Wednesday, April 28, 2010

Winning By Getting Outshot

In 2009-10, including the playoffs, the Montreal Canadiens are 9-1-0 when allowing 46 or more shots against. The same team is just 3-14-2 when allowing between 26 and 30 shots against.

That doesn't seem to make a lot of sense, especially given that the team's goal support numbers are not that far apart (2.86 goals per 60 minutes for 46+ SA games, 2.53 goals per 60 minutes for 26-30 SA games). The difference looks to be because the Habs (and their opponents) tend to adjust their play to the score. When Montreal beats better teams, especially in games where they take the lead early, they usually allow a lot of shots against. When Montreal loses, especially in games where they fall behind early, they tend to allow an average level of shots against.

Here are the period-by-period shooting percentage, save percentage and shots for/shots against numbers for Montreal in both 26-30 and 46+ shot games this season:

26-30 Shots Against:
First period: 6.0%, .881, 215 SF, 185 SA
Second period: 8.5%, .889, 216 SF, 201 SA
Third period: 4.8%, .924, 210 SF, 179 SA

46+ Shots Against:
First period: 18.6%, .957, 70 SF, 163 SA
Second period: 8.2%, .975, 85 SF, 158 SA
Third period: 7.4%, .947, 81 SF, 150 SA

I know En Attendant Les Nordiques has scoring chances for Montreal this season. I really wish my French was better because what I do understand of his analysis is always very interesting. I'd like to see how much the Habs' scoring chances are affected by the game score, and just how much the chance-to-shot ratio varies depending on whether they are leading or trailing. The percentages above make it look like the leading team, either Montreal or their opponent, managed with some success to reduce scoring chances for both teams in the third period.

His numbers from game 6, though, suggest that Montreal was not able to reduce shot quality very much in the late going. In the last two periods Washington outshot Montreal 36-12 and outchanced them 25-8. Jaroslav Halak had to make a ton of tough saves to keep his team in front. It should be noted that the Caps are a great team that would probably outplay Montreal anyway, which it makes it tough to tell how much of that chance differential was the impact of their tactics and how much of it was because of a difference in ability.

If Montreal gets up early tonight we can expect an avalanche of chances from the Caps, and they're going to need Halak to hold the fort again. If Washington goes out in front, they'll probably play a bit more defensively and be a bit more selective with their shots. As a result, if Halak makes 45+ saves tonight it will most likely be in a Habs win or overtime loss. If Washington's final shot total ends up in the 26-30 range, that's probably not going to be good news for Montreal.

Friday, April 23, 2010

Synergy

Scott Stevens in the playoffs without Martin Brodeur:
42 wins, 51 losses

Martin Brodeur in the playoffs without Scott Stevens:
16 wins, 26 losses

Scott Stevens and Martin Brodeur in the playoffs together:
83 wins, 56 losses

Apart, neither has won a game beyond the second round. Together they have 3 Stanley Cup rings. Just in case anybody needed a reminder that teams win in the playoffs, not individuals.

And no, I'm not criticizing Brodeur. He wasn't the reason New Jersey lost this year, just like Scott Stevens wasn't the reason that Washington never won anything in the 1980s. I feel like I could write the same post I did last summer, where I argued that Brodeur has been the same guy in the playoffs whether the Devils were winning or losing, the difference was the play of the team in front of him. When your team scores eight goals in five games on Brian Boucher, then you're very probably going to lose.

Thursday, April 15, 2010

Rookies vs. Veterans

Day one of the playoffs is in the books, and the big stories were upsets and playoff rookies doing well. Playoff newcomer Brian Elliott outplayed defending Champ Marc-Andre Fleury while Craig Anderson also won in his first taste of NHL playoff action.

Experience usually gets a lot of play in the media and around the barstools at this time of year. From a goalie perspective, there is a noticeable lack of experience among most of the playoff teams this year. Seven goalies will be making their playoff debuts, while another has just one start to his name and yet another has just one start in the last 10 years. Here are the career playoff games played prior to last night of the 16 expected playoff starters:

1. Martin Brodeur, 176
2. Evgeni Nabokov, 65
3. Jose Theodore, 49
3. Marc-Andre Fleury, 49
5. Ryan Miller, 34
6. Roberto Luongo, 22
7. Brian Boucher, 22
8. Ilya Bryzgalov, 16
9. Jaroslav Halak, 3
10. Pekka Rinne, 0
10. Jimmy Howard, 0
10. Jon Quick, 0
10. Craig Anderson, 0
10. Antti Niemi, 0
10. Tuukka Rask, 0
10. Brian Elliott, 0

Does that mean that New Jersey, San Jose or Pittsburgh are the most likely teams to win the Cup? Not in my book. There has not been much of a relationship between goalie experience and playoff performance, suggesting that experience is largely overrated as a factor in postseason success.

Since the lockout, here are the numbers for goalies in their first playoff season compared to the numbers for goalies who had prior playoff experience (goalies with less than 120 minutes of prior experience were still counted as rookies):

Veterans: 244-230, 2.46, .914
Rookies: 92-106, 2.58, .913

The save percentages for the two groups were essentially identical. It is even closer than the rounding makes it appear. The rookies were at .9134 while the veterans were at .9136, a difference of .0002 over a total sample size for both groups combined of over 20,000 shots. Assuming that differential represents the true skill difference from having been there before, experience would seem to account for one additional goal saved every 24 playoff series. That's one goal every 6 playoff seasons if the goalie's team went to the Finals in each and every one of them.

The rookies faced an extra 1.4 shots against per game, which accounts for the difference in GAA and likely much of the difference in win/loss record. It seems unlikely that the goaltenders in either group would have had an effect on the shots against, since the sample is made up of a diverse group of athletes and the goalies who made their playoff debuts between 2006 and 2008 appear in both samples. The most probable explanation is that the best teams in this period tended to have playoff veterans. This is supported by the fact that although several goalies had great performances in their first playoff seasons, the only one to make it to the Finals was Cam Ward in 2006.

Last night the rookies combined for a .902 save percentage while the veterans combined for .910. The main reason for the veterans' slight outperformance was Brian Boucher, a guy who played a total of two minutes in the playoffs from 2003 to 2009. Three of the five most experienced playoff goalies were in action last night, and those veterans combined for a dreadful .871 save percentage.

When a rookie goalie gets shelled, people usually make up stories about how he lacked experience and couldn't handle the pressure. When a veteran goalie who has won before plays poorly, it usually gets attributed to other factors (fatigue, the team in front of them, old age, etc.). The excessive focus by many on playoff experience appears to mostly be caused by confirmation bias. People only remember the data points that fit the sample. If you look at the entire sample in recent years, there is essentially nothing at all to suggest that a playoff newcomer will perform worse than a grizzled veteran.

Calling a goalie a playoff rookie in the first place is often a very NHL-centric view, especially with an ever-increasing crop of Europeans coming over to play goal in North America. Most of the netminders with zeroes beside their names above have extensive playoff experience in Europe, the AHL or international competition. The only two starters with essentially no playoff experience as a professional are Elliott and Quick, and they both played in the NCAA Championship.

If you want to predict playoff goalie performance, I don't think you'll lose much accuracy at all by focusing on a goalie's talent and regular season track record and ignoring the variable of experience.

Monday, April 12, 2010

Theodore

Is playing net for the 2009-10 Washington Capitals somewhat like playing for the mid-'80s Edmonton Oilers? Relative to league average scoring, this year's Caps are surprisingly competitive with Gretzky & co. Washington's average of 3.83 goals per game is 38% better than league average. The only edition of the dynasty Oilers to top that mark was 1983-84, when the team scored a record 446 goals and was 41% better than league average.

Jose Theodore has attracted some attention lately because of he has not lost a game in regulation over his last 24 games played (20-0-3). In one of those games Theodore was pulled after 3 early goals, and only escaped because his teammates game up with 3 late markers to pin the loss on replacement Semyon Varlamov. Even if we count that game, 1 regulation loss in 24 games remains a most impressive team record. I'm interested, as always, in how much of this record is because of the goalie and how much is because of the team.

If you look at the list of games, there are certainly some where the team bailed out their goalie. Theodore gave up the first goal about half the time in his starts, so he wasn't holding off the opposition early. The Caps had games where they trailed 5-3 in the second period, 4-1 in the second period, 5-2 in the third period, 3-0 in the third period, and 3-1 in the second period, and in every case they came back to either win or at least force overtime. Obviously the prolific offence had a big part to play in those big comebacks, which are still very unlikely in general in today's NHL.

On the other hand, Theodore's distribution of save percentage by period is pretty interesting over this streak.

First Period: 238/261, .912
Second Period: 218/243, .897
Third Period: 207/214, .967

For the most part, the Caps did not completely dominate their opponents for 60 minutes. Of all the third periods that Theodore played, the Caps only started with the lead in seven of them. They were trailing six times, and tied nine times. Without getting that high rate of saves from their goalie late in the games, the Caps probably wouldn't have won as many. On the other hand, it's at least possible that there were some playing to the score effects in there. The shots against drop as the game goes on, it could be that the team was focused on their personal counting stats in the early part of the game and around the 40 minute mark started to turn their attentions towards bringing home the "W".

I'm not sure whether this year's Caps are getting a bit lucky based on the percentages or whether they are collectively doing something to make them shoot/stop the puck at a higher rate than expected. Generally in those types of questions you can almost always assume that it's luck, but when you have such an extreme outlier as Washington is it makes you at least wonder.

I think Theodore has probably been running hot in a few key game situations. He's definitely contributed to the streak, but there have been more than a few times when the snipers in front of him did all the heavy lifting. I also think he got a bit lucky in terms of the goal distribution, winning quite a few games by 4-3 or 5-4 scorelines. Some in the media seem to not respect Washington because of their goaltending situation. I wouldn't write them off entirely because of that, but we should expect Theodore's playoff performance to be much closer to his career average of .908 than the .922 he has during his latest unbeaten streak.

Just as a bit of trivia, in his 2001-02 Hart Trophy season Theodore had a stretch where he had a .933 save percentage and 2.18 GAA yet was just 11-11-6.

P.S. Sorry about the silence around these parts lately, hope to be back on a more regular posting schedule as we head into the playoffs.