2007-11-21

Ryder Criticism. Alan Ryder, that is.

I read all of Alan Ryder's columns at the Globe & Mail. I'm a fan of his and it's great that he has a mainstream forum for his number crunching. Ryder's latest piece is on one of his hobby horses - respecting the shootout.

His conclusion is that there is little correlation between a goalie's hockey performance and shootout performance, therefore coaches and GMs need to spend more time thinking about the shootout. Especially those crotchety old-schoolers North of the border. To be honest, I find his dismissive tone of 'traditional hockey values' getting in the way of smart decisions a little insulting.

According to Ryder's table, there have been 337 shootout points awarded. Over the same time span, there have been 3373 non-shootout points awarded. That 3373 includes wins and OT/SO losses. So it's fair to say that the importance of the shootout is one ninth of the importance of regulation/OT NHL hockey.

For the most part, NHL coaches aren't morons. Say a coach has to decide between two players available to dress for a game. The two have identical regulation/OT performance but one happens to be a shootout ace. The decision is obvious in this case. The difficulty arises when there is a trade-off.

Let's compare Player A and Player B. Player A is fantastic in shootouts - he contributes 0.5 points more than Player B for every shootout. Unfortunately, he's not quite as good at regulation/OT hockey and contributes team 0.1 points per OT/regulation game less than Player B. Player A's net contribution:

0.5ppg*(1/10) - 0.1ppg*(9/10) = -0.04ppg

So Player A contributes 0.04 fewer points per than player B.

Similarly, if player C costs you 0.5 points per shootout but gains you an extra 0.1 points per regulation/OT, he's worth 0.04 points per game more than player B. Because of the rarity of the shootout, regulation/OT performance is much more valuable than shootout performance.

To sum up: Although 1/10th of all available points is a significant fraction, 9/10ths of all available points is nine times more significant. If you're facing a trade-off, you need to gain 9x more shootout value than you lose in regulation/OT value.

Next, let's look at Ryder's table of team shootout performance and compare it to regulation/OT performance:


Click the table for a clearer image. The SO record columns are copied from Ryder's table. I got the Non-SO record numbers from NHL.com.

The correlation between SO% and Non-SO% is -0.15. Good shootout teams are slightly less likely to be good 65-minute teams and vice versa.

The negative correlation is small but significant and completely unsurprising. It suggests the trade-off between shootout performance and 65-minute performance is real. Kiprusoff sucks in shootouts, but wins a lot of games for the Flames. Jussi Jokinen isn't that special in the first 65 minutes (recent 4-goal games notwithstanding), but is crackerjack in the shootout. Ryder's perspective that there is a lack of 'market competition' is explained by some teams having regular players that happen to be great shootout performers (i.e. Dallas), and some teams having great regular players who happen to be lousy shootout performers (i.e. Calgary).

To be fair, the main thrust of Ryder's article is that the backup goalie should come in as the closer for shootouts if he is the better shootout performer. It's a good point. The reason it doesn't happen is probably because coaches prefer not to put a 'cold' goalie in that situation, either due to injury concerns (seems goalies do the splits a lot in shootouts) or concerns about the performance of a 'cold' player. I'd like to see evidence that a player can be put in this situation regularly and perform at his usual 'warmed up' level without a greater risk of injury before I'd endorse the idea. Raycroft looks like the perfect guinea pig for the experiment.

Update: Ryder responded to my comment on his article thusly:

"You are absolutely correct that the shootout is 'only' about 10% of the game. But 'only' 3-4 players per team affect that 10%. And, of course, the goaltender is always one of those players. In the case of forwards the leading shootout contributions of all time have come largely from guys who don't perform at the same level in skating time. They are, in effect, shootout specialists. The shootout tests finesse. But the Leafs trot out their power forwards. The story with goaltending is more complex. You would not choose a single goaltender based on a stand-out shootout record. But, just maybe, you might choose a backup based on that metric AND use him as your closer."

Labels: , ,

3 Comments:

Blogger Dirk Hoag said...

If that correlation is -0.15, I think it says more that there is no inherent tradeoff between regular play and shootout performance, and the two are basically unrelated. If it was more strongly negative, that would indicate a tradeoff.

I think what's surprising is how casually some coaches approach the shootout, particularly in what shooters they select. Last week in Nashville, Ken Hitchcock threw out a motley collection that got stuffed in all 3 attempts, including Adam Foote as his final shooter. One point that seems worth noting is that in general, matching your shooters' hands to the catching glove of the goalie is one way to increase your chances.

11/21/2007 1:40 p.m.  
Blogger Jeff J said...

One would expect the correlation to be weakly negative for lots of reasons:
- a shootout involves as little as 21% of the participants in the 65 minute game.
- star shootout shooters with suspect regulation time skills can be hidden on the lower lines with minimal disruption to your 65-minute game, effectively further reducing that 21%.
- coaches get to select shooters and the shooting order. Any decent sample size will have it's anomalies. Even if your team is modeled after the Broad Street Bullies, some guys will probably be above average at breakaways.

Picking the cutoff point for significance of a correlation coeff. is arbitrary anyway.

I think the slight negative correlation is significant and is due to cases like Kipper whose weak shootout performance is tolerated because of outstanding reg/OT play. Real tradeoffs are out there - Raycroft/Toskala is a prime example.

Agreed on the handedness thing. I recall predicting that the first goalie to stop Jussi would be a right-handed catcher. I think I was wrong, but if I'd been right I'd still be gloating.

11/21/2007 3:15 p.m.  
Blogger James Mirtle said...

This post only took ya eight months!

Good to see you're back... (sort of).

11/27/2007 10:49 p.m.  

Post a Comment

<< Home