockey is in advanced stats. It seems hockey is slowly moving in that direction. The leafs have hired an analytics guy. I think a few other teams have as well.
-
- Posts: 35732
- Joined: Tue Mar 24, 2015 2:06 pm
- Location: All things must pass. With six you get eggroll. No matter how thin you slice it, it's still baloney.
- Contact:
hockey analytics
paging @mikey and @slappybrown for a fight to the death in this thread.
-
- Posts: 29522
- Joined: Wed Mar 25, 2015 1:45 pm
- Location: “MIMH is almost always correct” -ulf
hockey analytics
it's a tool, not a full solution. I guarantee the Leafs are still going to be terrible
hockey analytics
In their current state, theyre virtually useless. Maybe they'll eventually figure something out.
-
- Posts: 42660
- Joined: Thu Mar 26, 2015 10:58 pm
- Location: More of a before-rehab friend...
- Contact:
hockey analytics
What we see are not analytics. That's the part that confuses some people. Some people got wind of the fad of shot counting and used it as the be-all, end-all...the golden key to success. When really, it was obviously not...or at least, not significantly different than what you would expect for shots on goal, which have been tabulated for about 100 years...
A lot of teams have actually brought it back full circle, combining the actual events on the ice in terms of "quality" and matching it up with the numbers "quantity" to try to make some headway. No one really does all of one (eye test) or all of the other (numbers game)...maybe Colorado, they don't mess with this stuff, but outside of that, every team has (and has had) people working on this kind of stuff...and they have different mechanisms in how they deal with things...
The teams that made big splashes about hiring analytics people were teams like New Jersey, Edmonton, Toronto, etc. and all were murdered all season long...in fact, Edmonton already canned their guy(s) I believe...because they fell too hard for it. Just like the "advanced" numbers before it like save/efficiency percentage or plus/minus, in the first couple years it's like, "wow, this...this is the bee's knees, ankles and feet..." and then once you actually stop and process what it is your actually measuring you realize, over time, that it's really a lot of noise with only minor insightfulness or practical use...you still have to revert back to scouting the player...
The golden ticket, CF% (basically what percentage of shots you take in a game) was not overly helpful vs. previous "unadvanced" numbers:
Top 12 in CF%
1. Winnipeg - no playoff games won.
2. Chicago - Champs
3. Los Angeles - Missed playoffs
4. Tampa Bay - Runner up
5. Pittsburgh - one playoff game won.
6. Detroit - out in the first round.
7. Nashville - out in the first round.
8. Islanders - out in the first round.
9. Dallas - missed playoffs.
10. Boston - missed playoffs.
11. San Jose - missed playoffs.
12. Carolina - missed playoffs.
Only two teams even made it out of the first round. In all, 7 playoff series wins. The bottom 12, won 4 playoff series. Four qualified in the bottom 12, 7 qualified in the top 12. You'd just expect more I feel like given that it has "predictive" abilities. The Conf. Finalists were 2, 4, 15, 19.
This is regular old SOG...
SOG regular season 2014-15:
1. Chicago
2. Islanders
3. Nashville
4. Pittsburgh
5. San Jose
6. Rangers
7. Dallas
8. Boston
9. Ottawa
10. Los Angeles
11. St. Louis
12. Minnesota
4 failed to qualify
5 failed to win a series
1 of the finalists, 2 of the Conf. finalists
19. Detroit
20. Washington
21. Philadelphia
22. Toronto
23. Arizona
24. Columbus
25. Montreal
26. Edmonton
27. Colorado
28. Calgary
29. New Jersey
30. Buffalo
Even straight up goal differential I believe had the Conf. Finalists listed as 1, 2, 4, 17. Schedule-adjusted goal differential does even better: With 11 of the top 12 teams in that qualifying for the playoffs (all top 11, Kings were 12). The Conf. Finalists worked out to 1, 3, 4, 13. While all of the bottom 12 teams failed to qualify for the postseason.
Like Edmonton went out to try to improve their Corsi - and I believe they did - but they sucked anyway. You can coach shot attempts. If you want to lead the league in shot attempts, you can. It's not the shot attempts that are valuable, it's the right ones that are. The narrative of the game comes first, the stats will always come second. Which isn't to say that they are wrong, or useless, or anything like that...but the utter dependency on them by some is concerning. Like that weenie that slanders Brandon Sutter's good name all the time...he had a very good season this year, there isn't a number on this planet that could convince me otherwise. I watched every single second of his season, he played very well. I'm not asking you to believe it, but I'm not gonna be convinced because he didn't shoot enough that he was bad...that's not the game...it never was, it never will be...
A lot of teams have actually brought it back full circle, combining the actual events on the ice in terms of "quality" and matching it up with the numbers "quantity" to try to make some headway. No one really does all of one (eye test) or all of the other (numbers game)...maybe Colorado, they don't mess with this stuff, but outside of that, every team has (and has had) people working on this kind of stuff...and they have different mechanisms in how they deal with things...
The teams that made big splashes about hiring analytics people were teams like New Jersey, Edmonton, Toronto, etc. and all were murdered all season long...in fact, Edmonton already canned their guy(s) I believe...because they fell too hard for it. Just like the "advanced" numbers before it like save/efficiency percentage or plus/minus, in the first couple years it's like, "wow, this...this is the bee's knees, ankles and feet..." and then once you actually stop and process what it is your actually measuring you realize, over time, that it's really a lot of noise with only minor insightfulness or practical use...you still have to revert back to scouting the player...
The golden ticket, CF% (basically what percentage of shots you take in a game) was not overly helpful vs. previous "unadvanced" numbers:
Top 12 in CF%
1. Winnipeg - no playoff games won.
2. Chicago - Champs
3. Los Angeles - Missed playoffs
4. Tampa Bay - Runner up
5. Pittsburgh - one playoff game won.
6. Detroit - out in the first round.
7. Nashville - out in the first round.
8. Islanders - out in the first round.
9. Dallas - missed playoffs.
10. Boston - missed playoffs.
11. San Jose - missed playoffs.
12. Carolina - missed playoffs.
Only two teams even made it out of the first round. In all, 7 playoff series wins. The bottom 12, won 4 playoff series. Four qualified in the bottom 12, 7 qualified in the top 12. You'd just expect more I feel like given that it has "predictive" abilities. The Conf. Finalists were 2, 4, 15, 19.
This is regular old SOG...
SOG regular season 2014-15:
1. Chicago
2. Islanders
3. Nashville
4. Pittsburgh
5. San Jose
6. Rangers
7. Dallas
8. Boston
9. Ottawa
10. Los Angeles
11. St. Louis
12. Minnesota
4 failed to qualify
5 failed to win a series
1 of the finalists, 2 of the Conf. finalists
19. Detroit
20. Washington
21. Philadelphia
22. Toronto
23. Arizona
24. Columbus
25. Montreal
26. Edmonton
27. Colorado
28. Calgary
29. New Jersey
30. Buffalo
Even straight up goal differential I believe had the Conf. Finalists listed as 1, 2, 4, 17. Schedule-adjusted goal differential does even better: With 11 of the top 12 teams in that qualifying for the playoffs (all top 11, Kings were 12). The Conf. Finalists worked out to 1, 3, 4, 13. While all of the bottom 12 teams failed to qualify for the postseason.
Like Edmonton went out to try to improve their Corsi - and I believe they did - but they sucked anyway. You can coach shot attempts. If you want to lead the league in shot attempts, you can. It's not the shot attempts that are valuable, it's the right ones that are. The narrative of the game comes first, the stats will always come second. Which isn't to say that they are wrong, or useless, or anything like that...but the utter dependency on them by some is concerning. Like that weenie that slanders Brandon Sutter's good name all the time...he had a very good season this year, there isn't a number on this planet that could convince me otherwise. I watched every single second of his season, he played very well. I'm not asking you to believe it, but I'm not gonna be convinced because he didn't shoot enough that he was bad...that's not the game...it never was, it never will be...
-
- Posts: 16580
- Joined: Wed Mar 25, 2015 4:19 pm
- Location: Lifelong Alabama Football Fan
hockey analytics
Oh man oh man oh man
First round of this on 5AF. I have so so so so so much to disagree with I mean respond to in @mikey post.
Thanks @count2infinity for the 911 page
First round of this on 5AF. I have so so so so so much to disagree with I mean respond to in @mikey post.
Thanks @count2infinity for the 911 page
-
- Posts: 529
- Joined: Thu Jun 18, 2015 10:51 am
hockey analytics
I follow a ton of guys on twitter that seem to know what they are talking about. Just seems like the way of the future.
-
- Posts: 29522
- Joined: Wed Mar 25, 2015 1:45 pm
- Location: “MIMH is almost always correct” -ulf
-
- Posts: 42660
- Joined: Thu Mar 26, 2015 10:58 pm
- Location: More of a before-rehab friend...
- Contact:
hockey analytics
They seem to, yes. And I'm sure they do, in their relative area of expertise.I follow a ton of guys on twitter that seem to know what they are talking about. Just seems like the way of the future.
-
- Posts: 30609
- Joined: Thu Mar 26, 2015 8:53 am
- Location: I have four degrees and am a moron. Don’t let that fool you
hockey analytics
Oh boy
-
- Posts: 19041
- Joined: Thu Mar 26, 2015 2:37 pm
- Location: people notice my car when its shined up
hockey analytics
From what we've seen on this team, I do think Corsi is an accurate way to measure posession. I do not know if posession strongly translates into W's and L's however.
hockey analytics
Oh man oh man oh man
-
- Posts: 16580
- Joined: Wed Mar 25, 2015 4:19 pm
- Location: Lifelong Alabama Football Fan
hockey analytics
I don't know what this means. Reasoning out things from the original "Pittsburgh had 27 shots tonite" to the current soup of Corsi, Fenwick, PDO, QoC, GVT, HERO charts, etc. is "analytics"; its nothing more than taking what we see on the ice and translating that into a record of what happened through various measures. That's "analytics" -- systemic analysis of information and data.What we see are not analytics. That's the part that confuses some people. Some people got wind of the fad of shot counting and used it as the be-all, end-all...the golden key to success. When really, it was obviously not...or at least, not significantly different than what you would expect for shots on goal, which have been tabulated for about 100 years...
Again, not sure what "bringing it full circle" means here; if you're suggesting, as the rest of your post seemed to, that teams went 100% analytics, then we've been over this countless times, and this is the same strawman that you keep bringing up. No one in analytics has advocated for a "one or the other" type program (indeed, the only people who have relflexively dismissed one side are traditionalists who think this stuff is nonsense and say things like "do you even watch the games" in response to the use of analytics).A lot of teams have actually brought it back full circle, combining the actual events on the ice in terms of "quality" and matching it up with the numbers "quantity" to try to make some headway. No one really does all of one (eye test) or all of the other (numbers game)...maybe Colorado, they don't mess with this stuff, but outside of that, every team has (and has had) people working on this kind of stuff...and they have different mechanisms in how they deal with things...
If you're talking about quality meaning quality of teammates, competition, relative Corsi/Fenwick -- that's not really new. You could go on behind the Net 5 years ago and find QoC and QoT stats.
I cannot believe you cited Toronto -- TORONTO!!! -- as an example of how analytics haven't helped teams. That team has been dog **** for a long time, and was actively proud of its stupidity in eschewing analytics in favor of doing things like signing facepunching meatbags like Colton Orr and Fraser McLaren (within months of one another no less!) in a seemingly direct middle finger to the nerds with their fancy math degrees. Burke used to go to the Sloan Conference and proudly tell the audience he'd yet to see a single analytic framework that he deemed valuable in evaluating talent. Nonis continued that trend. They explicitly built a team predicated on things like "grit" and "toughness" and good ole' Canadian boys without reference to the fact that most every move they made was contrary to what an analytics guy would have told them to do.The teams that made big splashes about hiring analytics people were teams like New Jersey, Edmonton, Toronto, etc. and all were murdered all season long...in fact, Edmonton already canned their guy(s) I believe...because they fell too hard for it. Just like the "advanced" numbers before it like save/efficiency percentage or plus/minus, in the first couple years it's like, "wow, this...this is the bee's knees, ankles and feet..." and then once you actually stop and process what it is your actually measuring you realize, over time, that it's really a lot of noise with only minor insightfulness or practical use...you still have to revert back to scouting the player...
They hire analytics guys like Kyle Dubas and Cam Charron last summer, and essentially you lay the blame for them being a sht team at their doorstep? After years of Brian Burke and Nonis openly being hostile to this stuff, its the one season where they finally hired the analytics guy? Let's assume those guys said "75% of this roster is terrible according to our analysis, we should sign these guys instead" -- how do you accomplish that in a partial offseason and one full season? The answer, of course, is you don't. Pointing to Toronto doesn't help your argument that this stuff isn't particularly useful, because they ran their team into the dumpster while ignoring analytics.
Re: the Oilers, Chiarelli already has said that Dellow (I assume you're referencing Dellow) will continue to work with them, moving more to the front office role he should have been in in the first place (as opposed to the primarily coaching role they gave him which made absolutely no sense because the Oilers are the only team more poorly-run than Toronto). He's an incredibly abrasive person based on his Twitter and his interactions with people on the internet (and according to Eakins), but he's also very talented, so he's not been "fired."
Comparing the analytics work being done by people with PhD's in mathematics from MIT to plus/minus or save % (the latter is much more valuable than you think, but that's a different discussion) is not something that anyone can take seriously.
I don't know who told you this was the "golden ticket", because it isn't. Also, where are you getting these numbers? These are wrong.The golden ticket, CF% (basically what percentage of shots you take in a game) was not overly helpful vs. previous "unadvanced" numbers:
Top 12 in CF%
1. Winnipeg - no playoff games won.
2. Chicago - Champs
3. Los Angeles - Missed playoffs
4. Tampa Bay - Runner up
5. Pittsburgh - one playoff game won.
6. Detroit - out in the first round.
7. Nashville - out in the first round.
8. Islanders - out in the first round.
9. Dallas - missed playoffs.
10. Boston - missed playoffs.
11. San Jose - missed playoffs.
12. Carolina - missed playoffs.
Only two teams even made it out of the first round. In all, 7 playoff series wins. The bottom 12, won 4 playoff series. Four qualified in the bottom 12, 7 qualified in the top 12. You'd just expect more I feel like given that it has "predictive" abilities. The Conf. Finalists were 2, 4, 15, 19.
But, let's use CF% for now. 10 of the top 16 teams in CF% made the playoffs. MIN/OTT, two more playoff teams, were 50%+ in CF. The remaining 4 teams -- VAN, NYR, MTL, and CGY -- are driven either by Hart/Vezina Trophy level goaltending (MTL literally, NYR substantively between Talbot and Hank), and the other two played in a sht division and defied the odds; great for them. Your conclusion from this is that Corsi is not useful? CGY in 2015-2016 is very likely to be COL in 2014-2015, unless they clone Giordano.
The Kings have won two Cups after finishing as the 8th seed and the equivalent of a 6th seed in their Conference. They were, as you might put it, the Corsi champs both years as well.
Analytics said COL was a mirage; sure enough, they were.
I could go on and on, but you can very easily pull up the prior seasons which, in comparison to 2014-2015 relatively outlier status in this respect (eg, LAK missing the playoffs and CGY getting in with the worst CF% in history) and see that score-adjusted Corsi/Fenwick are very good indicators of team success.
7 game series are often times coin-flips, huh?This is regular old SOG...
SOG regular season 2014-15:
1. Chicago
2. Islanders
3. Nashville
4. Pittsburgh
5. San Jose
6. Rangers
7. Dallas
8. Boston
9. Ottawa
10. Los Angeles
11. St. Louis
12. Minnesota
4 failed to qualify
5 failed to win a series
1 of the finalists, 2 of the Conf. finalists
19. Detroit
20. Washington
21. Philadelphia
22. Toronto
23. Arizona
24. Columbus
25. Montreal
26. Edmonton
27. Colorado
28. Calgary
29. New Jersey
30. Buffalo
Even straight up goal differential I believe had the Conf. Finalists listed as 1, 2, 4, 17. Schedule-adjusted goal differential does even better: With 11 of the top 12 teams in that qualifying for the playoffs (all top 11, Kings were 12). The Conf. Finalists worked out to 1, 3, 4, 13. While all of the bottom 12 teams failed to qualify for the postseason.
This is where you fundamentally misunderstand the use of Corsi. No one is trying to "improve" their Corsi; the point is to acquire players that have the puck more often than the opposition does (why that specific player does is where scouting is invaluable); Corsi is a way to quantify that. It is self-evident that if I have the puck more than you, particularly given the relatively thin margins in talent that separate NHL teams (and the fact that longer time frames mitigate the impact of luck), I will, over the long run, be more likely to score more goals than you will score against me. Corsi and Fenwick, particularly the score-adjusted variants, are proxies for possession. There are countless articles that will show you the work on this. Players will inherently attempt to create the best possible scoring position for themselves or their teammates (subject to score effects). Analytics have no impact on that fact, which is the core of the sport.Like Edmonton went out to try to improve their Corsi - and I believe they did - but they sucked anyway. You can coach shot attempts. If you want to lead the league in shot attempts, you can. It's not the shot attempts that are valuable, it's the right ones that are. The narrative of the game comes first, the stats will always come second. Which isn't to say that they are wrong, or useless, or anything like that...but the utter dependency on them by some is concerning. Like that weenie that slanders Brandon Sutter's good name all the time...he had a very good season this year, there isn't a number on this planet that could convince me otherwise. I watched every single second of his season, he played very well. I'm not asking you to believe it, but I'm not gonna be convinced because he didn't shoot enough that he was bad...that's not the game...it never was, it never will be...
Absolutely no one is trying to "coach shot attempts" and telling their players to shoot from the redline to generate positive Corsi (which it wouldn't even do), but you know that. Couching it like this is just a purposeful misrepresentation.
hockey analytics
After almost 3 weeks, I'm kind of let down. It's like Pitt football mid September.
-
- Posts: 16580
- Joined: Wed Mar 25, 2015 4:19 pm
- Location: Lifelong Alabama Football Fan
hockey analytics
lol, the show is going to lose its luster once its gets into syndication here at 5AFAfter almost 3 weeks, I'm kind of let down. It's like Pitt football mid September.
ETA: H2P
-
- Posts: 30609
- Joined: Thu Mar 26, 2015 8:53 am
- Location: I have four degrees and am a moron. Don’t let that fool you
hockey analytics
I don't like it.
-
- Posts: 16580
- Joined: Wed Mar 25, 2015 4:19 pm
- Location: Lifelong Alabama Football Fan
hockey analytics
I don't like it.
-
- Posts: 29522
- Joined: Wed Mar 25, 2015 1:45 pm
- Location: “MIMH is almost always correct” -ulf
hockey analytics
yeah, slappy waited too long to swoop in.
hockey analytics
@slappybrown
-
- Posts: 529
- Joined: Thu Jun 18, 2015 10:51 am
-
- Posts: 19041
- Joined: Thu Mar 26, 2015 2:37 pm
- Location: people notice my car when its shined up
hockey analytics
ryan wilson probably had a happy accident in his pants when that sutter trade was announced
-
- Posts: 30609
- Joined: Thu Mar 26, 2015 8:53 am
- Location: I have four degrees and am a moron. Don’t let that fool you
hockey analytics
Ryan Wilson. What a bum.
-
- Posts: 529
- Joined: Thu Jun 18, 2015 10:51 am
hockey analytics
why don't you like him?
-
- Posts: 42660
- Joined: Thu Mar 26, 2015 10:58 pm
- Location: More of a before-rehab friend...
- Contact:
hockey analytics
From what little I've read of his, it's very unclear if he understands the game...he may know numbers and hell, some of those numbers may hold value...maybe even most...I don't know or care in the context of this discussion...but you gotta know your product, you gotta know what you're applying it too...
A salesman has to know his product better than anyone...he has not shown that ability in my limited views...maybe others that are more devoted followers will say otherwise...
A salesman has to know his product better than anyone...he has not shown that ability in my limited views...maybe others that are more devoted followers will say otherwise...
Who is online
Users browsing this forum: No registered users and 27 guests