Stats For The 21st Century?

[fourtrax]

http://www.montrealgazette.com/sports/Futu...2587/story.html

Be curious to see what stats heads like Moneypuck and roy_133 have to say about this.

Personally, I'm happy to see any evidence that the league is moving out of the dark ages with regards to statistics.

[Mathradio]

And how are we supposed to calculate Corsi +/-? Corsi +/- is but one of the predominant hockey stats of the future.

Sabermetrics will be a term that could probably cover sports statistics, not just baseball ones.

[SteenIsThaFuture]

http://www.montrealgazette.com/sports/Futu...2587/story.html

Be curious to see what stats heads like Moneypuck and roy_133 have to say about this.

Personally, I'm happy to see any evidence that the league is moving out of the dark ages with regards to statistics.

Seeing as the author is a friend of mine and I helped him with the article, not much to say

That article covers more mainstream feelings towards advanced metrics. I think I should have something up soon, but not sure covering things in more details. I can assure you though weep there's nothing in there that I don't deal with on a regular basis

[mtl1010]

And how are we supposed to calculate Corsi +/-? Corsi +/- is but one of the predominant hockey stats of the future.

yes, can someone explain how to calculate it???

I'd like to know why hossa would be the best player if this stat was used

[SteenIsThaFuture]

yes, can someone explain how to calculate it???

I'd like to know why hossa would be the best player if this stat was used

(Shots on net+ Shots blocked+ shots missed all while on the ice)- (Shots on net+ Shots blocked+ shots missed all while off the ice)/60

Can be found at behindthenet.ca

Here's my aforementioned piece from the Journal on the issue:

http://communities.canada.com/edmontonjour...ats-part-2.aspx

What are the problems with hockey stats right now?

Hockey stats at the moment lack two main things, details and isolation. Details in the sense that when looking at stats like points, or plus-minus in a box score it doesn’t tell you next to anything about how the player performed over the duration of the game as those stats only describe single events in a sport where there are many more events. On many occasions, the player’s actual performance will vary from the story that the simple statistics dictate.

The game of hockey may be determined in goals which come from shots, but it certainly does not revolve around goals and shots. There is so much more that contributes to the basic flow and movement of a hockey game and the stats just don’t capture this. They just capture the end results but not the process. The results are important, but they are determined by the process and not understanding how you got your results is a major hole in the data.

Isolation is the main problem now in my opinion which means isolating exactly what the player did. Goals and Assists are not as big a problem, but it still makes the assumption when totaling them at into a player’s final stats that all goals and assist are created equal which is nowhere near the truth. Plus-Minus is horrid with isolation, as it assume every player contributed equally to the goal for or against and awards them accordingly and that also is a horrid and completely false assumption on nearly every goal.

For goalies, they’re tracked for saves and goals let in, but external factors are not taken into account. Where did the shot come from? Was there a screen? Was the shot stoppable or was it a very high-difficulty save? Modern-day stats don’t capture that, it’s merely “Save” or “Goal”.

What are major developments in stat analysis?

Firstly people have discovered that in order to rid ourselves of some of the noise in Plus-Minus, people have started using a stat called Corsi which is a Plus-Minus but for shots instead of goals. The gains here are increasing your sample size to nullify statistical sampling issues and it takes the goalie out of the equation. There are still sample issues, but it’s vastly less so than with basic Plus-Minus.

Another development has been that statisticians have found there is a correlation on a player’s statistics depending on how many times they start a face-off in the offensive or defensive zone. Essentially stating a coach can influence a player’s raw statistical output by where and when he starts the play, a stat called Zone Starts.

Gab Desjardins at Behind the Net, has been able to make first-order attempts at trying to identify the opposition and linemate help/hindrance a player may have.

For goaltending and shooters, we’ve identified that, not shockingly, shot location has a large effect on the chances of a shot going in. It’s contributed to basic shot quality models, but that is still a ways off. We have seen though that all shots are not created equal and there is at least the chance goalie save percentages can be misrepresented, which is a big step.

Lastly, hockey statisticians have taken steps to try and create an all-encompassing stat to try and fully evaluate a player’s contributions on the ice. In simpler terms, the ability to say how many points in the standings/goals the team would or would not have without you. The two major references are Tom Awad’s Goals Versus Threshold used thoroughly at Puck Prospectus, and Alan Ryder’s Player Contribution at Hockey Analytics.

What will we see in the future with stat analysis?

Well at the moment we’re in a standstill with hockey stats. We have very smart people doing everything they can to uncover new things, but we just don’t have much to work with. Baseball had this same issue until they started tracking pitch velocity and location, tracking where the ball was being hit and just recently seeing how hard a ball was hit. This is called the Pitch/Hit F/X technology and it is so advanced and so helpful that some MLB teams hire people solely to analyze that data.

At the moment in hockey we have something called the RTSS (Real Time Scoring System) where a scorekeeper punches in things like shot location, shot type, giveaways/takeaways, missed/blocked shot and more. Problem with this is nearly all of those are prone to subjective bias and in some cases the bias is significant whereas the F/X technology is computerized and objective.

The answer is being able to expand the data we have and that won’t come until there is a demand. It started in baseball when people went to games and manually recorded things, something called Project Scoresheet and we’re seeing something similar with scoring chances and with you Defensive Errors being tracked by amateurs. Plenty of NHL teams do this as well, but on their own so I can see if people start pushing for more and new statistics, that eventually the league will respond and produce it. The obstacle here is finding a way to overcome the bias factor, which at the moment I’m not sure we can.

[mtl1010]

Thank you MP

[Mathradio]

Oh? Maybe the Corsi +/- could be applicable to soccer as well...

[ChiLla]

Oh? Maybe the Corsi +/- could be applicable to soccer as well...

What's the point? You don't have line changes in soccer, meaning the on the ice/off the ice part of the stat is rendered irrelevant.

There's so much more to soccer that shots on goal and blocked shots, it doesn't make much sense to track those stats in isolation (it's being done but the actual value of those stats is extremely limited). As a player, you can absolutely dominate without recording a single shot on goal, let alone scoring or assisting on one, which makes the game so interesting IMO. Go and watch guys like Iniesta, Xavi or Schweinsteiger for example, they're capable of controlling the pace of the game and influence the way how it is played by their teammates and opponents alike, yet you'll rarely find their names on the scoresheet. Passing lanes and keeping possession are key and they absolutely excel in that department.

[Mathradio]

Please forgive me. I know very little about soccer but since they keep track of shots on goal and shift individual players during games, the Corsi +/- would still be calculable.

Now I recognize that it has more value in hockey, though...

[SteenIsThaFuture]

What's the point? You don't have line changes in soccer, meaning the on the ice/off the ice part of the stat is rendered irrelevant.

There's so much more to soccer that shots on goal and blocked shots, it doesn't make much sense to track those stats in isolation (it's being done but the actual value of those stats is extremely limited). As a player, you can absolutely dominate without recording a single shot on goal, let alone scoring or assisting on one, which makes the game so interesting IMO. Go and watch guys like Iniesta, Xavi or Schweinsteiger for example, they're capable of controlling the pace of the game and influence the way how it is played by their teammates and opponents alike, yet you'll rarely find their names on the scoresheet. Passing lanes and keeping possession are key and they absolutely excel in that department.

Well if their ball possession skills lead to shots by teammates while on the pitch even if not by them it would apply to their Corsi.

That was a point I made was the flow of a hockey game goes beyond shots and the flow/puck movement is a major part of hockey not captured in the stats.

[ChiLla]

Please forgive me. I know very little about soccer but since they keep track of shots on goal and shift individual players during games, the Corsi +/- would still be calculable.

Now I recognize that it has more value in hockey, though...

No worries, it's all good. Corsi may work for hockey but it's of very limited value for soccer. Sure you can substitute a player but it's not even remotely comparable to rolling line changes in hockey, especially since substitutions are limited in terms of numbers and they largely depend on the current score in a given game. It's not uncommon to take a forward off the field to add another defender when trying to protect a lead towards the end of a game. However, you won't see Cammalleri benched in favor of Hal Gill because our head coach decides to go with 3 defenders and 2 forwards with 10 minutes left in order to protect a 1 goal lead.

Eeehhh... wait. Maybe I shouldn't post this. Just in case Jacques is reading along

Well if their ball possession skills lead to shots by teammates while on the pitch even if not by them it would apply to their Corsi.

That was a point I made was the flow of a hockey game goes beyond shots and the flow/puck movement is a major part of hockey not captured in the stats.

In fact they do. Directly as well as indirectly. I understand your point but the whole /60 (or /90 in case of soccer) wouldn't really work, since the whole substitution concept is entirely different or am I missing something? On top of that, blocked shots in soccer aren't a major factor really. Sure, a defender may be able to perform a great shot block here or there but it's really not that common.

[SteenIsThaFuture]

In fact they do. Directly as well as indirectly. I understand your point but the whole /60 (or /90 in case of soccer) wouldn't really work, since the whole substitution concept is entirely different or am I missing something? On top of that, blocked shots in soccer aren't a major factor really. Sure, a defender may be able to perform a great shot block here or there but it's really not that common.

Then you go to something called Fenwick which is Corsi without the blocked shots

[ChiLla]

Then you go to something called Fenwick which is Corsi without the blocked shots

Mhhh... it's entirely possible this could work for soccer. Still, I'm not sure what the correlation to wining would be. I don't want to sound ignorant and you know I really appreciate your work, but IMO soccer and hockey are so different, I doubt using advanced metrics from one sport in the other holds much water. Pass completion ratios, one-on-one duels won/lost (tackles, headers, dribbling), possession percentages, fouls committed vs. clean tackles etc. go a longer way than shots on goal/missed/blocked IMO. The funny part is you could still lose the game, even if your team is dominating in every possible aspect yet fails to score a goal.

[SteenIsThaFuture]

Mhhh... it's entirely possible this could work for soccer. Still, I'm not sure what the correlation to wining would be. I don't want to sound ignorant and you know I really appreciate your work, but IMO soccer and hockey are so different, I doubt using advanced metrics from one sport in the other holds much water. Pass completion ratios, one-on-one duels won/lost (tackles, headers, dribbling), possession percentages, fouls committed vs. clean tackles etc. go a longer way than shots on goal/missed/blocked IMO. The funny part is you could still lose the game, even if your team is dominating in every possible aspect yet fails to score a goal.

Well Desjardins did some in-depth soccer analysis, showing a Corsi/Fenwick wouldn't work because shot sample is too low

http://www.behindthenethockey.com/2010/6/1...pected-goals-in

And that territory/ball possession is the most correlative to winning:

http://www.behindthenethockey.com/2010/8/1...lling-territory

I'd argue puck possession is the most correlative thing to winning in hockey too, just due to the nature of the game more possession usually results in shots.