All Purpose Analytics and Extended Stats Discussion

[CapitalsCupReality]

Every study comparing other means of puck possession data to shot attempt metrics has concluded that there is little difference between the two.

Links to these "studies"?

 

[caps8]

4th best Corsi/Fenwick =/= 4th best possession player. I really wish people would stop using those terms interchangeably.

Not only is there negligible difference, shot attempts are far more relevant than zone time it has been proven. A team can have it in the other end forever but the point is to get pucks on net and that's what shot attempts measure.

http://www.pensionplanpuppets.com/2012/7/15/3161638/measuring-puck-possession-the-end-part-1

 

[Xaroc]

Not only is there negligible difference, shot attempts are far more relevant than zone time it has been proven. A team can have it in the other end forever but the point is to get pucks on net and that's what shot attempts measure.

http://www.pensionplanpuppets.com/2012/7/15/3161638/measuring-puck-possession-the-end-part-1

This all day. Corsi and Fenwick are the best two measures of puck possession we have. You can talk about other things like quality of competition and quality of teammates and zone starts (although these are somewhat of a red herring in most situations) but when Semin is on the ice the puck is headed towards the opponents net more than his. Does he make some mistakes? Sure, but all players do that from time to time.

 

[Xaroc]

Prompted by this tweet:



I took a look here and found some interesting stuff:

http://war-on-ice.com/teamtable.htm...05&splitseasons=1&usedaterange=0&tablegroup=4

Caps are playing at what looks like a really low pace. 3rd lowest in the league. However, they are generating attempts at the 6th highest in the league. So what is driving this low pace is the fact they are number 1 in suppressing attempts in the entire NHL. Oh and to top it off Caps have the highest score adjusted CF% in the league right now at 56.1%. That right there is championship caliber possession.

So in other words they are really doing what you want to be doing. Both generating a ton of shot attempts and holding other teams well in check.

You can check out the scoring chance data as well and it doesn't look quite as stellar but Caps are still 4th highest in SCF%, 10th highest in pace, 6th highest in SCF/60, 8th lowest in SCA/60.

 

[Hivemind]

Wasn't it proven that score adjustments led to less predictive stats? Just going with the raw numbers is better.

 

[caps8]

Wasn't it proven that score adjustments led to less predictive stats? Just going with the raw numbers is better.

I think you're thinking of the "close" possession which left out a lot of data and thus, score-adjusted numbers were introduced to use all of the non-close data in a meaningful way.

 

[Xaroc]

Wasn't it proven that score adjustments led to less predictive stats? Just going with the raw numbers is better.

http://www.hockeyprospectus.com/bur...ose-vs-score-adjusted-metrics-for-prediction/

Here is the TLDR quote:

In the end it appears that score and venue adjustments at 5v5 capture enough important information that they drastically improve on raw Corsi or Fenwick when it comes to predicting future goals and wins (by approximately 5% around the 20 game mark).

 

[NobodyBeatsTheWiz]

Every study comparing other means of puck possession data to shot attempt metrics has concluded that there is little difference between the two.

Nope. Not on an individual level. Not a single study has shown that a player with a good Corsi/Fenwick is a good possession player. It only indicates what the players team does while he's on the ice.

 

[NobodyBeatsTheWiz]

Not only is there negligible difference, shot attempts are far more relevant than zone time it has been proven. A team can have it in the other end forever but the point is to get pucks on net and that's what shot attempts measure.

http://www.pensionplanpuppets.com/2012/7/15/3161638/measuring-puck-possession-the-end-part-1

Shot attempts are far more relevant in what context? There are specific tactics coaches employ to cycle the puck in the corners. That's more relevant to them than shot attempts in those contexts.

This is my problem with #fancystats. People attribute too much real-world meaning to them.

Sure, over a large sample size, teams that have good Corsi/Fenwick tend to possess the puck more. But those metrics are like +/-. They give little context for individual players.

"Good possession player" is a subjective evaluation of what a player does with the puck, and it can vary from person-to person. For me, a good possession player is one who protects the puck well, passes accurately, and doesn't turn it over in dangerous areas. Has absolutely nothing to do with shot attempts.

 

[ChibiPooky]

I think you're thinking of the "close" possession which left out a lot of data and thus, score-adjusted numbers were introduced to use all of the non-close data in a meaningful way.

Ah, this is what I was thinking about.

 

[Xaroc]

Sure, over a large sample size, teams that have good Corsi/Fenwick tend to possess the puck more. But those metrics are like +/-. They give little context for individual players.

"Good possession player" is a subjective evaluation of what a player does with the puck, and it can vary from person-to person. For me, a good possession player is one who protects the puck well, passes accurately, and doesn't turn it over in dangerous areas. Has absolutely nothing to do with shot attempts.

That is just quibbling with terminology. They give plenty of context to individual players. +/- is a bad stat because goal scoring is infrequent and if you happen to get unlucky you can get buried. Shot attempts happen a lot so are far less prone to randomness as far as analyzing what is happening on the ice. And looking at WOWY you can see how good someone is with or away from a given player. That provides plenty of context. You can see if someone is being carried by teammates or if two players work well together etc.

 

[BobRouse]

Adding Oshie, Williams and Schmidt/Orlov as 3rd pairing has had significant impact on our 5v5 play so far.

 

[Xaroc]

Adding Oshie, Williams and Schmidt/Orlov as 3rd pairing has had significant impact on our 5v5 play so far.

And to this point you can see why Justin Williams is a possession monster. Just watching him play he is super smart, never gives up on a play, and has a great stick.

 

[Ridley Simon]

Your 100% Fancy Stats Discussion Thread

I agree with the posters that dont like the Roster Building threads getting bogged down with the fancy stats talk (these days....Orpik v the world).

Can we have its own thread for that, please?

Its not a crime to have multiple threads, and Fancy stats deserves it own, from where I sit.

So...get back to your Orpik v Orlov etc discussion....but have it here?

 

[Ovechkins Wodka]

Team Orpik

 

[g00n]

I agree with the posters that dont like the Roster Building threads getting bogged down with the fancy stats talk (these days....Orpik v the world).

Can we have its own thread for that, please?

Its not a crime to have multiple threads, and Fancy stats deserves it own, from where I sit.

So...get back to your Orpik v Orlov etc discussion....but have it here?

Merged with the existing thread that's been inactive for a bit. Feels like we're going over the same old ground, maybe some can catch up on what's already been discussed, or add to it.

We're going to have talk about stats when talking about players, and vice versa. We've had these kinds of debates that dominate roster or player conversations in the past. If we're talking about the value of the players, it's fine. If we're almost exclusively talking about the value of the stats, that should be in this thread.

If someone wants to try an Orpik vs Orlov thread we can see how it goes, though I think it'd be nothing but an exercise in piling on one or two posters.

 

[Hivemind]

Can we stop calling possession stats fancy stats? They're rather simple. We're only a couple years away from them being viewed right along side "traditional" stats.

Fancy/advanced stats would be the ones that actually involve calculations.

 

[CapitalsCupReality]

Don't all stats by definition involve calculations?

 

[Ovechkins Wodka]

Can we stop calling possession stats fancy stats? They're rather simple. We're only a couple years away from them being viewed right along side "traditional" stats.

Fancy/advanced stats would be the ones that actually involve calculations.

Until possession is on the score sheet it won't be viewed as traditional.

Nylander would have been the Babe Ruth of possession with his circles... Prolly is why his kids are such highly ranked prospects.

 

[NobodyBeatsTheWiz]

Can we stop calling possession stats fancy stats? They're rather simple. We're only a couple years away from them being viewed right along side "traditional" stats.

Fancy/advanced stats would be the ones that actually involve calculations.

We can stop calling them "fancy stats" when people stop extrapolating things that aren't there, like possession.

 

[Hivemind]

We can stop calling them "fancy stats" when people stop extrapolating things that aren't there, like possession.

We've been over this before. Every single study has shown the difference between shot attempts and zone time is negligible.

 

[NobodyBeatsTheWiz]

We've been over this before. Every single study has shown the difference between shot attempts and zone time is negligible.

And we've been over this before. The only study that's ever been linked on here was a 10 game sample size of one team. In fact, that "study" said this:

This does not conclusively prove that shot rate statistics are an accurate proxy for puck possession. I measured only 10 games out of the 1230 that are played every NHL season.

There are actual possession metrics used by teams. The NHL calls Corsi/Fenwick shot attempt statistics. No idea why that's such a difficult thing for everyone else to do.

 

[Raikkonen]

Those guys from sportlogiq will bring the light to stats community. But I doubt they will make it freely available

Btw, we should ask Vic Ferrari about it. He's working for GMBM after all.

 

[twabby]

It's largely irrelevant because shot attempt metrics predict success, not possession.

 

[Hivemind]

And we've been over this before. The only study that's ever been linked on here was a 10 game sample size of one team. In fact, that "study" said this:



There are actual possession metrics used by teams. The NHL calls Corsi/Fenwick shot attempt statistics. No idea why that's such a difficult thing for everyone else to do.

I've linked multiple different studies myself. Including one by the Caps' own Vic Ferrari, which showed strong Pearson Coefficients between zone time and all three shot attempt metrics.