Advanced Stats vs Don Cherry

[Jtabo]

I don't know why a comparison of shot attempts is considered a posession stat. By shooting the puck aren't you giving away posession essentially? Wouldn't a team that takes the time to set up and find a better scoring chance be more of a "possession" team than one that steps over the blueline and attempts the first shot they can get no matter its percentages? I see such flaws with Corsi more than any other advanced stats. And I hate when people jump in and defend it and say "of course its gunna have it flaws". Well no, it shouldn't. You know why it shouldn't? Because if your gunna treat hockey like MoneyBall you better be damn certain its flawless or you will be mocked.

 

[rojac]

Of course, you could be great possession team and not have a great Corsi if you don't take a lot of shots.

 

[Duke Silver]

I don't know why a comparison of shot attempts is considered a posession stat. By shooting the puck aren't you giving away posession essentially? Wouldn't a team that takes the time to set up and find a better scoring chance be more of a "possession" team than one that steps over the blueline and attempts the first shot they can get no matter its percentages? I see such flaws with Corsi more than any other advanced stats. And I hate when people jump in and defend it and say "of course its gunna have it flaws". Well no, it shouldn't. You know why it shouldn't? Because if your gunna treat hockey like MoneyBall you better be damn certain its flawless or you will be mocked.

The simplistic way people look at it is "you can't shoot the puck if you don't have it".

 

[thewave]

If you could only include shots taken within 30 ft of the net. It becomes more accurate I would suspect.

 

[leafs in five]

HAHA Mirtle does have a hard on for Corsi.

Good Corsi = taking a ton of bush league shots from the perimeter that basically pads the tendys stats (see Reimer/Bernier).

Quality scoring chances, goals and wins are all that matters. With most emphasis on the latter two.

you don't think that Reimer and Bernier have played remarkably well in putting up the numbers they have? because they've just been facing easy bush league shots from the perimeter all year?

 

[The Missing Piece]

50% of the time it works everytime.

Great OP.

 

[Duke Silver]

you don't think that Reimer and Bernier have played remarkably well in putting up the numbers they have? because they've just been facing easy bush league shots from the perimeter all year?

Why can't it be a little bit of both?

They have made the hard saves when they needed to: a testament to them.

But one of the major reasons their save percentage totals are so high is because the Leafs have done well at keeping the quality of shots low.

There have been so many times this season where an opposing team has had only 3-5 Grade-A scoring chances in a game against the Leafs. For the most part, our goaltenders have been up to the task.

There have also been so many times where the Leafs get up a few goals and the opposing team just starts throwing shots on net from anywhere. For example, the Leafs were outshot by the Oilers 24-21 going into the third period on Tuesday. They came into the third period up 3-0, then got outshot 19-5 in that period. Same thing happened in Calgary: Up 2-0 going into the second, outshot 28-15 in last 40 minutes.

 

[Downtown27]

I still don't see how shot differential directly corresponds to possession. Wouldn't it correspond to, say, shot differential?

Why not just count actual possession time? It's been done in soccer for decades and is a useful but not determinate stat to gauge team performance.

 

[Duke Silver]

I still don't see how shot differential directly corresponds to possession. Wouldn't it correspond to, say, shot differential?

Why not just count actual possession time? It's been done in soccer for decades and is a useful but not determinate stat to gauge team performance.

It's more difficult with hockey because of how often possession changes.

If it were possible, though, I'd love to have a "time on attack" measurement like they have in video games. It would count how much time a team has possession of the puck in the offensive zone.

 

[The Legend]

HAHA Mirtle does have a hard on for Corsi.

Good Corsi = taking a ton of bush league shots from the perimeter that basically pads the tendys stats (see Reimer/Bernier).

Quality scoring chances, goals and wins are all that matters. With most emphasis on the latter two.

I've been working on some shot quality as statistical analysis models for a while now in my spare time, mainly to attempt to prove one way or the other if this theory is true. It's not ready yet, but I'll just give you a bit of an indication of where it might be heading:

Last year, the Leafs took 111 shots within 10 feet of the net. They gave up 114 shots within 10 feet of the net. Within 20 feet of the net, they took 358 shots. They gave up 393 shots.

On the perimeter, which I would safely consider to be 30 plus feet from the net, the Leafs took 746 shots. They gave up 968 shots.

In the interest in my future piece on the matter, I won't provide any analysis or thoughts. Just consider those stats.

 

[wulfio*]

It's more difficult with hockey because of how often possession changes.

If it were possible, though, I'd love to have a "time on attack" measurement like they have in video games. It would count how much time a team has possession of the puck in the offensive zone.

I'm pretty sure it wouldn't be difficult if they stuck rfid chips or the like in the puck and tags of jerseys.

 

[LogieTrice]

you don't think that Reimer and Bernier have played remarkably well in putting up the numbers they have? because they've just been facing easy bush league shots from the perimeter all year?

They have played fantastic this year - aside from Ottawa game (Reimer) and Edmonton game (Bernier). They truly shut the door on quality scoring chances.

Carlyle was quoted saying the quality scoring chances he counted vs Calgary was something much closer like 12 vs 16. Granted the Flames still out 'quality chanced' us but its not as bad as the 2:1 shot differential. I would also argue teams get a lot more shots because the leafs play with the lead very often. So Randy shuts down the offense and tells them to play hard D and limit the good chances.

All I am saying is corsi is useless when it comes to teams that play defensively (aka Randy Carlyle hockey) when they have a lead because it takes in to consideration all the garb shots getting thrown at goalies.

 

[Damisoph]

They have played fantastic this year - aside from Ottawa game (Reimer) and Edmonton game (Bernier). They truly shut the door on quality scoring chances

I don't know much about corsi but my eyes have told me they've been giving up far too many quality scoring chances in the last couple games. They've been really good eliminating those second chances, clearing rebounds, etc...but they were lax against Edmonton and especially Calgary. Canucks will burn us tomorrow if we don't smarten up a bit.

 

[LogieTrice]

I don't know much about corsi but my eyes have told me they've been giving up far too many quality scoring chances in the last couple games. They've been really good eliminating those second chances, clearing rebounds, etc...but they were lax against Edmonton and especially Calgary. Canucks will burn us tomorrow if we don't smarten up a bit.

One thing I have noticed when it comes to quality scoring chances - Because of the viewing angle when watching on tv, the shot looks more dangerous than it really is (excluding the chances within 10x10 ft in front of the net - those are juicy and if they save it, that is a bonus).

This is most true when is a big clapper coming from the point. You can't tell how much traffic Bernier or Reimer has till the replay. The leafs do a great job of clearing the way so our goalies get a great look at all shots coming from the point. As a result, that makes the saves easier for the Leafs goalies.

 

[Mimico]

I still don't see how shot differential directly corresponds to possession. Wouldn't it correspond to, say, shot differential?

Why not just count actual possession time? It's been done in soccer for decades and is a useful but not determinate stat to gauge team performance.

Thats why I used CorsiFor% ((CorsiFor/CorsiAgainst) X 100) in my spread sheet.

 

[Espher]

Yes but people interpret that to mean that Team A is better than Team B and extrapolate from there when all they should really be worrying about is the score.

Correlation does not imply causation, and one game wouldn't be enough to demonstrate a strong positive or negative correlation anyway.

Advanced stats are extremely useful and -- over a large sample size -- they can show a strong correlation, but they do not guarantee anything, and can't be taken in a vacuum. It's the same for any sport.

A team consistently defying a correlative analysis that is an accurate predictor for 25-29 other teams makes them an outlier and, while it's great that they're an outlier and may tell us that we need to look at and consider more elements (as anyone who cares about stats already knows), that doesn't mean the analysis is absolutely guaranteed 100% wrong.

The pro-stats guys who parrot them blindly are just as bad as the anti-stats guys who dismiss them blindly.

With all that being said... keep calm and Corsi's wrong.

Edit: One thing I would like to see plotted out is CorsiFor% but divided into three groups: CorsiFor%Tied, CorsiFor%Leading, CorsiFor%Trailing. The second, in particular, could be an accurate reflection of how the teams play with a lead (sit back and turtle, keep playing as they were, or press the attack). Even better if we could divide it by the spread (e.g. leading/trailing by 1, leading/trailing by 2, leading/trailing by 3+).

 

[getyourselfsomerest]

Corsi can't be wrong, it's a statistic. What's wrong(so far) is the people who have the stat sheet stapled to their foreheads saying the Leafs are bound to lose.

The way I look at it, Corsi is just a numerical value for what coaches already know. More often than not, Carlyle has said that he isn't happy about many aspects of the Leafs' game. And he's right. They've been consistently out shot and out worked. What's the difference? Great goaltending and skilled forwards. Most advanced stats people recognize this and say "duh, good goaltending is they key to everything." Even Don Cherry, who doesn't use advanced stats, knows that there are times when your goalie won't be as good. And if they keep playing this way, they'll more than likely lose those games. Advanced stats guys aren't saying it's a fluke that the Leafs are winning, they're saying it's unsustainable BECAUSE it's rare that goaltending is this good for an extended period of the season. I know this, you know this....if the Leafs goalies falter, they're not currently playing well enough to outscore the other problems. That's why Carlyle says time and time again that they need to tighten up defensively.

Corsi is great on an individual level in a large sample when looking at a couple other stats. On a team level, I don't think it's as good because I think there is such a thing as quality shots and Jason Blake shots and I think our goalies have stopped a TON of Jason Blake shots.

 

[Kyle Doobas*]

Yeah, getting outshot 2-1 every night isn't concerning at all... This'll never backfire....

That's not even remotely what I said..?

 

[dougieg93]

I'm sure there are teams that are more built for corsi and still win hockey games, the leafs are still built for the rush with a little more defensive responsibility than with Wilson as head coach. They are winning games through PP + PK + SV% + SH%

 

[thewave]

Just take a standard 30ft circle, center and move it so it lines with the back of the net. That's the primary striking zone pretty much. Eliminates the super sharp low percentile shots. If you mock it up and overlay it on the NHL illustration you can get a great idea. I wish they had a prime zone feature so you can see just that and eliminate the perimeter junk.

 

[Mystifo]

Yes but people interpret that to mean that Team A is better than Team B and extrapolate from there when all they should really be worrying about is the score.

Well it is intuitive to think that if Team A has the puck for more time than Team B then they have a better chance of winning.

 

[likeabosski]

Corsi stats also come with on-ice sh% and on-ice sv% stats. Don't knock on Corsi just because that one stat alone does not tell you everything. It's useful when used in combination with other stats. And when you look at context, who they are playing with, etc. Scott Gomez is a good Corsi player but I would not put him on the ice with other playmakers. He would need to play with a goal scorer to be productive.

In the NHL, goal scorers typically get more dough than playmakers but for good reason though. You can teach a goal scorer to be less selfish with the puck but you can't easily teach a "pure" playmaker to score goals. Some playmakers have high shooting %s but that's only because they tend to pick their shots and feed the lower percentage shots to a guy who is more capable. Being a quality passer is not something that you can just become easily though. So playmakers are still important to have in ice hockey.

 

[Havoc]

Must be the Leafs skewing your correlation LoL

 

[James Mirtle]

The correlation between points percentage and Fenwick Close the last six seasons is actually .58. Score effects matter if you're doing possession analysis. So does sample size. If you take out last season, which was a bit unique due to only having 48 games, it rises to .61.

That's higher than a lot of other statistics thought to be fairly important at the team level. It correlates higher with winning than power play percentage, penalty kill percentage, shooting percentage, save percentage, shot differential, faceoff percentage etc. It's roughly the same as goals per game.

No one has ever argued teams cannot win games without possession. Possession is only 5-on-5 play and doesn't make any measurement of goaltending, which is obviously huge for any team's success. You can be a great possession team with horrific goaltending and you'll miss the playoffs. Or, like the Leafs, you can be a weak possession team with great goaltending and special teams and you'll make them.

But there's no argument that poor or middling possession teams do as well as good ones. Teams with 52 per cent or better Fenwick Close are, on average, dominant teams. Teams below 48 per cent are, again on average, quite poor.

For the Leafs to be a consistent 100-point team in this league, it's likely they're going to have to spend more time in the opposition end than they have been. The good news is it looks like they've got goaltending and special teams figured out.

 

[mix1home]

If you look at Corsi as nothing more than shot attempt differential it's a useful statistic in terms of quantifying possession. Nothing more. It doesn't have to be predictive to be of value.

Wanna win games with poor possession? Easy, just maintain higher shooting and save percentages than your opponent. Sadly, we know that over time any teams PDO (save + shooting %) will normalize towards 1000.

We may very well maintain our freakishly high PDO but history tells us that more often than not things will regress to the mean.

This regression to 1000 means that you cannot have all good forwards AND good goalie on your team at the same time, or you cannot be particular at taking your shots AND forcing opposition to take bad shots, which both assumptions being quite silly.