Thursday, January 29, 2009

Was Glenn Anderson Clutch?

As someone who has a reasonable degree of familiarity with the situational playoff game results for the 1980s Edmonton Oilers dynasty, from my investigation of Grant Fuhr, I felt compelled to respond to a recent post at Mc79hockey about Glenn Anderson's clutch play.

I disagree with Tyler Dellow's conclusion. There is some pretty good evidence that Glenn Anderson did score more in close games and less in blowouts.

First of all, I'm not sure regular season results matter at all if you are trying to prove somebody is relaxing in less important situations, since somebody who only cared about championships probably thought everything from October to March was less important. Anderson is one of those players that is highly regarded for his playoff performance, and I doubt there are any regular season stats you could pull out that would sway his supporters at all. To me, it makes sense to address this question by looking at playoff performance.

I happen to have the scoring results for Oiler playoff games from 1981-1987 for Gretzky, Kurri, Messier, and Anderson, broken down into results for the first two periods and the third, just like I did for the goaltending numbers. The source is the Hockey Summary Project. Unlike my goalie numbers I haven't gone through the tedious step of double-checking the results, so the totals don't quite match Hockey-Reference, but they are pretty close.

In the first two periods, here is the scoring breakdown:

Wayne Gretzky: 46 goals, 92 assists, 138 points
Jari Kurri: 40 goals, 55 assists, 95 points
Glenn Anderson: 32 goals, 41 assists, 73 points
Mark Messier: 32 goals, 38 assists, 70 points

Now here is third period and overtime:

Wayne Gretzky: 19 goals, 44 assists, 63 points
Mark Messier: 22 goals, 25 assists, 47 points
Glenn Anderson: 20 goals, 24 assists, 44 points
Jari Kurri: 20 goals, 18 assists, 38 points

In the first two periods, Gretzky/Kurri scored 86 goals compared to 64 for Messier/Anderson . In the third period that flipped, and the second line guys actually scored more goals, 42-41. I would assume that Gretzky was double-shifted a lot in the third, since there is much more of a gap between his results and Kurri's. If so, Gretzky may be partially responsible for Messier's and Anderson's results if he played together with them more late in games. However, there is no doubt that Messier and Anderson were much more likely to score in the third period.

That's pretty good evidence for the Anderson fans right there, but the main allegation is that Gretzky piled up the points in blowouts while Anderson did not. To check that, I considered any game where the score differential after two periods is 3 or more goals to be a blowout scenario.

Third periods starting with a score differential of 3+:
Wayne Gretzky: 4 goals, 15 assists, 19 points
Jari Kurri: 9 goals, 5 assists, 14 points
Mark Messier: 5 goals, 2 assists, 7 points
Glenn Anderson: 4 goals, 3 assists, 7 points

What about in close games? Did Anderson's scoring rise? Let's change it to take all third period results that started either tied or with a one goal margin to see how they did in close games:

Third periods starting with a goal differential of 1 or less:
Wayne Gretzky: 10 goals, 20 assists, 30 points
Mark Messier: 9 goals, 17 assists, 26 points
Glenn Anderson: 8 goals, 15 assists, 23 points
Jari Kurri: 8 goals, 7 assists, 15 points

Now, let's convert everything to a per-game rate and summarize Gretzky's scoring vs. Anderson's:

First two periods:
Gretzky: 0.72 GPG, 2.16 PPG
Anderson: 0.50 GPG, 1.14 PPG

Third periods and OT:
Gretzky: 0.59 GPG, 1.97 PPG
Anderson: 0.63 GPG, 1.38 PPG

Third periods in blowouts:
Gretzky: 0.52 GPG, 2.48 PPG
Anderson: 0.52 GPG, 0.91 PPG

Third periods in close games:
Gretzky: 0.61 GPG, 1.84 PPG
Anderson: 0.49 GPG, 1.41 PPG

Looks like both assessments are generally correct (Gretzky cherry-picked more in blowouts, Anderson scored more in tight games). Many people are probably at this point tempted to attribute the gap to effort level or clutch play, but I'm still not sold on that. I think the Oilers simply played to the score, especially Messier's line. They played offensively when tied or behind and defensively when they were well ahead. In my Fuhr post you can see the shot for/against numbers in various situations that support this point. Note that because the Oilers were so good, the blowout numbers consist almost entirely of games where the Oilers were the ones blowing out the opposition. Since all we are measuring here is scoring, somebody who stopped trying when the score is not close would be pretty much indistinguishable from somebody who simply played defensively when their team was way ahead. I think the latter reason is more likely to be the case with professional athletes in playoff situations, but I think media and fans are more likely to come up with the former explanation.

There are still questions that remain about Anderson's HOF status, such as for example whether it was Messier or Anderson who was really driving the results on that line, but that his scoring was situational seems to me to be an open and shut case.

Monday, January 26, 2009

Back to Backs

Doogie2K did a recent article on the physiology of goaltending, focusing on the effects on the body of playing goal and why it is difficult to recover in time for optimal performance in back-to-back games. I found the article to be very interesting, and wanted to do a bit of additional research to try to quantify the back-to-back effect.

I don't have a quick way of figuring that out, so I decided to use the brute force method of scrolling through game logs. I took a sample of the top 10 goalies in games played since the lockout, and compared how they did in the first game vs. the second game of all of their back-to-backs over the last 3 seasons. What I found, somewhat surprisingly, was that the numbers were pretty much identical in both halves of the back-to-backs:

First Game: 131-80-13 (.614), 2.40, .918, 29.2 SA/60
Second Game: 124-72-27 (.617), 2.43, .918, 29.8 SA/60

However, we need to account for two major lurking variables: travel and strength of opposition. If teams are aware of the potential factors influencing back-to-back performance, they are more likely to avoid starting their goalies in back-to-back games. Therefore it is likely that the goalie will face only weaker opponents in the back-to-back games he does start. We would also expect that goalies would recover better if they did not have extensive travel between back-to-back games.

Here is the breakdown by Conference:

East: 2.40, .919, .628 in 1st game; 2.36, .922, .598 in 2nd game
West: 2.42, .914, .585 in 1st game; 2.53, .912, .651 in 2nd game

I expected the numbers from the Western Conference goalies to be especially influenced by weaker opposition, because of the more extensive travel teams have to endure. That seems to be supported by the increase in winning percentage in the second game, despite a drop in goaltending statistics.

The Eastern numbers suggest that the goalies have similar performance in back-to-backs, but the rest of the team of the team plays a bit worse (scoring rate drops slightly and shots against go up by 0.7 shots per game, resulting in a worse record).

The weaker opposition effect is less likely to be present for goalies who play in nearly every game. I broke it down by only looking at back-to-back results from seasons where a goalie played 70 games or more.

First Game: .641 win %, 2.21, .922, 28.4 SA/60
Second Game: .609 win %, 2.41, .916, 28.8 SA/60

Compare that to goalies who didn't play 70 games or more that season:

First Game: .563 win %, 2.71, .910, 30.2 SA/60
Second Game: .634 win %, 2.45, .922, 31.4 SA/60

These numbers suggest that schedule strength is a factor, and that teams do avoid starting goalies in back-to-backs against quality opponents. As always with goalie stats, these numbers are dependent on the rest of the team - we don't know what the impact is on the skaters, so maybe the numbers for the workhorse goalies have a lot to do with the type of shots being allowed. However, I think it is likely that goalies can be expected to perform somewhat worse in back-to-back situations, and the evidence suggests the effect is probably somewhere around .005-.010 in save percentage.

The sample sizes are fairly small for each goalie individually. However, for what it is worth Turco, DiPietro and Vokoun did better in the second half of back-to-backs, while Luongo, Kiprusoff, Nabokov, Brodeur and Giguere did worse, and Miller and Lundqvist were about the same. A lot of that probably had to do with strength of opposition.

Thursday, January 22, 2009

Destroying a Brodeur Excuse

One oft-repeated excuse used by Brodeur fans (one of my commentors brought it up today, in fact) is that Brodeur was not the starter on Canada's national team in 1996, 1998 and 2002 because the other goalies had their coaches managing the squad.

Oh, really?

In the 1981 Canada Cup, Scotty Bowman started Mike Liut over his own goalie, Don Edwards.

In the 1984 Canada Cup, Glen Sather started Pete Peeters over his own goalie, Grant Fuhr.

In the 1987 Canada Cup, Mike Keenan started Grant Fuhr over his own goalie, Ron Hextall.

In the 1991 Canada Cup, Mike Keenan started Bill Ranford over his own goalie, Ed Belfour.

Looks to me like the Canadian coaches certainly seem like they are able to evaluate their own guys pretty well. Maybe they were vocal in securing their guy a roster slot at least, but when the games started they played the guy they thought had the best chance of helping their country win.

Now what happened when Brodeur entered the picture?

In the 1996 World Cup, Glen Sather started Curtis Joseph, his own goalie, ahead of Martin Brodeur.

In the 1998 Olympics, Marc Crawford started Patrick Roy, his own goalie, ahead of Martin Brodeur.

In the 2002 Olympics, Pat Quinn started Curtis Joseph, his own goalie, ahead of Martin Brodeur in game one vs. Sweden, before playing Brodeur in all the remaining games.

Do coaches always go with their own goaltender? Quite clearly they do not. In fact, there has never been a Canada Cup or best-on-best Olympic tournament where a Canadian head coach picked his own current goalie as the starter, except when Martin Brodeur was the backup. Either it is one big conspiracy against Brodeur, or the coaches simply didn't think he was good enough. From the stories I have read from around those time periods, those coaches were not making a rogue decision against popular support - the consensus from journalists and fan polls seems to be that Brodeur just wasn't the best option.

If you want to argue that Martin Brodeur deserved to be the starting goalie at the 1998 or 2002 Olympics, then argue it based on his merits. Don't use the lame and insulting excuse that it was because of some sort of favoritism from the head coach, because the historical record is pretty clear that Canadian coaches do not have a record of being biased towards their own goaltenders in major international hockey tournaments.

Sunday, January 18, 2009

Which Goalie Was Best in the Clutch?

After having dealt with the aggregate numbers, it's time to delve into the individual data to take a detailed look at the situational performance of Belfour, Brodeur, Hasek, Joseph and Roy. The last time I tried to rank them by their high-leverage performance, I put them in the following order: 1. Belfour, 2. Roy, 3. Hasek, 4. Joseph, 5. Brodeur. This time I'm armed with substantially more detailed data, having broken every playoff game they played between 1994 and 2008 (except for 1997) by period and game situation, so it's time to review and update those rankings.

I expected that since all these goalies were pretty good, there wouldn't be a huge difference in their play late in the game. Turns out I was wrong. Here is how the goalies did during the first two periods:

Patrick Roy: 2.16, .923
Ed Belfour: 2.19, .921
Dominik Hasek: 2.11, .920
Martin Brodeur: 1.97, .919
Curtis Joseph: 2.37, .916

Very similar performance all around. Now let's add in their results for the 3rd period and overtime, and I'll also include my "close and late" save percentage, which includes overtime and all third periods that began tied or with a one goal differential.

Dominik Hasek: 1.77, .935 in 3rd/OT, .939 close and late
Ed Belfour: 1.75, .932 in 3rd/OT, .936 close and late
Martin Brodeur: 1.92, .919 in 3rd/OT, .923 close and late
Patrick Roy: 2.22, .919 in 3rd/OT, .905 close and late
Curtis Joseph: 2.08, .918 in 3rd/OT, .912 close and late

Hasek and Belfour significantly outperformed their peers late in games, or at least they appear to have done so. We need to evaluate the team factors before we can make a conclusive statement.

The first situation to look at is when the goalie's team is leading by one goal after 2 periods. How a goalie performs when his team is leading late in the game is probably one of the main measures people use to determine how "clutch" someone is. If a goalie can hold the other team off the scoresheet in this scenario, his team wins, which is a pretty valuable contribution.

Up By One Goal After 2 Periods:

GoalieSFSASV%SH%WL
Hasek21.028.0.9707.9%162
Brodeur22.426.4.94210.9%325
Joseph19.130.0.9384.9%124
Belfour21.228.8.90711.6%226
Roy19.730.6.89211.6%209

In third periods his team entered leading by one goal, Dominik Hasek had an 0.83 GAA and a .970 save percentage. Did his team's style of play contribute to that? It probably did, but you can factor in an awfully strong team effect and those numbers are still disgustingly good. In Buffalo the shot splits indicate that the Sabres were pretty much hanging on for dear life whenever they got a lead - in all the third periods they started with the lead combined, the Sabres were outshot by nearly a two-to-one ratio and scored on only 5% of their shots, yet they went 21-1 because the opposition almost never scored on Hasek.

The numbers indicate that Curtis Joseph's teams were similar in terms of trading off offence to try to hold the lead. Cujo did pretty well with a .938 save percentage despite getting almost no goal support.

The numbers show that Brodeur, Belfour and Roy all benefitted from teams that were very good at counterattacking when in the lead. Brodeur had a very good save percentage, although the Devils had the best shots for/shots against ratio and probably were mostly outplaying the opposition even while ahead late. I would guess that, with the Devils' strength combined with the opposing team's likely heavy shot bias, Brodeur was probably facing a relatively low shot quality against here. The Devils were noteworthy for having a few third periods where they led but still completely shut down the opposition to the tune of only 1 or 2 shots against in the third period. However, all goalies probably faced somewhat easier than average shots when their teams were ahead by a goal in the third, and Brodeur likely did contribute to his team finishing out games.

Both Belfour and Roy had fairly mediocre save rates. Roy in particularly did quite poorly in this scenario, at least in the portion of his career included in the study, posting a sub-.900 save percentage and allowing the other team to come back to win 9 times.

Next up, how our 5 netminders did when the game was tied:

Tied After 2 Periods:

GoalieSFSASV%SH%WL
Belfour26.423.0.9437.3%1812
Joseph23.625.2.9245.9%916
Hasek25.225.7.9159.0%1522
Brodeur26.625.0.9137.6%1825
Roy25.327.4.90310.5%2012

The tied results are a little trickier to evaluate, because both the shot rates and save percentages depend a fair bit on which team ended up scoring first to break the tie. Roy's win/loss was very good in these situations, yet the numbers indicate that the Avalanche snipers were probably the ones driving the results.

This was the only situation that Hasek's results were not outstanding. His save percentage was not too bad relative to his peers, but he may have let in a few goals at the wrong time since his win/loss record was worse than expected.

Joseph again did pretty well but got very little goal support, and that is reflected in his record.

Belfour significantly outperformed everyone else here, but his numbers show that the Dallas Stars were an elite team in tied games late.

Belfour in DAL: 31-18 SF/SA, .952 Sv%, 2.17 GF/60
All other teams: 20-32 SF/SA, .934 Sv%, 1.55 GF/60

Since it is a similar scenario, let's look at overtime results as well:

Overtime

GoalieSFSASV%GFAGAAWL
Belfour26.831.8.9491.911.642118
Hasek25.528.7.9391.891.751413
Roy35.732.4.9383.102.001711
Brodeur29.327.3.9161.482.291117
Joseph32.327.2.8891.773.03712

Put these two scenarios together, and Ed Belfour was the guy with the most success in tie games. Belfour and Hasek were both strongly outshot on average in OT, yet played well enough to help their teams to a winning record.

Roy's overtime legend is well established, and these numbers do not disappoint. The shots for and against numbers seem to indicate that Colorado trusted their goalie enough to play a more open style of game in overtime, and their offence and Roy's goaltending combined for some pretty good results.

I've been critical of Martin Brodeur's overtime record before, but to be fair he has had abysmal goal support. His save percentage has not been outstanding in OT, but most of the blame should fall on the shooters. Once again Curtis Joseph did not get much goal support, but he also didn't make as many saves as he should have in sudden death play.

The final game situation was when a team is trailing. Which goalie was best able to hold the other team off and allow his team a chance to tie the game?

Down by One Goal After 2 Periods:

GoalieSFSASV%SH%WL
Hasek30.223.6.9558.2%89
Belfour25.622.2.9449.3%1118
Brodeur30.122.1.9187.9%522
Roy28.420.0.8909.2%411
Joseph32.619.8.8839.6%517

The answer, once again, is Dominik Hasek. Hasek faced the most shots of any of the goalies, and had a dominating save percentage (.955). Hasek's goal support was about average, but his team went 8-9 in games they entered the third period trailing by a goal. For comparison's sake, the average winning percentage of the other 4 goalies combined was just 26%. Belfour again joined Hasek well clear of the rest of the field.

Somewhat interestingly, the goalie that got the most support in this scenario was Curtis Joseph, the same guy who had the least goal support at pretty much all other times. Joseph's teams had a very strong outshooting rate when trailing, but Cujo's performance was not very good (.883).

If I had to rank the goalies based on their overall performance in high-leverage situations, the top choice is pretty obvious: Dominik Hasek. Hasek was great in OT, dominating when his team was trying to mount a comeback, and virtually unbeatable when they had the lead. Hasek's career was great, but his results in Buffalo were even better - as a Sabre, Hasek's "close and late" playoff save percentage in 1,167 high-leverage third period and OT minutes was an astonishing .949.

Ed Belfour takes the second spot comfortably, with Brodeur and Joseph pretty close for 3rd and 4th. Somewhat surprisingly, Patrick Roy ends up in 5th.

Roy and Joseph both suffer a bit because the 1997 playoff season is missing here. I have no doubt that Roy's playoff results in Montreal would look very strong, although they would need to be adjusted somewhat to the league scoring averages to make for a fair comparison with the Colorado numbers. I do suspect that we would see some strong team factors at play with Roy's numbers as well, however, since the Canadiens had a strong defence. The evidence here suggests that the Avalanche did not have a particularly strong team defence, but their high shooting percentages were a big help for Roy.

This whole exercise helps describe a bit more of the team context these guys were playing in. Most of all, however, it shows that Dominik Hasek was the best goalie of his generation, and that his advantage over his peers was even greater when the chips were down.

Thursday, January 15, 2009

Shots and Save Percentage Revisited

I wanted to do some more study of the relationship between shots and save percentage. Earlier looks at this topic have shown little to no relationship, but my recent look at third period play shows a large effect of teams playing to the score. This is likely to be at least partially blocking the underlying shots/save percentage relationship, if there is any, so I thought to use my sample of playoff shots by period to revisit this issue.

I have compiled shot numbers from all playoff games played by Belfour, Brodeur, Hasek, Joseph and Roy from 1994-2008 (not including 1997). I was especially interested in the results from the first two periods, which we would expect to be less affected by scoreboard effects. All the shot rates in this post are expressed in terms of shots against per 60 minutes of play, both because that is a familiar scale for shot numbers and also to allow us to compare to the third period results. For the first two periods, the breakdown was as follows:

0-15 shots faced/60: .906
15-22 shots faced/60: .911
23-30 shots faced/60: .918
31-37 shots faced/60: .927
38-45 shots faced/60: .930
45+ shots faced/60: .930

Those numbers certainly do suggest that more shots against result in a higher save percentage, although we have to be a bit careful with conclusions from this sample since the numbers reflect the performances of just 5 goalies playing on 10 different franchises. Also, even though the correlation seems pretty clear, there are some other variables that we need to take into account.

One such variable is special teams play. Power play shots are more likely to go in than even strength shots, and if power play shots are a higher percentage of the total shots taken in a more defensive contest we would expect this to lower the average save percentage. If this is true, we would expect that save percentage would be correlated with the total shots in the game from both teams, so I ran the numbers on that one as well (again, first two periods only) and they support the theory:

less than 45 total shots per 60 mins: .912
from 45 to 60 total shots per 60 mins: .919
more than 60 total shots per 60 mins: .923

Running this type of breakdown on a larger sample consisting of only even strength shots during the first two periods would remove both the special team and playing to the score variables, and would therefore be a good way to test if there is any direct relationship between number of shots and save percentage.

What about the third period numbers? Does the relationship continue when teams start playing to the score? Here are the third period numbers, again expressed in terms of shots per 60 minutes (e.g. 10 third period shots would be equivalent to 30 shots per 60).

0-10 shots/60: .873
11-15 shots/60: .931
16-20 shots/60: .890
21-25 shots/60: .918
26-30 shots/60: .935
31-35 shots/60: .931
36-40 shots/60: .924
41+ shots/60: .927

The relationship is not as clear, but there still seems to be some evidence that more shots means a higher save percentage. The sample size is smaller so a bit more randomness should be expected, but we also know there is a strong game score effect.

The logical step is to break the third period numbers down by game situation. The numbers are still expressed as shots per 60 minutes, but the sample sizes are smaller so I'm going to compress the groups.

Third Period Shots When Leading After 2 Periods:
0-20 shots/60: .927
21-30 shots/60: .934
31-40 shots/60: .923
41+ shots/60: .939

Third Period Shots When Tied After 2 Periods:
0-20 shots/60: .895
21-30 shots/60: .911
31-40 shots/60: .943
41+ shots/60: .923

Third Period Shots When Trailing After 2 Periods:
0-20 shots/60: .900
21-30 shots/60: .927
31-40 shots/60: .909
41+ shots/60: .882

It looks like a similar relationship when tied. However, there is not a clear relationship for either the shots while leading or the shots while trailing. The shots when trailing look, if anything, to be negatively correlated with save percentage, but I would expect that this could partly be explained by strength of opposition, as only dominant teams would be badly outplaying and outshooting the opposition while holding a one goal lead late in the game.

The problem with the tie-game numbers are that they are from third periods that start out tied, but don't usually stay that way. When one of the teams scores a goal to take the lead, both teams would then adjust their style of play according. Based on what we know about score effects the numbers may make sense, if the teams that faced few shots are teams that got scored on early and then had to attack for the last period, and the teams that faced a lot of shots scored early and then sat back to defend the lead. Note that special team factors are not very likely to be at play here, as referees do not generally make many penalty calls in the third periods of close playoff games.

The "playing to the score" effect may be similarly impacting the first and second period numbers as well. I assumed that teams would start playing to the score in the third period, but in a playoff game that they were trying their hardest to win they would quite possibly start playing to the score earlier. That seemed to be the case in many games that I saw, especially when one team went up by a few goals early on and then got outshot for the next two periods. It would be pretty unlikely for two teams that were tied 0-0 in a playoff game through 2 periods to have allowed 25 shots each to that point, as generally teams play cautiously until they are forced to open it up.

Some have argued that goalies are less focused and more likely to let in goals when shots are less frequent. This one is very difficult to test without play-by-play data. As a goalie myself, I can believe that this would have some small effect, but I really doubt it is anything significant enough to even worry about. One possible way to test this would be to look at the performance of backup goalies who come into the game off the bench and see how they do in their first couple of shots against. I'm not sure what kind of situational effects they would be facing, since that would usually be a blowout scenario, but that would be an interesting number to see.

A final consideration is whether chances against the run of play are more dangerous than shots generated in the offensive zone, since the defence is less organized and there are less defenders to beat. There is some evidence in favour of this one from the scoring and save percentages I collected for teams when leading and trailing in the third period. However, team strength is also a factor - if the stronger team is more likely to be leading in the third, and the stronger team is more likely to score on any given shot, then we would expect that teams leading in the third period would have better shooting percentages.

I'd like to see this theory tested by some more advanced metrics like measures of shot quality or average shot distance, with game score taken into account. Note that the third period leading and trailing numbers suggest that this effect, if real, would only apply if one team was substantially outplaying the other team. It seems that in the third period, the leading team usually plays it safe while the trailing team is just putting more pucks on the net, and in that situation we don't see any relationship between the number of shots against and save percentage. Again, however, we may expect that the team leading is more likely to be the better team, which would be counterbalancing any effect.

I think that special teams, playing to the score, and strength of opposition effects are mainly responsible for the stats that show a relationship between shots against and save percentage. More advanced studies could be done using play-by-play data to test each of these variables individually and see the effect on the results, which would allow us to come up with a more definitive answer to this question.

One thing that seems clear from this is that shots taken are at least partially discretionary. The way we see both shots against and save percentage increase for the goalie on the leading team in the third period is evidence of that. This finding has implications for the ability to compare goalies based on shots against totals. A goalie who faces more shots per game than his teammates might be every bit as good in terms of puckhandling and rebound control and so on, and the difference might stem entirely from the team being in the lead much more often with him in net. Or maybe opposing coaches told their shooters to fire from everywhere on that particular goalie (see this game for an example of how that can happen to anyone). There may be certain goalies that have inflated reputations which cause opposing shooters to pass the puck rather than shoot from all angles, or, on the flip side of the coin, there may be goalies who are falsely considered to be weak or have poor rebound control that end up facing a barrage of rubber by opposing teams who are eager to challenge them. In both of those cases, it seems to me a bit unfair to give the goalie the credit or blame for that extra shot differential.

Monday, January 12, 2009

The Third Period Updated

After getting some feedback from post on playing to the score, I went back and took out all the empty-netters from my sample of third period statistics. Here are the revised totals for Belfour, Brodeur, Hasek, Joseph and Roy combined in the third period in the playoffs, from 1994-2008 (1997 not included):

Leading by 1 after 2 periods:
21 SF, 29 SA, 1.96 GF/60, 2.02 GA/60, 9.4% SH%, .930 SV%

Tied after 2 periods:
25 SF, 25 SA, 2.06 GF/60, 2.02 GA/60, 8.1% SH%, .920 SV%

Trailing by 1 after 2 periods:
29 SF, 22 SA, 2.59 GF/60, 1.76 GA/60, 8.8% SH%, .918 SV%

For comparison's sake, the average shooting percentage in the first period was 9% and the average save percentage was .920.

This is evidence that the team leading in the third period may have slightly better scoring and save percentages than the team trailing. The shots will usually favour the trailing team, meaning the trailing team is more likely to score the next goal. Despite this the team trailing after 2 periods won only 14% of the time, which shows the effectiveness of the tactic.

What is nice about these numbers is that the huge majority of the shots were taken 5 on 5, since the refs usually put away the whistles in the third period of playoff games.

Thursday, January 8, 2009

And The Vezina Goes To...Whoever Never Had to Face Gordie Howe

I have done a bit of research into the Original Six lately, which has become a point of interest for me because of the close relationship between team strength and goalie stats in that very unbalanced league. I mentioned before that goalies on the best teams never had to play against their own Hall of Fame shooters. Here is a some evidence that proves this point. From 1946-47 to 1955-56, Toronto, Montreal and Detroit were the three best defensive teams almost every single season. That makes it easy to divide the league in two with a strong group and a weak group. I looked at the results of games when the strong teams played the weak teams, and here are the goals against averages for each team against the Bruins, Hawks and Rangers combined:

1. Detroit 2.22
2. Montreal 2.22
3. Toronto 2.25

That is stunningly similar for a 10 year sample size. Based on those results, I think it is reasonable to conclude that the overall goal prevention ability of each of those three teams (team defence and goaltending combined) was roughly equal. Here's how they did against each other:

1. Detroit 2.15
2. Montreal 2.30
3. Toronto 2.44

As we can see, the overall difference between the teams came entirely from their results against each other. Now we don't have save percentages for those seasons, and we can't necessarily rule out the possibility that, say, Detroit's defence was worse than Toronto's but they made up for it with superior goaltending. Still, it makes one wonder whether Terry Sawchuk's biggest advantage was really his goaltending skill, or merely his good fortune to avoid having Gordie Howe and Ted Lindsay spoil his shutout bids.

This is certainly something to keep in mind for older goalies, but a similar effect can still show up in more recent years as well whenever there is a single dominant team in the league, or a few teams that are a level above everyone else. A goalie on a great team often has great playoff stats because he never had to face his own team. It is also a potential flaw with using playoff results adjusted compared to league average. Grant Fuhr's league-adjusted playoff stats, for instance, probably look relatively better than they should since his Oilers were out there wrecking everyone else's. I bet Fuhr's numbers would be a lot closer to average if I took out all of the playoff games involving Gretzky's Oilers and then recalculated the year-by-year league averages. I'd like to see a similar calculation for guys like Billy Smith or Chris Osgood as well.

Wednesday, January 7, 2009

Why Canadians Hate Hasek

Watching the world junior hockey final, it became clear why the average Canadian hockey fan hates (and therefore underrates) Dominik Hasek. It is because the cardinal sin in Canadian hockey is diving.

Canadians hate diving so much that they often see it even when it doesn't exist. Go down lightly once or twice, and every single time thereafter will be seen as a dive. Whenever Swedish goalie Jacob Markstrom was run in the final the fans thought it was a dive, and they got on Markstrom's case, booing him when he touched the puck and even when he was called up to accept the best goalie of the tournament award. Apparently the general consensus is that when you come out of your net to play the puck, and get hit during your follow-through with most of your weight on your lead foot, and that same plant foot is taken out by one of the fastest skaters on the ice plowing into you at top speed, knocking your stick out of your hands and spinning you around so fast that your helmet flies off and nearly clears the zone -- it must have been a dive. Bloggers everywhere are trying to outdo each other in hyperbole - according to one, it was "the worst dive in the history of sport." Give me a break.

I'm not saying Markstrom was completely innocent by any means, but I don't see how he had any chance of winning either of those collisions with Esposito or Della Rovere. How much he embellished is I guess a matter of opinion. I say not really that much on all except one of them, but you can judge for yourself from the Youtube clip.

Dominik Hasek usually left no doubt about his unsporting manipulation of referees. He knew exactly what he was doing when he was drawing contact or feigning injury, not to mention things like complaining to the referees and knocking the net off its moorings when the opposing team was just about to score. Those type of antics just do not mesh with the Canadian hockey worldview, and as a result Hasek pays the price in the court of public opinion. Take a look at this comment thread if you want to see what the average Canadian hockey fan thinks about Dominik Hasek. I recommend you don't, as it is basically a bunch of things like: "Hasek is the best at whining like a little girl" and "He is a diver and embellisher which could make him one of the all time greats if he played soccer" and "Domenik the quitter, you have got to be kidding me. He wouldn't even make it onto my top 15 all time list."

I think that sentiment is generally correct (i.e. Hasek had a greater-than-usual predisposition towards diving, embellishing, whining and possibly even quitting), but at the end of the day you have to evaluate the whole package, and for Hasek that includes his dominating puck-stopping abilities. It would have been better if Hasek did not do those things, but he grew up in an Eastern European country where they don't have a similar stigma against diving and he was also a bit of a jerk, that's just who he was. It is easy to exaggerate the negative tendencies, but diving (while regrettable and unsportsmanlike) really has little negative cost in terms of winning or losing games. It is really more of a style issue, and at the end of the day, performance is far more important than style for ranking hockey players.

Monday, January 5, 2009

Playing to the Score

The outshooting and scoring numbers for the 1980s Oilers were so interesting I was inspired to gather similar data for other top goalies of the modern era. Using the same method as I did for Grant Fuhr, and restricting my sample to the playoffs only (because that is when teams really play to win), I collected shot and goal data for the first two periods combined, and then for the third period and overtime (if any). I then compared the results based on game score to see how teams played to the score. The data are not complete, because the Hockey Summary Project only has playoff data from 1994 to 2008 (with 1997 missing), so the numbers will not exactly match the goalies' career totals.

The goalies in my sample were Patrick Roy, Dominik Hasek, Ed Belfour, Curtis Joseph and Martin Brodeur, the same goalies I had included in a previous study of playoff performance, so I was hoping to revisit that analysis with some updated information as well as to test the effectiveness of some of my estimates. I will get to the individual numbers in another post shortly, but first I'd like to present the aggregates because they have value in addressing some fundamental hockey questions about how teams play to the score and the effect of strategy on goaltender play.

1st two period stats:

28 SF, 27 SA, 2.52 GF/60, 2.16 GA/60, 9.0% SH%, .920 SV%

All of the 5 goaltenders had a lot of wins in their careers, and this line shows why. Some of the goalies played on outshooting teams while others were outshot, and some of the teams were better at scoring than others, but overall the percentages worked out to an average edge of a 1% higher scoring rate, which accounts for 0.27 goals per game. The rest of goal differential comes from the slight advantage in outshooting.

3rd period stats:

These third period stats are for the entire third period that began with a specific score. Ideally we would like to count based on actual game score throughout, rather than just at the start of the period, but that requires play-by-play data. My "box score method" is much simpler, but achieves it through a loss of accuracy. However, there were relatively few lead changes in the third period, especially since most of the sample was during a low period for goalscoring league-wide. The strength of the observed effect, even in this limited sample, is strong enough that I feel confident that these results do approximate the real effect. I would certainly encourage someone with the programming chops to break it down in exact detail for some recent playoff seasons to see if my results are typical.

When leading by one goal after 2 periods:

21 SF, 29 SA, 2.36 GF/60, 2.03 GA/60, 11.8% SH%, .928 SV%

When up a goal late in the game, the teams drastically cut back on offence. Similarly, the trailing team started putting more shots on net. Interestingly, however, save percentage went up and so did shooting percentage, implying that the trailing team is taking more shots of lower quality, whereas the leading team is generating higher-than normal shot quality. Another possible explanation for this effect is that the leading goalies simply played better (by focusing more, bearing down more, etc.) in this important situation. However, I would imagine that the goalies who were down 1 would also be highly-focused, and yet their save percentages went down almost 3%. Whatever the explanation, this resulted in a huge percentage gap (almost 5%) in scoring rates, which made the leading team actually slightly more likely to outscore its opposition despite the one-sided shot differential.

I doubt that is generally the case (remember, we are only looking at strong goalies here, and any save percentage below .919 would result in the leading team getting outscored), however it certainly helps to have a great goalie when you are trying to hang on to a one goal lead late in the game.

When tied after 2 periods:

26 SF, 25 SA, 2.34 GF/60, 2.08 GA/60, 9.2% SH%, .916 SV%

When tied, the outshooting results are similar, with some evidence of teams playing more cautiously. Many of these periods started out tied but then had one of the teams take the lead, so could be quite different than what the picture would look like if we looked at just the time when the game was actually tied. We will, however, see that situation that when we look at the overtime results.

When trailing by one goal after 2 periods:

30 SF, 22 SA, 2.64 GF/60, 2.01 GA/60, 9.0% SH%, .916 SV%

When trailing, the teams started throwing pucks on net and taking greater risks offensively. The scoring rate was identical to the rate in the first two periods, although save percentage was slightly lower. We would need to look at a larger sample to see if the shot quality for the trailing teams is generally lower. I would suspect that it might be, but I am not sure. The evidence does appear to be there that the leading team will probably have higher than average shot quality for. The goalies' save percentage was .004 lower than during the first two periods combined, and, and earlier we saw how the shooting percentage went up almost 3% when the teams were up by one.

This shows again why these teams were winning a lot of games - their scoring rates were better than the opposition's in all three situations. Part of this is probably that the teams sampled were more likely than usual to tie the game up and therefore some of this time would have actually been played with the score tied. However, the team leading after two periods had a very good record (86% of the time the team that was leading after 2 periods ended up winning the hockey game).

Overtime stats:

28 SF, 28 SA, 2.03 GF/60, 2.14 GA/60, 6.7% SH%, .926 SV%, 14 W, 14 L

In overtime, save percentage rose and shooting percentage went way down. It was interesting that these goalies' teams were likely to get outscored in overtime. Even though all the teams had a good goalie in net they were just as likely to lose as to win once the game went into the fourth period. This suggests that overtime is quite random. Note: Patrick Roy was a major outlier in terms of overtime results through the course of his career, but most of his overtime outperformance came in Montreal pre-1994 and is not included here.

Outshooting results:

Given how shooting results are apparently strongly dictated by the scoreboard, shooting results for an entire game are not particularly useful. If a team dominates the first two periods and goes up a goal or two entering the third, they are likely to sit back while the other team takes a lot of long-range shots in the third. The shots might have been something like 25-15 through two periods, but then get reversed 13-3 in the third, and both teams end up with 28 on the night even though Team A was clearly the better team. I observed many games with that type of distribution when going through the box scores.

These effects were less strong in the first two periods, however, so we can look at how outshooting determined the results over those two frames. I calculated a "winning percentage" based on the score after 2 periods, giving 2 points for being ahead and 1 point for a tie game.

When outshooting the opposition, the teams had a .635 win percentage. When outshot by any amount they were just .481. Given that we are only looking at the first two-thirds of a hockey game here, that is a lot of evidence in favour of outshooting driving results. Even with a Hall of Famer in your net (Joseph is debatable, but let's just assume he is for the sake of a more punchy sentence), evidence suggests that your team is more likely to be trailing than leading if you get outshot over the first two periods of a playoff game. TV analysts take note: The best way to predict a playoff series is not to simply pick the team with the better goalie, but to pick the team more likely to have the edge in puck possession and shots. Certainly teams can have the percentages go their way and ride a combination of a hot goalie and goals against the run of play for a few games or even a series or two, but past history suggests that the the inevitable hand of regression nearly always shows up to end the fairy tale short of the final prize.