"True Possession" stat?

[Matthews316]

Is their a possession stat or metric that keeps track of the amount of possession, based on time with the puck, a team has when a certain player is on the ice? To go even further, splits based on time spent in the def/neu/off zones?

Click to read more...

 

[fiveonfive]

There isn't anything that is publicly available as far as I know. Current possession stats aren't really meant to be taken as actual possession either, that's just a colloquial name they were given in their infancy. A lot of people prefer not to call them that anymore so that people aren't confused between "true possession" and shooting metrics.

 

[Creativero]

I have a hard time understanding how anyone could think counting the number of times you willingly directing the puck to the opposition would be a good proxy for actual time possessing the puck. Watch a Manchester City game.

 

[Creativero]

If I had to guess the ratio of penalties for and against is probably the best proxy for actual time of possession.

 

[Creativero]

How count passes completed or touches? I bet turnovers and shots actually have a simlare relationship to actual time of possession.

 

[BoredBrandonPridham]

I have a hard time understanding how anyone could think counting the number of times you willingly directing the puck to the opposition would be a good proxy for actual time possessing the puck. Watch a Manchester City game.

It's not estimating time possessing the puck, it is estimating a ratio of having the puck v.s. without. And people can think it because I believe data has shown that there is some correlation. But it's really just looked at for lack of better data when it comes to possession.

How count passes completed or touches? I bet turnovers and shots actually have a simlare relationship to actual time of possession.

Does this data exist? If so, sure, is there correlation? If it doesn't exist yet and we're going to find a way to track possession, why not just track the amount of time the teams are in possession of the puck?

With turnovers and shots, the data becomes much more coarse. You get like 100 Corsi events to work with per game, but with shots you have half that. With turnovers you have even less.

If you think shots could be good, then I would expect Corsi to be much better.

 

[fiveonfive]

I have a hard time understanding how anyone could think counting the number of times you willingly directing the puck to the opposition would be a good proxy for actual time possessing the puck. Watch a Manchester City game.

Shooting metrics > actual possession. No one cares how accurate a proxy they are for possession, because they are more useful information than actual possession time would be.

 

[Henkka]

Is their a possession stat or metric that keeps track of the amount of possession, based on time with the puck, a team has when a certain player is on the ice? To go even further, splits based on time spent in the def/neu/off zones?

No, because that's one real advanced stat.

These shot distribution stats like Corsi are just made to fool some fans to believe something "advanced", which is very much flawed especially with individuals.

Puck control time + High danger Area scoring chances could be something.

 

[93LEAFS]

I have a hard time understanding how anyone could think counting the number of times you willingly directing the puck to the opposition would be a good proxy for actual time possessing the puck. Watch a Manchester City game.

Very different sports. A missed shot in soccer while occasionally resulting in a quality rebound generally results in lost possession. In general, shooting in hockey creates chaos that leads to high-percentage shots, or ever if the goalie freezes it, a 50/50% chance of regaining possession.

The reason Corsi is valuable over actually timing puck possession is because it measures aggressive possession. Remember Van Gaal's United? They held the ball most of the game but never had any intent.

 

[AlienWorkShop]

While it does slightly irk me Corsi and the like are casually referred to as a possession, wasn't there some Leafs blogger a couple years ago who actually used a stop timer of puck control for some Leafs games and showed it decently correlated with Corsi? Has that been repeated and/or refuted?

 

[seventieslord]

While it does slightly irk me Corsi and the like are casually referred to as a possession, wasn't there some Leafs blogger a couple years ago who actually used a stop timer of puck control for some Leafs games and showed it decently correlated with Corsi? Has that been repeated and/or refuted?

I don't know if it was just some leafs blogger, but yes, I think the whole point of using corsi is that it was demonstrated early on to correlate closely with actual possession to the point that it's a much less labour-intensive way to calculate possession that yields the same useful results.

 

[urho]

I don't know if it was just some leafs blogger, but yes, I think the whole point of using corsi is that it was demonstrated early on to correlate closely with actual possession to the point that it's a much less labour-intensive way to calculate possession that yields the same useful results.

On the other hand, stylistically there are some lines that cycle a lot and try to create chances by wearing the opponent down and then trying to find that high percent shooting area. It doesn't produce as many shots but it's possession that doesn't necessarily show up as high corsi numbers. Huberdeau - Barkov - Jagr-line from last year comes to mind as a perfect example.

 

[OvermanKingGainer]

On the other hand, stylistically there are some lines that cycle a lot and try to create chances by wearing the opponent down and then trying to find that high percent shooting area. It doesn't produce as many shots but it's possession that doesn't necessarily show up as high corsi numbers. Huberdeau - Barkov - Jagr-line from last year comes to mind as a perfect example.

Low shot attempts for, if you have the puck, will be compensated for by even lower shot attempts against. The line you're referring to had 54.51 CF60 and 51.63 CA60 which is positive corsi(51.36%), but not the dominating sort of a line like Marchand-Bergeron-Pastrnak, who are posting 73.86 CF60 and 40.02 CA60 for a remarkable 64.86% If your assertion were true, that the Barkov Jagr line spent all that much time in the offensive zone, their CA60 would be 40 as well. A 54.51 CF60 and 40.00 CA60 would be a 57.7% CF% which would be remarkable without them taking any more shot attempts.

 

[fiveonfive]

I don't know if it was just some leafs blogger, but yes, I think the whole point of using corsi is that it was demonstrated early on to correlate closely with actual possession to the point that it's a much less labour-intensive way to calculate possession that yields the same useful results.

That is actually incorrect. Corsi is better information than time of possession is. It's only called possession because of colloquial/casual naming, because it does vaguely track possession time, but the shot information is the key not the time of possession.

On the other hand, stylistically there are some lines that cycle a lot and try to create chances by wearing the opponent down and then trying to find that high percent shooting area. It doesn't produce as many shots but it's possession that doesn't necessarily show up as high corsi numbers. Huberdeau - Barkov - Jagr-line from last year comes to mind as a perfect example.

See above, high corsi is > than high time of possession

 

[ijuka]

Corsi's fine and all as long as it's just an output. When you start actually coaching with nothing in mind but maxing the amount of corsi actions you have, that's when it goes south fast. And that's the issue with the stat.

 

[Stephen]

Why they can't find a way to just put a transponder on players and into the puck and just let the data gather by itself? Stopwatches? Come on.

Just give us real data on the following:

Percentage of time spent in the offensive, neutral and defensive zone for each player.
Number of touches of the puck.
Time spent with puck on the stick while in possession of the puck.
Distance skated per shift and per game.
Peak pace, average pace, stops, starts, whatever.

 

[fiveonfive]

Corsi's fine and all as long as it's just an output. When you start actually coaching with nothing in mind but maxing the amount of corsi actions you have, that's when it goes south fast. And that's the issue with the stat.

Not sure what you mean. If a coach pays attention to how his team does with respect to the output, and then tried to maximize the situations in which the output improves, what is the problem?

I'm not sure there has ever been a team which has tried "maxing the amount of corsi actions", succeeded at doing so, and then still went "south fast". Not sure why so many people think this is a problem with corsi.

Why they can't find a way to just put a transponder on players and into the puck and just let the data gather by itself? Stopwatches? Come on.

Just give us real data on the following:

Percentage of time spent in the offensive, neutral and defensive zone for each player.
Number of touches of the puck.
Time spent with puck on the stick while in possession of the puck.
Distance skated per shift and per game.
Peak pace, average pace, stops, starts, whatever.

Yeah, it's weird that the NHL can't get those things tracked and available in this day and age. However, I would suspect that corsi might still pack more info than a lot of the stats you mentioned as an aggregate of "you must be doing many of those things well, if your corsi metricks are decent".

 

[Stephen]

Yeah, it's weird that the NHL can't get those things tracked and available in this day and age. However, I would suspect that corsi might still pack more info than a lot of the stats you mentioned as an aggregate of "you must be doing many of those things well, if your corsi metricks are decent".

I'd just rather see the raw data and make comments:

"I prefer Tavares to Stamkos because he holds the puck for an average of 3 minutes per game to 1 minute 20 seconds, which represents 18% of the total time his team has the puck versus 11%, and he has a 75 :15 : 10 ratio of time in each zone versus 90 : 5 : 5 with the puck compared with Stamkos which means he factors into the transition and is able to factor in on all three zones significantly better."

I want possession talk to be that kind of discussion.

 

[urho]

Low shot attempts for, if you have the puck, will be compensated for by even lower shot attempts against.

But that's not necessarily true because seeking the high percentage chance for a minute might not lead to a shot on goal (or maybe just one) and at the end of the shift opposition might get a shot from far out (or something like that). I fail to see how counting shot attempts is much more useful than +/- (Barkov was +18, Bergeron +12...). Both are more like banal stats for me rather than "advanced". It's not totally irrelevant but it's not that revealing either.

 

[fiveonfive]

I'd just rather see the raw data and make comments:

"I prefer Tavares to Stamkos because he holds the puck for an average of 3 minutes per game to 1 minute 20 seconds, which represents 18% of the total time his team has the puck versus 11%, and he has a 75 :15 : 10 ratio of time in each zone versus 90 : 5 : 5 with the puck compared with Stamkos which means he factors into the transition and is able to factor in on all three zones significantly better."

I want possession talk to be that kind of discussion.

I think that data would be interesting too. But I think the root of you dislike is the name of these stats more than their application. Just imagine it was called "pucks to the net efficiency" or something, because that is what it is. It's not actual time on stick possession.

 

[fiveonfive]

But that's not necessarily true because seeking the high percentage chance for a minute might not lead to a shot on goal (or maybe just one) and at the end of the shift opposition might get a shot from far out (or something like that). I fail to see how counting shot attempts is much more useful than +/- (Barkov was +18, Bergeron +12...). Both are more like banal stats for me rather than "advanced". It's not totally irrelevant but it's not that revealing either.

I think your biggest problem with not understanding shooting metrics is the fact that you don't realize that meaningful information comes from 10+ games of shot attempts, not one play. Yes, there are gonna be play's where a player technically displayed great possession skills (in terms of puck being on a stick) and then did not get rewarded for it. That would just be noise though, because over time that player will almost certainly have a favourable CF%.

Another major issue with your dilemma is that "Possession" stats are a lot less about time on stick possession (than most people seem to understand), and a lot more about the outcomes of possessions (than most people seem to understand). It is both! Think about it, what use is cycling the puck for 5 minutes if the other team ends up with more chances on your net at the end of the play?

 

[seventieslord]

Has anyone tried adjusting Corsi for score effects? I understand we have Corsi close, but then we have a smaller sample of data to go by and we toss out the data from when the deficit is 2+ goals. But couldn't we look at what typically happens to possession in these situations and adjust them individually? (example, if it's shown that teams are typically 10% better at corsi when down 2 goals compared to when close, then adjust all corsi stats by 10% each way for periods of time where teams that are up or down by 2 goals.)

For us to have a "true" possession stat I think we need all data, and I think we need more than just a full season's worth too. I'd like to see what Corsi would look like if adjusted for:

- Zone starts (relative to team of course)
- strength of competition
- strength of linemates
- quality of shots/chances (something that brings LA down to earth, for example, since it was shown that these "possession kings" produce lower quality shots)
- score effects

I think that all five of these things have an impact and these impacts are all measurable and can be adjusted for. It would be nice to see stats match up a little more with the eye test and with the opinions of NHL coaches who are paid to win hockey games. For example, Torey Krug and Shayne Gostisbehere lead NHL defensemen in relative Corsi, but they're sheltered offensive defensemen whose coaches don't trust them with the most minutes on the team, or the hardest minutes. They're clearly not even the best all-around defensemen on their own teams, let alone the league.

Thoughts?

 

[morehockeystats]

Corsi has one really big issue.
It's called "Goodhart's Law".

 

[seventieslord]

Corsi has one really big issue.
It's called "Goodhart's Law".

I didn't even know what this meant and had to look it up, but before I did, I suspected that it meant exactly what it does.

 

[Golden_Jet]

Corsi has one really big issue.
It's called "Goodhart's Law".

Good point, I've read some players are now directing the puck towards the net a few times a game, regardless if there is a play there, just to up their corsi, just like Goodhart said.