Puck possession vs Shot attempts

[TOML]

I don't see why possession time would be any better than Fenwick. I mean, there could still be a bunch of teams who have the puck a lot but don't do much with it.

These simple stats don't really look at what's actually being done with the puck on each play, but what if the stat is defensive zone possession time, like what the Avs-related video outlined? That may indicate some solid defensive play. But then again, without looking at the context we don't really know. Same with Fenwick.

But if puck possession is generally viewed as a positive and if Fenwick is just being used as a proxy to determine puck possession, then yes I think actual possession time numbers would be better.

 

[The Mighty Duck Man]

These simple stats don't really look at what's actually being done with the puck on each play, but what if the stat is defensive zone possession time, like what the Avs-related video outlined? That may indicate some solid defensive play. But then again, without looking at the context we don't really know. Same with Fenwick.

But if puck possession is generally viewed as a positive and if Fenwick is just being used as a proxy to determine puck possession, then yes I think actual possession time numbers would be better.

That's exactly how it started out, and probably still is, but there are a select few so unconditionally in love with corsi/fenwick that they've deluded themselves into believing those stats are better than possession.

 

[Hivemind]

Please provide evidence that possession is a better predictor of winning hockey games than shot attempt-based metrics.

 

[TOML]

Please provide evidence that possession is a better predictor of winning hockey games than shot attempt-based metrics.

The video in the OP goes briefly into it, (Chicago, Boston, LA, StL, Colorado tops in d-zone possession time), but if you know of a source for possession time data, per zone, that would be a big help.

It's a pretty common thing to hear among players, coaches and GM's say that puck possession is a key to winning.

What I don't understand is rather than using proxies to determine possession, why not just use actual possession?


This guy goes on to ask some of the same questions.
http://www.tsn.ca/naylor-puck-possession-is-the-talk-of-the-nhl-1.106934

 

[Doctor No]

It's a pretty common thing to hear among players, coaches and GM's say that puck possession is a key to winning.

Of course it's a key to winning. Is it the most important key to winning? (That's a large part of the question being asked).

 

[charlie1]

What I don't understand is rather than using proxies to determine possession, why not just use actual possession?

Because possession data don't exist publicly. As far as I know, possession data (in the form of Zone Time) are only available from 1999/2000 through 2001/2002. So Corsi/Fenwick is used as a proxy.

 

[TOML]

Because possession data don't exist publicly. As far as I know, possession data (in the form of Zone Time) are only available from 1999/2000 through 2001/2002. So Corsi/Fenwick is used as a proxy.

This is a mystery. Apparently the data is available, otherwise how was the OP's video made? Did the personally use a stopwatch for all the games to determine the data? Or it's private data collected by the league? Or is he guessing?

 

[TOML]

Of course it's a key to winning. Is it the most important key to winning? (That's a large part of the question being asked).

Well, consider the ultimate puck possession scenario where the losing team literally never has the puck: The winning team wins every faceoff, completes every pass, wins every puck battle and pops it into the net with ease because of this ultimate ability. I think it's safe to say that puck possession is easily the most important key to winning. But this concept isn't novel.

I guess the question is whether possession time is more important than shot-related proxies at determining possession. Need to see the actual data for that determination, but I suspect that time data would be very, very useful.

 

[Doctor No]

Well, consider the ultimate puck possession scenario where the losing team literally never has the puck: The winning team wins every faceoff, completes every pass, wins every puck battle and pops it into the net with ease because of this ultimate ability. I think it's safe to say that puck possession is easily the most important key to winning. But this concept isn't novel.

Okay, consider the ultimate shot attempt differential scenario where the losing team never gets a shot on goal.

How does that play out differently than the example that you describe?

Boundary cases are always interesting, but you have to be careful with how you apply them.

 

[TOML]

Okay, consider the ultimate shot attempt differential scenario where the losing team never gets a shot on goal.

How does that play out differently than the example that you describe?

Boundary cases are always interesting, but you have to be careful with how you apply them.

Because to get a shot you first need the puck. Losing the faceoff means never getting the puck back in the ultimate puck possession scenario.

 

[Doctor No]

Because to get a shot you first need the puck. Losing the faceoff means never getting the puck back in the ultimate puck possession scenario.

I understand that (and it seems like you agree that in the extreme, the two scenarios produce the same result).

"Be careful with boundary conditions" is what I'm saying.

 

[charlie1]

This is a mystery. Apparently the data is available, otherwise how was the OP's video made? Did the personally use a stopwatch for all the games to determine the data? Or it's private data collected by the league? Or is he guessing?

The NHL doesn't release it publicly. That video was made by the NHL network. The rest of us are constrained to the public data.

 

[Sojourn]

I think both have their place, and the importance of it can vary from team to team. Possessing the puck doesn't mean you're going to take every shot opportunity you get, and shooting the puck every time it's on your stick doesn't mean you're possessing the puck more than the other team.

Some teams seem to rely on shooting from everywhere, and going hard to the net. Other teams seem to hold the puck more, and look for the higher quality chances. For both of these teams, puck possession and shot differential is going to be important, but to varying degrees. Both systems can be effective, so there is no right or wrong way. What matters is winning the game. It's also worth noting that a scoring opportunity doesn't always result in a shot on goal, and a shot on goal doesn't always mean there was a scoring opportunity

I don't think either stat, alone, is really going to give you an accurate picture.

 

[TOML]

The NHL doesn't release it publicly. That video was made by the NHL network. The rest of us are constrained to the public data.

Perhaps they should. Fans are constantly thirsty for more stats to mull over. And it's not like anything shocking will be revealed. Good teams/players are good at keeping the puck? You're going to know this when just watching the games.

 

[The Mighty Duck Man]

Please provide evidence that possession is a better predictor of winning hockey games than shot attempt-based metrics.

Corsi was developed as a proxy for puck possession, Fenwick evolved from it to credit shot blocking as a skill, and both are referred to "possession metrics". So, considering that's exactly what they are, tell me why a proxy for possession would be a better tool than possession itself.

Don't delude yourself. Most would love actual possession numbers and are simply using the easiest, most readily available substitute.

 

[Doctor No]

Corsi was developed as a proxy for puck possession, Fenwick evolved from it to credit shot blocking as a skill, and both are referred to "possession metrics". So, considering that's exactly what they are, tell me why a proxy for possession would be a better tool than possession itself.

Sometime life's funny. It's not intuitive that shots would predict future performance better than goals, but it does.

If you have the proof that seems obvious to you, please post it. The entire point of this subforum lies in replacing impressions with facts, no matter how obvious those impressions may be.

Telling others not to "delude" themselves isn't particularly helpful.

 

[The Mighty Duck Man]

Sometime life's funny. It's not intuitive that shots would predict future performance better than goals, but it does.

If you have the proof that seems obvious to you, please post it. The entire point of this subforum lies in replacing impressions with facts, no matter how obvious those impressions may be.

Telling others not to "delude" themselves isn't particularly helpful.

The entire theory behind Corsi and it's correlation with forecasting has everything to do with possession of the puck. Otherwise, why not just use shots on goal, as if Corsi isn't measuring puck possession, then missed and blocked shots are highly irrelevant.

Obviously I can't provide the "proof" that actual time of possession is a better measure for possession than Corsi, as those numbers unfortunately aren't available. That doesn't mean we should eschew common sense.

Anyway, it is a delusion, and I would say it's very helpful to call it out. Many refuse to accept that Corsi is flawed and will not always be right, and choose instead to follow it blindly. In this case, to the point of believing it's a better measure than the thing it's substituting for. In my opinion, types like that aren't much better than the Steve Simmons' of the world. They deserve to be called out on it.

 

[Doctor No]

Anyway, it is a delusion, and I would say it's very helpful to call it out.

Maybe it's a delusion, and maybe it's not.

Anyone who believes that Corsi is perfect shouldn't be listened to.

On the other hand, that's not what is being argued in this thread. As many people as there are in the world who believe Corsi is flawless, there seems to be a far greater number who like to ascribe that belief to anyone who disagrees with them (usually accompanied by terms such as "advanced stats fans" or whatnot). Those who do that are far from being "helpful".

 

[Hivemind]

Corsi was developed as a proxy for puck possession, Fenwick evolved from it to credit shot blocking as a skill, and both are referred to "possession metrics". So, considering that's exactly what they are, tell me why a proxy for possession would be a better tool than possession itself.

Don't delude yourself. Most would love actual possession numbers and are simply using the easiest, most readily available substitute.

The game summaries exist for some old seasons, and they contain zone time. Show me that zone time correlates more heavily with winning than shot attempt-based metrics. Otherwise you're just huffing and puffing.

 

[Mubiki]

Corsi was developed as a proxy for puck possession, Fenwick evolved from it to credit shot blocking as a skill, and both are referred to "possession metrics". So, considering that's exactly what they are, tell me why a proxy for possession would be a better tool than possession itself.

Don't delude yourself. Most would love actual possession numbers and are simply using the easiest, most readily available substitute.

Because reducing it to shots removes a lot of pointless clutter. Possession is meaningless unless you are getting shots and denying shots.

Time of possession matters more in sports like soccer and football, where there is a legitimate attempt to fatigue your opponent.

 

[TOML]

The entire theory behind Corsi and it's correlation with forecasting has everything to do with possession of the puck. Otherwise, why not just use shots on goal, as if Corsi isn't measuring puck possession, then missed and blocked shots are highly irrelevant.

Obviously I can't provide the "proof" that actual time of possession is a better measure for possession than Corsi, as those numbers unfortunately aren't available. That doesn't mean we should eschew common sense.

Anyway, it is a delusion, and I would say it's very helpful to call it out. Many refuse to accept that Corsi is flawed and will not always be right, and choose instead to follow it blindly. In this case, to the point of believing it's a better measure than the thing it's substituting for. In my opinion, types like that aren't much better than the Steve Simmons' of the world. They deserve to be called out on it.

I think we're in agreeance that if this simple possession time data were made as available as corsi and fenwick is then it would be very beneficial to those who want to follow the game more closely without actually having to watch every second of every game. From what I read, this is essentially the purpose of looking at advanced stats to begin with.

There is no reason to be biased towards using one set of data or another. It's data, it's being collected, and I think it should all be made publicly available. Why not? It would only create more fan interest in the games. Those who already look at corsi and fenwick numbers per game should definitely be thrilled.

 

[Doctor No]

There is no reason to be biased towards using one set of data or another. It's data, it's being collected, and I think it should all be made publicly available. Why not? It would only create more fan interest in the games. Those who already look at corsi and fenwick numbers per game should definitely be thrilled.

I'd love to see more of the data be made public.

As I understand it, the data that was originally made public by the league was then pulled back because they were worried that it would be used in salary negotiations. I'm not sure if I can see any reason why those fears would have become reduced in the years since.

 

[Hammer Time]

The entire theory behind Corsi and it's correlation with forecasting has everything to do with possession of the puck. Otherwise, why not just use shots on goal, as if Corsi isn't measuring puck possession, then missed and blocked shots are highly irrelevant.

Corsi and Fenwick may originally have been developed as a possession proxy, but I also see Fenwick being used as a proxy for scoring chances. Obviously, some adjustment is going to be required to compensate for teams e.g. Boston who are good at keeping shots against to the outside. And yes, shots on goal might be even better for a scoring chance proxy. (Although some will say that if you leave someone open 10 feet in front of the net and they blow it, you should still get dinged ... so there's a case that missed shots should still be counted)

I know Extra Skater last season had a stat for SOG% which revealed, for instance, that Anaheim's low Corsis were largely misleading because they did a lot of shot-blocking.

Obviously I can't provide the "proof" that actual time of possession is a better measure for possession than Corsi, as those numbers unfortunately aren't available. That doesn't mean we should eschew common sense.

I don't think we're debating that possession time measures possession time better than Corsi, that's sort of obvious. I'm just not sure that having offensive zone possession time is the most important factor in winning hockey games - it's definitely important, but you still have to do something with the possession.

 

[TOML]

I'd love to see more of the data be made public.

As I understand it, the data that was originally made public by the league was then pulled back because they were worried that it would be used in salary negotiations. I'm not sure if I can see any reason why those fears would have become reduced in the years since.

Given the amount of insane contracts being handed out already, I'm not sure how this information being made public would affect anything negatively in that regard.

Besides, 'market value' has trumped any stats-related justification of salary when push comes to shove. "He would've been paid more on the open market." etc.

 

[Canadiens1958]

Common

I'd love to see more of the data be made public.

As I understand it, the data that was originally made public by the league was then pulled back because they were worried that it would be used in salary negotiations. I'm not sure if I can see any reason why those fears would have become reduced in the years since.

Common. Believed to be the reason why the original SV% data published weekly during the 1956-57 season into 1957-58 was pulled back. Also the format for the original +/- did not reflect individual responsibility but a units (5 players) role, thereby leveling the impact on salaries.