Primary Sidebar
Bio
Gabriel Desjardins is the author of the Behind the Net blog and runs the statistical hockey website Behindthenet.ca. He grew up in Winnipeg and was a huge Jets fan and Dale Hawerchuk fan. These days, with no hometown team, he analyzes the entire NHL.
His writing has appeared in the Wall Street Journal, New York Times, San Francisco Chronicle, National Post, Globe and Mail and Toronto Star, as well as on ESPN.com, CNNSI.com, MLB.com, Puck Prospectus and Protrade. He has also worked as a statistical analyst for professional teams and agents in the four major sports.
In real life, he has a BSc from Queen's University and a MSc from UC Berkeley, and works as an Electrical Engineer in Palo Alto, CA. His last job before becoming an engineer was as a lumberjack. He recently stopped using wooden sticks and can't believe he waited this long to make the switch!
Contact him at info at behindthenet.ca.
Forget the Avs: let's worry about the Sharks!
by Hawerchuk on Mar 19, 2010 9:00 AM EDT
16 comments
For several seasons, the San Jose Sharks have combined strong 5-on-5 offense with defense that prevented their opponents from getting shots against them. In each of the last two seasons, the Sharks have had the 4th-best shot differential at 5-on-5 in the league. But this year, they can't play defense anymore:
| GF/60 | SF/60 | OSV% | GA/60 | SA/60 | SV% | Goal+/- | Shot+/- | PDO | |
| 2007-08 | 2.2 | 28.4 | 923 | 2.1 | 23.4 | 911 | 0.1 | 5.0 | 988 |
| 2008-09 | 2.2 | 30.9 | 928 | 2.0 | 25.9 | 924 | 0.3 | 5.0 | 996 |
| 2009-10 | 2.8 | 30.8 | 909 | 2.2 | 30.5 | 928 | 0.6 | 0.3 | 1019 |
This is surprising because San Jose has five skilled face-off men and has put up the 2nd-best faceoff percentage since the NHL started recording the data in 1997:
| Team | Season | FO% |
| CAR | 2001-02 | 56.1 |
| SJS | 2009-10 | 55.9 |
| PHI | 1998-99 | 55.4 |
| ANA | 2003-04 | 55.3 |
| WSH | 2000-01 | 55.2 |
| ANA | 2002-03 | 55.2 |
| DET | 2008-09 | 55.1 |
| DAL | 1997-98 | 54.9 |
| CAR | 2000-01 | 54.9 |
| CAR | 2003-04 | 54.9 |
It's amazing that San Jose can be so prolific in the face-off circle yet control shots so poorly. With just average face-off takers, San Jose would be a complete disaster. Actually they're already a disaster - here's their rank in Corsi% (=SF/(SF+SA)) by period in one-goal and tie game this season:
| Corsi% | Up 1 | Tied | Down 1 |
| 1st/2nd | 23 | 30 | 27 |
| 3rd<18 | 14 | 19 | 23 |
The only thing that's keeping the Sharks in the game is Evgeni Nabokov's high save percentage.
| EV | PK | |
| 2007-08 | 918 | 895 |
| 2008-09 | 920 | 854 |
| 2009-10 | 930 | 911 |
There's no doubt about it: Nabokov is playing way over his head. If we expect Craig Anderson to come down a bit over the next couple of months, we should logically expect Nabokov to crash hard. The irony is that Nabokov put up his best numbers in San Jose's weakest season in recent memory - had he played in good luck like this in either of the last two seasons, the Sharks could be very well be laughing at their many detractors while drinking from the Stanley Cup.
16 comments | 0 recs |
Corsi Percentage by Period
by Hawerchuk on Mar 18, 2010 9:00 AM EDT
14 comments
I got a few questions about Monday's piece on Corsi Percentage at even-strength with the scored tied. In particular, people wanted to see how Corsi Percentage varied by lead and by period. There is a significant effect:
| Up 1 | Tied | Down 1 | |
| 1st/2nd | 48.1 | 50 | 51.9 |
| 3rd<18 mins | 43.3 | 50 | 56.7 |
| Total | 46.1 | 50 | 53.9 |
Being up or down one goal has only the most minimal effect on shot rates during the first two periods. But in the third period, the team that's up goes into a defensive shell, while the team that's down pinches and takes chances. Here are team statistics and rankings in the first two periods:
| Up 1/1-2 | Tied/1-2 | Down 1/1-2 | Rank Up | Rank Down | |
| chi | 54.6 | 56.5 | 64.1 | 1 | 1 |
| pit | 50.9 | 54.4 | 51.8 | 4 | 15 |
| phi | 49.1 | 53.7 | 50.2 | 11 | 21 |
| tb | 50.7 | 53.1 | 51.5 | 5 | 18 |
| stl | 50.0 | 52.8 | 54.7 | 7 | 6 |
| nj | 49.6 | 52.5 | 49.1 | 10 | 26 |
| pho | 48.1 | 52.5 | 52.8 | 18 | 12 |
| buf | 49.9 | 51.3 | 51.7 | 9 | 17 |
| bos | 48.1 | 51.2 | 45.2 | 16 | 29 |
| ana | 47.2 | 51.1 | 53.0 | 21 | 10 |
| tor | 50.2 | 50.9 | 52.9 | 6 | 11 |
| ott | 48.1 | 50.8 | 55.5 | 17 | 5 |
| min | 48.2 | 50.7 | 49.4 | 15 | 23 |
| la | 53.8 | 50.6 | 56.2 | 2 | 3 |
| van | 46.4 | 50.6 | 54.1 | 25 | 9 |
| nsh | 47.2 | 50.4 | 51.2 | 22 | 19 |
| dal | 51.0 | 50.2 | 54.2 | 3 | 8 |
| car | 47.6 | 49.8 | 57.6 | 19 | 2 |
| det | 49.9 | 49.7 | 46.6 | 8 | 28 |
| cls | 44.9 | 49.6 | 49.2 | 26 | 25 |
| mon | 42.9 | 49.6 | 54.6 | 30 | 7 |
| atl | 48.7 | 49.4 | 51.7 | 12 | 16 |
| cgy | 48.4 | 49.2 | 50.2 | 14 | 20 |
| edm | 44.8 | 47.3 | 56.1 | 27 | 4 |
| was | 43.5 | 46.1 | 52.5 | 29 | 13 |
| nyr | 44.6 | 45.9 | 44.4 | 28 | 30 |
| nyi | 48.6 | 45.8 | 49.2 | 13 | 24 |
| fla | 47.3 | 45.7 | 50.1 | 20 | 22 |
| col | 47.1 | 44.8 | 52.5 | 24 | 14 |
| sj | 47.1 | 44.1 | 47.4 | 23 | 27 |
Chicago dominates, as usual. The 3rd period looks a little different:
| Up 1/3rd | Tied/3rd | Down 1/3rd | Rank Up | Rank Down | |
| chi | 47.4 | 61.3 | 68.8 | 6 | 1 |
| det | 50.0 | 58.4 | 64.0 | 3 | 4 |
| phi | 49.7 | 57.5 | 63.6 | 4 | 6 |
| nj | 49.2 | 56.3 | 52.9 | 5 | 24 |
| buf | 37.3 | 56.2 | 65.2 | 28 | 2 |
| tor | 52.5 | 55.7 | 64.8 | 1 | 3 |
| stl | 39.4 | 53.4 | 55.8 | 22 | 18 |
| pit | 40.8 | 51.6 | 52.6 | 20 | 25 |
| la | 44.5 | 51.6 | 56.3 | 13 | 15 |
| was | 42.6 | 51.3 | 54.2 | 18 | 20 |
| cgy | 45.6 | 51.3 | 62.9 | 11 | 7 |
| nyr | 40.5 | 51.1 | 56.6 | 21 | 14 |
| nsh | 43.1 | 50.9 | 63.9 | 17 | 5 |
| atl | 28.2 | 50.7 | 49.6 | 30 | 28 |
| ott | 45.6 | 50.6 | 61.3 | 10 | 9 |
| tb | 46.0 | 50.6 | 52.2 | 8 | 26 |
| cls | 43.5 | 50.2 | 55.9 | 16 | 17 |
| van | 46.2 | 50.0 | 59.1 | 7 | 11 |
| sj | 44.2 | 48.5 | 53.3 | 14 | 23 |
| car | 41.8 | 48.0 | 53.9 | 19 | 21 |
| bos | 50.3 | 47.7 | 62.0 | 2 | 8 |
| edm | 44.6 | 47.5 | 54.4 | 12 | 19 |
| ana | 38.4 | 47.3 | 61.1 | 26 | 10 |
| pho | 45.7 | 47.0 | 56.0 | 9 | 16 |
| dal | 37.2 | 46.9 | 56.9 | 29 | 13 |
| fla | 38.1 | 45.4 | 44.0 | 27 | 30 |
| min | 43.6 | 43.7 | 53.7 | 15 | 22 |
| mon | 38.6 | 42.3 | 58.9 | 24 | 12 |
| col | 39.1 | 40.8 | 49.0 | 23 | 29 |
| nyi | 38.6 | 36.7 | 51.7 | 25 | 27 |
The Colorado Avalanche were really the impetus for this post, and I don't think there's a method to their outshooting madness: they're 29th in Corsi % early and late in tie games. They back off when they're up no matter which period it is, and they're 29th in the league at outshooting their opponents in the 3rd period. They're essentially the anti-Leafs: good goaltending instead of bad, skilled shooting instead of unskilled, and a very low shot volume as opposed to a very high one.
14 comments | 0 recs |
Empty-Net Goals and One-Goal Games
by Hawerchuk on Mar 17, 2010 9:00 AM EDT
0 comments
In the last 90 seconds of a one- or two-goal hockey game, the trailing team gets much more value out of scoring a goal than they do out of preventing one. Hence the strategy of pulling the goalie. The net result is that lots of one-goal games become two-goal games, and lots of two-goal games become three-goal games. If we eliminate empty-net goals, we can see how many "true" one- and two-goal games there are:
| No ENG | ENG | |
| 1-goal | 32.5 | 24.3 |
| 2-goals | 20.6 | 22.5 |
| 3-goals | 11.9 | 17.9 |
This table includes games from 2001-2009.
There is one minor wrinkle in this analysis: teams are more likely to tie a game or go from two down to down one goal when they pull the goaltender. So if we ignore the disadvantage they have when they pull the goalie, we also need to ignore the advantage they get from doing it. We can remove the effects of pulling the goalie by looking at the score before the last two minutes of the third period:
| Prior to 18:00 | No ENG | ENG | |
| 1-goal | 33.3 | 32.5 | 24.3 |
| 2-goals | 21.7 | 20.6 | 22.5 |
| 3-goals | 12.2 | 11.9 | 17.9 |
It turns out to be a very minor difference. But fundamentally, there are a lot more one-goal games and a lot fewer three-goal games than we'd think based on the score at the end of regulation.
0 comments | 0 recs |
Colorado Avalanche: Worried Yet?
by Hawerchuk on Mar 16, 2010 9:00 AM EDT
70 comments
Following on yesterday's post about Corsi percentage: The Colorado Avalanche have been one of the year's surprise success stories. After finishing 28th in the league last season, they came out of the gate 10-1-2 and they're on pace to hit 100 points this year. Much of the improvement came from the brilliant offseason signing of goaltender Craig Anderson - despite being outshot by more than 4.5 shots per game, the Avalanche have managed to score just as many goals as their opponents in the 54 games since their hot start. I know nobody wants to hear this, but I'm concerned that Colorado's success is illusory.
Craig Anderson has posted a .923 save percentage, which has netted the Avs three more wins than if he put up his career-average save percentage. It's instructive to look at career save percentages for active goaltenders:
| Goalie | SV% |
| Backstrom | 0.919 |
| Hiller | 0.919 |
| Luongo | 0.919 |
| Thomas | 0.918 |
| Lundqvist | 0.917 |
| Vokoun | 0.917 |
| Anderson | 0.915 |
| Miller | 0.914 |
| Huet | 0.914 |
| Kiprusoff | 0.914 |
| Mason | 0.914 |
| Bryzgalov | 0.914 |
Anderson is a good goaltender and he was a solid pickup, but we shouldn't expect a .923 save percentage going forward from him or anyone else.
But goaltending isn't even half the story: not only do the Avalanche have one of the best save percentages since the lockout, they also have one of the highest shooting percentages. This table shows the raw numbers and their rank for even-strength play when the game is tied, both with and without missed shots:
| Excl Miss | Rank | Incl Miss | Rank | |
| SH% | 9.2 | 14 | 6.8 | 9 |
| SV% | 936 | 15 | 954 | 18 |
| PDO | 1027 | 7 | 1021 | 4 |
| Corsi% | 44.5 | 144 | 48.0 | 124 |
There have been 150 team-seasons since the lockout, so ranking 144th in Corsi% is troubling. Sometimes the very best teams can get outshot because they spend so much time with the lead and don't need to pinch and press for shots. But as the table shows, Colorado gets outshot when the game is on the line. Obviously a team can get outshot and still win with a great goaltender, but in the long-run, teams simply do not sustain high shooting percentages. If Colorado's shooters had merely been league average, the Avalanche would have 72 points today and a playoff spot would feel very much out-of-reach.
It may not seem inherently unreasonable for the Avs to have a higher-than-average shooting percentage. But the bottom of the roster - filled with players who've been healthy scratches or shuttled back and forth to the minors - has a much higher shooting percentage than the top three lines. There's no reason to think that players who aren't really part of the plan are going to shoot the lights out, and there's no reason to think that guys like Brandon Yip or David Jones (when he returns) will continue to post the best shooting percentages of the post-lockout era.
Let me be clear: it's not like Colorado should finish 28th in the league this season. But even with Anderson playing well in goal, they seem like more of a playoff bubble team. It will be very difficult for the Avs to win a playoff series if they don't start dominating their opponents on the shot tables.
70 comments | 2 recs |
Corsi Percentage: Up One, Down One and Tied
by Hawerchuk on Mar 15, 2010 9:00 AM EDT
10 comments
I just wanted to re-visit league-wide Corsi percentages at even-strength for teams in tied or one-goal games. Chicago continues their dominance on the shot board, while Toronto is improbably second:
10 comments | 0 recs |
Biggest Brawls of the Last 23 Years
by Hawerchuk on Mar 12, 2010 9:00 AM EST
16 comments
We've only got data back to 1987, so a lot of the really big brawls of the past are missing. Here are the five-biggest regular season brawls plus the biggest playoff brawl:
#1. Ottawa vs Philadelphia, March 5, 2004 - 21 fighting majors, 20 game misconducts, 417 PIMs, Most Penalties in a Single Game Ever:
Ottawa Senators vs. Philadelphia Flyers Brawl March 5, 2004 (via Yoeddy1)
16 comments | 0 recs |
ESPN: wrong and bad with small sample sizes, too
by Hawerchuk on Mar 11, 2010 7:23 PM EST
0 comments
Malcolm Gladwell has written extensively about the birthday phenomenon in sports: when youth leagues use a birthday cut-off (say January 1st), the majority of players tend to be older rather than younger. This phenomenon persists through junior hockey and seemingly into the NHL. As ESPN's Alvin Chang notes:
"People have tested this theory on several of sample groups, and it almost always holds true -- in junior leagues, the NHL and, of course, NHL draft classes."
It was nice of Alvin to link to my work on junior hockey, but unfortunately, he never read my follow-up, which noted that if we restrict our dataset just to high-scoring NHL players, the birthday bias almost completely disappears. While there might be a lot more first quarter birthdays in junior, players born late in the year are much more likely to make the jump to the NHL. It's not surprising either - if you're born in December, then you've been playing against bigger and stronger and better players for your entire life, and you have an extra year of development before you reach your peak relative to a guy who was born 11 months earlier than you. Sure, the minors are full of players born early in the year, and they are also over-represented among goons, but for the players we care about, it's not a significant effect.
Oddly,while Alvin doesn't acknowledge this effect, he draws a strange conclusion from a small number of players in the 2010 draft class:
"We tested [the 2010] class, fully expecting the usual results. But we found an anomaly. This draft class isn't dominated by North American skaters born between January and March. Instead, it's dominated by skaters born between April and June."
That's why you take more than one year worth of data, my friend! Sometimes weird things happen. Thankfully, the statisticians he consulted made no mistake:
"So, finally, we had to ask: Is this a fluke? Several statisticians told us an occasional anomaly is to be expected. Even Gladwell said, via e-mail, "One would expect wide variation from year to year. It's really long-term trends that we are interested in.""
Wow. Long-term proven trend experiences one-year anomaly.
Next time, it might be more interesting to ask general managers what they think of the late-early birthday disparity...
0 comments | 0 recs |
Campaign 2010: Fire Brian Sabean
by Hawerchuk on Mar 11, 2010 6:09 PM EST
9 comments
It's not often that I write about baseball anymore, but I have decided to campaign for one thing in 2010: the firing of Brian Sabean. If 2010 is going to be a bad year for incumbents, it should be a bad year for the most incompetent of them all.
A brief summary of the charges:
1) Once a successful general manager, Brian Sabean's recent record features four-straight losing seasons followed by a fluke 3rd place 88-win finish. If you think the 2009 Giants were that good, I have some Icelandic bonds to sell you.
2) While an ok evaluator of pitching talent, Sabean had zero home-grown hitters of consequence on his roster between 2004 and 2008. And it's not as though he traded them away - there are no Giants draftees hitting well for any other team. Sabean was so convinced that his first and second round draft picks were worthless that he intentionally gave them away for several years by signing free agents a day early; a day later and he could have had both the free agent and the pick.
3) Sabean doesn't seem to have a clue about player valuation. He has signed many mediocre players to bad 3- or 4-year contracts, and normally, his crowning achievement would be Barry Zito's 7-year, $126 million dollar contract, which BZ has already underperformed by 40% in his first three seasons. Of course, Sabean also consummated this deal:
11/14/2003: P Joe Nathan, P Francisco Liriano and P Boof Bonser to Minnesota for C A.J. Pierzynski
Fangraphs estimates that Pierzysnki was worth about $2.3M more than the Giants paid him. Of course, Sabean non-tendered A.J. at the end of his first season in San Francisco. In true Giants fashion, they needed after-the-fact justification for releasing Pierzynski and the San Francisco Chronicle dutifully reported that he had kneed trainer Stan Conte in the nuts.
Regardless, the three players Sabean sent away for a mediocre catcher have already been worth $80 million to the Twins.
The bottom line: Brian Sabean is so incompetent that he gave away nearly $200 million in value through bad free agent signings and bad trades since the 2003 off-season. For this, he deserves to be fired.
9 comments | 0 recs |
Secondary Sidebar
FanPosts
Community blog posts and discussion.
Certain photos copyright © 2010 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
