Saturday, November 15, 2008

ES Save % While Tied

There has been a very insightful investigation of the effects of possession on scoring success going on at Mc79hockey (check it out here and here). The key takeaway is that teams play to the score, and in-game tactics appear to have a large effect on both outshooting and scoring percentages.

To remove game score as a factor, Vic Ferrari wrote an app at his website timeonice.com to look at even-strength play with the score tied. I ran the numbers for each team in the league during the 2007-08 season. The full table is below, but here are a couple of comments:

Correlation between number of shots faced and save percentage: .097. Correlation between shot differential and save percentage: .063. Based on this sample, outshooting appears to have little effect on save percentage at even strength in a tie game.

Correlation between blocked shots and save percentage: -0.239. I'd like to see more data on this, but it seems to suggest there could be a weak negative relationship between blocked shots and save percentages (i.e. it is tougher for a goalie to play on a team that blocks more shots).

These results are more sensitive to luck/randomness since the sample size is smaller. There appears to be a clear shot quality effect, however, since most of the elite defensive teams show up near the top with weaker teams near the bottom. There is also a gap of .041 between the best and worst teams.

I compared the teams' results with Alan Ryder's shot quality numbers. Most teams had similar rankings, but there were a few teams with large discrepancies. Either those teams had very good or very bad goaltending, or they were very good or bad at protecting leads or playing from behind, or there was something else going on.

Better than expected:
Anaheim: 29th in shot quality, 1st in ES SV% while tied
N.Y. Rangers: 25th in shot quality, 6th in ES SV% while tied
Dallas: 27th in shot quality, 17th in ES SV% while tied
Montreal: 17th in shot quality, 8th in ES SV% while tied

Worse than expected:
Columbus: 1st in shot quality, 20th in ES SV% while tied
Washington: 4th in shot quality, 22nd in ES SV% while tied
Buffalo: 13th in shot quality, 30th in ES SV% while tied
Calgary: 3rd in shot quality, 18th in ES SV% while tied
Tampa Bay: 14th in shot quality, 29th in ES SV% while tied

Anaheim and Buffalo are the most interesting ones here. I don't think that either Giguere is that good or Miller is that bad. It could be that both teams have extreme results in the opposite direction when they are leading or trailing. However, I suspect that for both of those teams shot quality simply may not accurately measure the true difficulty of scoring chances against.

Assuming the RTSS shot distance figures that make up the shot quality rating are more or less correct (and these are road numbers only, to reduce rink reporting bias), the difference would therefore have to be either because of the average shot angle, the average defensive pressure on the shooter, or the quality of the average opposing shooter. Behind the Net has a shot quality measure that takes into account shot angle. According to those numbers, Giguere faced shots of about average difficulty, while Miller faced slightly easier than average shots against. It looks like the Ducks probably kept more of their shots away from the middle of the ice than normal, but that doesn't explain the entire difference.

I would suspect that as an offensive team Buffalo might be giving up more rush chances and more space to opposing forwards, which allows them more time to evaluate their options and make their shots. Anaheim's defence-first philosophy likely creates more pressured shots, which are easier to stop. Another possible explanation is that Anaheim has an elite checking line which may help their goalies by reducing shots against from the opposition's best players.

I would love to see these numbers for when teams were leading and trailing as well, to see both the overall save percentage effects and which teams were great at locking it down or opening it up.

Here are the full numbers for 2007-08 at even-strength, game tied:

RankTeamSAGASv%
1.Anaheim64836.944
2.Boston73843.942
3.New Jersey67541.939
4.N.Y. Islanders74546.938
5.Vancouver70846.935
6.N.Y. Rangers59940.933
7.Nashville69447.932
8.Montreal68947.932
9.Minnesota72450.931
10.Phoenix72350.931
11.Atlanta88162.930
12.Detroit51738.926
13.Pittsburgh70253.925
14.Florida78860.924
15.Ottawa60446.924
16.San Jose58845.923
17.Dallas63349.923
18.Calgary66953.921
19.Carolina62950.921
20.Columbus66153.920
21.Edmonton82367.919
22.Washington67556.917
23.St. Louis70459.916
24.Colorado69258.916
25.Toronto69158.916
26.Philadelphia68962.910
27.Chicago62156.910
28.Los Angeles70865.908
29.Tampa Bay71369.903
30.Buffalo63662.903

5 comments:

Anonymous said...

just the tip of the iceberg i would assume. i'd like to see the follow ups for save % when ahead, and when behind. my guess is particularly with brodeur (who has an excellent save % when the game is tied) also excels when his team has the lead. contrarily i can see his numbers being a lot more average when his team is trailing, for obvious reasons.

overpass said...

The key takeaway is that teams play to the score, and in-game tactics appear to have a large effect on both outshooting and scoring percentages.

I read those discussions, and I don't see where in-game tactics were found to have an effect on scoring percentages.

There has been no causal relationship found between outshooting or playing to the score and SV% in those threads, as far as I can tell. The only relationship found was between outshooting and playing to the score. This was making it appear as if there was a relationship between outshooting and SV%, but that was found to be an effect rather than a cause.

Further comparison of SV% while tied to SV% while leading or trailing might reveal a relationship between playing to the score and SV%, and I'd be very interested in seeing that comparison. Without any evidence of such a relationship, however, I think you are just reducing your sample size for analysis. I'm not sure you are gaining anything by analyzing ES SV% while tied as opposed to all ES SV%.

The Contrarian Goaltender said...

You're right, I shouldn't have included scoring percentages in that statement. I am expecting some differential between leading/trailing in terms of SV%, but we still need to analyze the data to be sure.

sunnymehta.com said...

Huh? The way I understand the recent findings is this:

When a team outshoots their opponent at even strength, their own S% and SV% are negatively impacted. However, when a team outshoots their opponent at even strength in a tie game, their S% and SV% are not negatively impacted.

overpass said...

However, when a team outshoots their opponent at even strength in a tie game, their S% and SV% are not negatively impacted.

Yes, I understand it this way too, based on Vic Ferrari's analysis of tied games.

When a team outshoots their opponent at even strength, their own S% and SV% are negatively impacted.

I don't think this is necessarily the case. It may be so, but the data presented at mc79hockey.com and by Vic aren't sufficient to prove this.

Suppose teams play to the score. Specifically, suppose that a team that is trailing will take more shots than the team that is leading. If this is the case, some of the outshooting teams will be outshooting teams only because they have previously allowed goals in the game. When looking at the whole game, the outshooting team will have the lower SV% and shooting% because of this, even if both teams had the same shot quality after the lead was taken. IOW, poor shooting% and SV% is a cause of outshooting, and not an effect. It's a selection bias issue.

It may also be true that teams that outshoot their opponent while leading and trailing tend to take low-quality shots, as you say.

The relationship between outshooting and shooting/SV% may be because of the selection bias issue. It may be a genuine relationship only when teams are leading or trailing. It may also be some combination of the two. However, without further analysis (such as Vic running the numbers for leading teams and trailing teams) it's impossible to tell which it is. As a result, I don't think we can't really draw any conclusions about shot difficulty for goaltenders on teams that outshoot or get outshot yet.