Sunday, November 20, 2011

Anomalies: Why Are Crosby's Possession Numbers Merely Good

This will perhaps be the start of a continuing series in which we analyze and discuss strange things we've noticed about hockey statistics. A feature of these articles is that we won't have all the answers, so comments and discussion are strongly encouraged.

In this installment, I want to discuss something that has irked me since I first got into hockey analytics. After reading about Corsi and Fenwick stats I went to BTN to look at the numbers for my favorite team, the Pittsburgh Penguins. I was quite surprised by how bad the team was according to these metrics, this was in the Therrien days, and in particular Crosby's numbers were downright mediocre. They have improved since but over the last two seasons, in which he played a season and a half, his even-strength Corsi/Fenwick stats are merely good instead of great. To state the obvious, he is considered by many to be the best player in the game and nobody reasonable would put him outside the top 5-10 players, head injury aside. What gives?

5-on-5 Corsi

Here are his Corsi numbers from 2007-2008 through 2010-2011. I include his rank among forwards that played 40 games or more for the given year.

SeasonCorsi On/60Rank
2010-20118.5564
2009-201012.5529
2008-2009-0.21186
2007-20082.35140

Even with the improvement, including the huge points streak taking up much of the 2010-2011 season, you can see that his stats aren't close to the elite level most everyone would put him at, including us nerds. I have a couple possible explanations, but would love to hear from you if you've got more.

Weak Linemates

For Pens fans, a frustrating part of having so much strength at center is that there is not a lot of money to go around for wingers. If you look at the guys consistently at the top of the Corsi rankings, you tend to see pairs or groups of top guys that play together or perhaps an elite player with at least one good player. Examples include a number of Detroit combinations, whichever combination of Kane, Toews, Sharp and Hossa you like, the Sedins, Ovechkin with Bäckström and Kesler with Raymond. Crosby has spent most all of his time with guys like Dupuis and Kunitz who aren't bad but are definitely role players and don't compare well to those names.

Let's take a look at the Corsi QoT for selected players in the last 3 full seasons. Corsi QoT is the average Corsi rating off all skater teammates for the player's ice time. It isn't perfect because a player influences his teammates' ratings but gives a rough idea how good the teammates are.

Player2008-20092009-20102010-2011
Sidney Crosby-3.6285.2975.779
Alexander Ovechkin12.9538.4384.013
Pavel Datsyuk17.24612.2297.363
Jonathan Toews11.62914.8588.437
Henrik Sedin1.3687.4358.28
Ryan Kesler-0.1497.0217.624

You can see that going by this metric Crosby's teammates have not been as strong as those of other elite players.

An exception to this rule is that he has played a bit of time with Malkin. How has that duo been? I haven't pulled data from 2007-2008, but I do have it from the last three years. Here is a chart with their Corsi rates together. Unsurprisingly, they are all much better than Crosby's overall numbers.

YearCorsi%Corsi/60Minutes
2008-200947.5%-5.647393
2009-201056.2%16.377202
2010-201162.0%32.047163
Total53.4%8.317758

I'll have more on that 2008-2009 season in a bit, but in the last two years for the time they've been on the ice together they have dominated. I didn't list the score-tied numbers but they are actually pretty similar.

Coaching

People are making similar arguments for Ovechkin this year, which I think is at least in part an overreaction to some brutal shooting luck, but it seems like in the early years the cautious, defensive approach espoused by Michel Therrien may have held the Penguins back. It's beyond my area of expertise to analyze the particulars of this, but there are data to back this up.

Courtesy of time on ice here are the Corsi percentages (Corsi shots for divided by Corsi shots for and against) the seasons before during and after Bylsma took over for Therrien. While the Pens were and still are a young team, the jump is pretty big and the 2008-2009 season is damning.

SeasonCoachCorsi%
2007-2008Therrien45.6%
2008-2009Therrien46.0%
2008-2009Bylsma52.8%
2009-2010Bylsma52.2%

A problem with this analysis for the 2008-2009 season is that Gonchar, Pittsburgh's best defenseman that year, was out most of the early part of the season coming back right around the time Bylsma took over 25 games from the end. Here are the numbers from 2008-2009 for all time in which Gonchar was off the ice, whether or not he was in the lineup, under Bylsma and Therrien.

CoachCorsi%
Therrien46.0%
Bylsma52.8%

For that 2008-2009 season, here are Crosby's on-ice Corsi numbers with each coach:

CoachCorsi%Corsi/60
Therrien48.8%-2.836
Bylsma50.5%1.205

Of course there are many many other factors like overall team skill going up and variance may have played a role. That said, based on these splits I think some of the blame for the first couple years goes to Therrien, or you might say the praise for the last few goes to Bylsma.

Putting It Together

Let's look at Crosby's Corsi numbers with and without Malkin since Bylsma took over:

SituationCorsi%Corsi/60
With Malkin57.7%20.239
Without Malkin53.0%7.12

That's in 421 minutes of ice time with Malkin, or over 7 full games worth of time. If Crosby played 75% of his time with Malkin then his Corsi% would be about 56.5% and his rate about 17. Both of these numbers would be right up at the top of the Corsi rankings.

Conclusion

These are my ideas, I'd love to hear more if you've got them. I'd also be interested in your thoughts on what I'm putting out there. I think the biggest factor by far is the relative weakness of the wingers. This is supported by how dominant Crosby was when Evgeni Malkin played the top-winger role for him over the last couple seasons.

7 comments:

  1. Interesting analysis, and something I've always wondered about too. I do think the coaching change is largely responsible for Crosby's improvements territorially. Malkin's numbers there have been much better too, which is weird because he hasn't seemed nearly as dominant under Blysma post 08-09 as he was under Therrien. Not sure what to make of that.

    As you alluded to linemates seem to have a pretty major impact on Corsi numbers. Having two highly skilled players capable of driving the play leads to the outrageous Corsi's we see from guys like Datsyuk / Zetterberg, Kane / Toews, Backstrom / Oveckin etc.

    One thing worth noting though is that Sid tends to play with Malkin in favorable offensive circumstances which may skew the numbers a bit. Talking specifically even strength here, of course. Often times after the other team ices the puck, or when Bylsma sees their 4th line on the ice, he will attempt to turn the tables on the opposition by rolling out the two headed beast. So we should naturally expect their numbers to improve just because of the easier ice time they tend to be getting together. The question is to what degree does this impact their numbers, and that's a tough question to answer without more data.

    ReplyDelete
  2. Sid and Geno also play together a lot more when the team is trailing as well, so score effects are another factor.

    ReplyDelete
  3. Hey a5. Definitely thought you would enjoy this one. Thanks for the comments.

    Malkin is very interesting with the same sorts of questions. His stats have improved quite a lot; last season he was among the league's best in his half season or so and he's been strong this year but in his time without Crosby I think he had a negative Corsi every season before that. There it's easy to point to defensive liabilities, being bad at faceoffs and so on. I think a lot of it is the same stuff I mentioned above.

    I meant to put this in there but Malkin-Crosby-X has a Corsi of 22.806 with the score tied since Bylsma took over. While I do agree that they get put out there for a lot of really nice situations, it seems like score effects aren't a big factor.

    ReplyDelete
  4. Ehh, not sure about the linemates. Kunitz is a top line wing on every team save Philly and Anaheim and Boston, I'd think.

    All the dominant possession players have great help from the back end, as well. Toews/Kane have Keith, Backstrom/Ovechkin Green, Datsyuk/Zetterberg Lidstrom, etc. Not sure the Pens have had that until last season, with Letang emerging as a true #1 all-minutes type player.

    Also, maybe that 11%+ on-ice shooting percentage isn't a fluke. (http://www.behindthenet.ca/nhl_statistics.php?ds=1&f1=2007+2008+2009+2010+2011&f2=5v5&f3=SIDNEYCROSBY&f4=C+LW+RW+D&c=1+3+7+8+13+15+17+20+23+24+25+10+29+30+32+33+34+63+67)

    ReplyDelete
  5. I think it is pretty easy

    1.) When your line is shooting 12.74% then you lose possession of the puck earlier then you would have if you didn't score. The same thing goes if a line is shooting 3%, a lot of those rebounds stay in the zone and you continue firing at the net. I think high sh% will lower your CORSI slightly.

    2.) 47% zone start in 2010-11 will hurt his possession numbers.

    3.) Pittsburgh was ahead so often last year that they were in that defensive posture more then most teams. They aren't a team like Detroit who plays aggressive with the lead. This hurt his overall numbers.

    ReplyDelete
  6. Not to mention Kunitz and Dupuis are certainly not 1st line player or even 2nd line players in my opinion.

    ReplyDelete
  7. Dupuis is not a 1st or 2nd line player, but Kunitz is a legitimate play driver - check out his numbers without Crosby, they're legitimate.

    ReplyDelete