• Moderators, please send me a PM if you are unable to access mod permissions. Thanks, Habsy.

The Stats Thread

Re: Habs News & Notes: 2014-15 Season

TheCountofMonteCristo said:
How is that diff from corsi which will credit all on ice for shot for and punish shot against even if wasn't your guy who shot against or you who took shot?
Click to expand...

Simply put - the more data points you have available to you the more those factors even out over the long run.

You could say the same thing about SV% I suppose.
 
Re: Habs News & Notes: 2014-15 Season

UWHabs said:
I think corsi makes more sense because it has larger numbers. Every shift (or almost every shift) there's 1 or multiple shot attempts. But for actual goals, players are on for very few over the course of a game. So the randomness will even out quicker.

+/- to me is not a useless stat, but it's useless without context. If you say Hossa is -14, that tells me nothing. But if he was -14 when nobody else on the team was above -20, maybe that was good. I think relative +/- can be a non-useless stat, but people who just quote is as if it's a proxy for something else without mentioning the context should be shot.
Click to expand...

Beat me to it, nicely said.
 
Re: Habs News & Notes: 2014-15 Season

hockeylover said:
Simply put - the more data points you have available to you the more those factors even out over the long run.

You could say the same thing about SV% I suppose.
Click to expand...

Yes. That's also why over the course of a season, +/- starts to make a bit more sense, when applied with context of course.
 
Re: Habs News & Notes: 2014-15 Season

GrandWazoo said:
Yes. That's also why over the course of a season, +/- starts to make a bit more sense, when applied with context of course.
Click to expand...

It might make a bit more sense over the long haul, but it's still too small of a sample size in a single season IMO. And you see it in the results. When you look at both of our teams over the past 10 years, the best +/- players in a single season were Mark Fraser and Alex Kovalev, both +18.

I really think we've evolved way past +/- at this point. I was a little disappointed to open "The Stats Thread" and it be simply a post about +/-. You could look at so many more valuable numbers.
 
Maybe a few will find these useful/interesting. Taken from the Leaf board (up to date before the last game).

C T.Plekanec (32): 19gms, 7gls, 15pts, 46.3cf%, -2.4rel, 18:56 (82gms, 30gls, 65pts) --- 50.3fo%
W M.Pacioretty (26): 19gms, 8gls, 16pts, 51.8cf%, +5.6rel, 19:24 (82gms, 35gls, 69pts)
W A.Galchenyuk (20): 19gms, 5gls, 13pts, 46.6cf%, -1.9rel, 16:17 (82gms, 22gls, 56pts)

C D.Desharnais (28): 19gms, 2gls, 10pts, 48.0cf%, -0.2rel, 17:10 (82gms, 9gls, 43pts) ---- 55.1fo%
W P.Parenteau (31): 19gms, 5gls, 9pts, 52.7cf%, +7.2rel, 15:31 (82gms, 22gls, 39pts)
W B.Gallagher (22): 19gms, 5gls, 10pts, 46.2cf%, -2.8, 15:48 (82gms, 22gls, 43pts)

C L.Eller (25): 19gms, 5gls, 8pts, 55.2cf%, +7.5rel, 14:22 (82gms, 22gls, 35pts) ---- 54.5fo%
W J.Sekac (22): 12gms, 3gls, 5pts, 54.3cf%, +5.7, 12:13 (82gms, 21gls, 37
4pts)
W D.Weise (26): 16gms, 3gls, 8pts, 39.7cf%, -7.6rel, 10:33 (82gms, 16gls, 41pts)

C M.Malhotra (34): 19gms, 0gls, 0pts, 35.0cf%, -10.8rel, 11:37 (82gms, 0gls, 0pts) ---- 61.3fo%
W B.Prust (30): 19gms, 1gls, 4pts, 49.3cf%, +3.3rel, 12:5% (82gms, 4gls, 17pts)
W D.Bowman (25): 1gms, 0gls, 0pts, 0.0cf%, -45.2rel, 6:06 (82gms, 0gls, 0pts)

C M.Bournival (22): 2gms, 0gls, 0pts, 60.0cf% +12.0rel, 6:16 (82gms, 0gls, 0pts)
W R.Bourque (33): 13gms, 0gls, 2pts, 50.3cf%, +1.0rek, 12:20 (82gms, 0gls, 13pts)

D P.Subban (25): 19gms, 5gls, 12pts, 49.3cf%, +2.4rel, 25:02 (82gms, 22gls, 52pts)
D A.Markov (36): 19gms, 1gls, 9pts, 46.5cf%, -1.5rel, 24:54 (82gms, 4gls, 39pts)

D A.Emelin (28): 17gms, 0gls, 6pts, 48.2cf%, +0.4rel, 21:49 (82gms, 0gls, 29pts)
D T.Gilbert (32): 19gms, 1gls, 3pts, 46.4cf%, -2.0rel, 20:21 (82gms, 4gls, 13pts)

D S.Gonchar (40): 6gms, 0gls, 3pts, 47.9cf%, -2.3rel, 15:29 (82gms, 0gls, 41pts)
D M.Weaver (36): 18gms, 0gls, 3pts, 45.1cf%, -3.3rel, 15:43 (82gms, 0gls, 14pts)

D N.Beaulieu (22): 13gms, 0gls, 2pts, 51.9cf%, +5.5rel, 13:55 (82gms, 0gls, 8pts)
D J.Tinordi (22): 9gms, 0gls, 2pts, 48.1cf%, +1.1rel, 12:01 (82gms, 0gls, 18pts)

G C.Price (27): 15gms, .919sv%
G D.Tokarski (25): 5gms, .937sv%
 
Just to add a little explanation for those numbers.

Zeke is the one who compiles them and here's what he said about them.

zeke said:
1) yup. It's fairly simple. A player is on the ice for a certain number of corsi events (shots, missed shots, blocked shots) for and against. The corsi for percentage tells us that a player who is on the ice for 6 corsi events for his team and 4 against his team has a 60% corsi for percentage.

2) the cf relative I'm using is actually pretty complex, and measures how each player's corsi % compares to the teammates he's on the ice with AND also compares those teamates performances on and off the ice. It's not at all simple but it's a very nice stat to look at to try and separate one player's performance from the team's overall performance.

3) I'm not using any quality of competition, both because i've read enough convincing arguments that overall it has a tiny effect on these numbers, and also just looking at the numbers we get such a tiny difference in the numbers that i don't even think we'd learn anything at all from them. most every player is between 49-51% in opponents' CF%, so it's kind of a useless number to use. the zone starts thing seems to be a much more significant number that actually has a big impact on CF% - which is why I'm using Zone-start adjusted stats all the time, because IMO the stats are almost meaningless in terms of player comparisons without accounting for that factor somewhat at least.
Click to expand...
 
Re: Habs News & Notes: 2014-15 Season

hockeylover said:
It might make a bit more sense over the long haul, but it's still too small of a sample size in a single season IMO. And you see it in the results. When you look at both of our teams over the past 10 years, the best +/- players in a single season were Mark Fraser and Alex Kovalev, both +18.

I really think we've evolved way past +/- at this point. I was a little disappointed to open "The Stats Thread" and it be simply a post about +/-. You could look at so many more valuable numbers.
Click to expand...

I agree but I don't think +/- are as useless as people think. I also feel that corsi is only slightly better than +/-. One of the biggest flaws to it IMO is that it considers all shot attempts equally.

Some stats that I would be interested in :

Something like Corsi but for scoring chances
Scoring chances per player
Time of possession
Opposition time of possession
Time of possession per zone
Time of possession per player
Number of time a player had possession of the puck
Pass attempts
Passes completed
Possession loss (anytime a player lose possession of the puck, excluding shot attempts and pass attempts)
 
Fenwick is the better approximation of "scoring chances" than Corsi, and while still not perfect, it actually correlates pretty well

Corsi/Fenwick als correlate very highly with Time of Possession in the offensive zone, when you measure it with a stopwatch and compare to those numbers.
 
BKerr said:
Fenwick is the better approximation of "scoring chances" than Corsi, and while still not perfect, it actually correlates pretty well

Corsi/Fenwick als correlate very highly with Time of Possession in the offensive zone, when you measure it with a stopwatch and compare to those numbers.
Click to expand...

I still would like to see a real scoring chances stat, shots are only an approximation as many shots are not real scoring chances.

Also the time of possession per player, and what did he do with the puck during that time would be very interesting to me. How long is he able to hold on the puck? How often did he made a play (shot or pass) with the puck, versus losing it? Those would tell me a lot about a player.
 
I've been looking for pass attempt/pass completion/direct turnover -- basically treat every player like an NFL QB (to tell that part of the story)

BTW, I'm sure teams keep all types of stats that we don't have access to. All they need is a few computers and some eager fans to watch all the games for $15/hour. It's like scoring a baseball game for Bill James (which I did back in the day, live of course as there was no internet)
 
GrandWazoo said:
I still would like to see a real scoring chances stat, shots are only an approximation as many shots are not real scoring chances.

Also the time of possession per player, and what did he do with the puck during that time would be very interesting to me. How long is he able to hold on the puck? How often did he made a play (shot or pass) with the puck, versus losing it? Those would tell me a lot about a player.
Click to expand...

CH1 said:
I've been looking for pass attempt/pass completion/direct turnover -- basically treat every player like an NFL QB (to tell that part of the story)

BTW, I'm sure teams keep all types of stats that we don't have access to. All they need is a few computers and some eager fans to watch all the games for $15/hour. It's like scoring a baseball game for Bill James (which I did back in the day, live of course as there was no internet)
Click to expand...

Yes, the stats we have aren't perfect. I agree 100%.

And the types of stats you two want would be nice to know, and would likely help us. The more info we have is always a good thing in my mind... you just have to know how to use that info.

As more technology is available and more data recorded there are likely a bunch of new fancystats to come in the next few years that will take us beyond what we know right now.

All I'm saying is that with what is available today, corsi and fenwick, used in context with other stats... are good approximations for possession and scoring chances.

Fenwick is a good approximation for scoring chances as a % base. What you find is that in short runs shots at net and scoring chances may not line up. But over the long run (ie 82 game season), If your fenwick percentage is 55%, your scoring chance percentage will likely be pretty close to that same percentage.
 
I'm a stat guy so the more stats I can get, the better. I like Corsi and Fenwick (I think I like Fenwick more), zone starts % and quality of opposition (although this one is quite imperfect). They are a much needed upgrade on the traditional stats to help evaluate a player. Even +/- was an upgrade at one point to simply goals and assists. There is still ways to go though.
 
the quality of opposition seems to be bullshit to me as the quality is determined by someone saying what is quality iirc or it is based on a 1st, 2nd 3rd line designation which of course again is somewhat subjective
 
GrandWazoo said:
I'm a stat guy so the more stats I can get, the better. I like Corsi and Fenwick (I think I like Fenwick more), zone starts % and quality of opposition (although this one is quite imperfect). They are a much needed upgrade on the traditional stats to help evaluate a player. Even +/- was an upgrade at one point to simply goals and assists. There is still ways to go though.
Click to expand...

Yup. As far as hockey goes, the analytics are lagging way behind other sports. They've come a long way in just the last 2-3 years.

I used to ignore analytics because I used to have the same opinion a lot here have: I preferred the eye test. And to this day, I still use the eye test when it comes to certain players. So when analytics say that Benoit Pouliot is a really good player, sorry, I just can't agree with that. I've watched him play enough times to know that he's average at best. Anyway, a few years ago, there was a team that squeaked into the playoffs and were set to face the Vancouver Canucks in the first round. I didn't give them much of a chance, I thought Vancouver would lose one game at most. But then I saw a lot of the people who were into analytics talk about how that team had the best possession stats in the whole league and should be considered at the very least a darkhorse to win the Cup. All that made me think was, welp, another reason that fancy numbers nonsense will never beat the eye test. And then that team ended up sweeping the Canucks. Holy shit. Then swept the Blues. Holy shit. Then beat the Coyotes in 5. Is this real life??? The LA Kings then beat the NJ Devils in 6 games to end up having the most dominant playoff run in my lifetime. That's when I knew there was something legit to this and I got into it.
 
The eye test vs stats dichotomy is a canard... just a product of people's insecurities when they first learn about 'advanced' stats.

Stats, like words, are only useful if the reader knows how to interpret and gain knowledge from them.
 
GGpX said:
Yup. As far as hockey goes, the analytics are lagging way behind other sports. They've come a long way in just the last 2-3 years.

I used to ignore analytics because I used to have the same opinion a lot here have: I preferred the eye test. And to this day, I still use the eye test when it comes to certain players. So when analytics say that Benoit Pouliot is a really good player, sorry, I just can't agree with that. I've watched him play enough times to know that he's average at best. Anyway, a few years ago, there was a team that squeaked into the playoffs and were set to face the Vancouver Canucks in the first round. I didn't give them much of a chance, I thought Vancouver would lose one game at most. But then I saw a lot of the people who were into analytics talk about how that team had the best possession stats in the whole league and should be considered at the very least a darkhorse to win the Cup. All that made me think was, welp, another reason that fancy numbers nonsense will never beat the eye test. And then that team ended up sweeping the Canucks. Holy shit. Then swept the Blues. Holy shit. Then beat the Coyotes in 5. Is this real life??? The LA Kings then beat the NJ Devils in 6 games to end up having the most dominant playoff run in my lifetime. That's when I knew there was something legit to this and I got into it.
Click to expand...

Kyle Dubas described using advanced stats as a system of checks and balances and I sorta liked that description. If you're really off about something on your eye test, if any of your personal biases are getting in the way of a fair assessment, or the simple fact that you can't possibly have equal viewings of everyone in the league... numbers will usually help to catch it. But if the numbers say something that really isn't logical or doesn't correlate to what you're seeing (and a lot of the fancy stats guys have said stuff that is just bizarre to me) then I definitely think it's fair to reassess the validity of the stat.
 
You must log in or register to reply here.
Back
Top