Feb 182012
 

The other day at Pension Plan Puppets there was a discussion about the merit so Steve Downie and whether the Leafs should go after him if Tampa made him available.   In it I brought up the fact that when Steven Stamkos or Martin St. Louis or Vincent Lecavalier play with Downie their offensive numbers increase, sometimes dramatically.  The following table shows each players goals for per 20 minutes of ice time in 5v5 zone-start adjusted situations with and without Steve Downie on the ice with them.

Teammate With Without % Impr.
2011-12 Stamkos 1.481 1.19 24.5%
St. Louis 1.441 1.182 21.9%
Lecavalier 1.408 1.139 23.6%
2010-11 Stamkos 1.297 1.122 15.6%
St. Louis 1.246 1.06 17.5%
2009-10 Stamkos 1.284 0.903 42.2%
St. Louis 1.257 0.868 44.8%
Lecavalier 1.167 0.849 37.5%
2008-09 Stamkos 1.113 0.685 62.5%

This year Downie has improved their offensive production between 20 and 25% and it isn’t because Downie is an elite offensive player relative to those three guys.  He only has 12 goals and 28 points in 54 games this year and is Tampa’s 6th leading scorer behind the three listed above and Purcell and Malone.  In his best season, 2009-10, he only had 22 goals and 46 points so he is far from an elite offensive player at the individual level and yet he does something that makes his skilled linemates better.

I have always been interesting in exploring the optimal way to build a team and so this prompted me to look a little deeper to see if mixing in some toughness in the form of a winger with good size (not a ‘goon’) but not necessarily elite offensive skill with more pure skill players makes the skill players better.   So, here are a few more examples that I found interesting.

Alex Burrows

Burrows Teammate With Without % Impr.
2011-12 D. Sedin 1.052 1.017 3.4%
H. Sedin 1.112 0.92 20.9%
2010-11 D. Sedin 1.365 1.06 28.8%
H. Sedin 1.277 0.982 30.0%
2009-10 D. Sedin 1.481 2.126 -30.3%
H. Sedin 1.321 1.994 -33.8%
2008-09 D. Sedin 1.534 0.851 80.3%
H. Sedin 1.434 0.94 52.6%

Milan Lucic

Lucic Teammate With Without % Impr.
2011-12 Krejci 1.182 0.607 94.7%
2010-11 Krejci 1.232 0.613 101.0%
2009-10 Savard 0.768 0.676 13.6%
2008-09 Savard 1.277 1.098 16.3%

Drew Stafford

Stafford Teammate With Without % Impr.
2011-12 Roy 0.939 0.585 60.5%
2010-11 Roy 1.137 0.926 22.8%
Connolly 1.407 0.765 83.9%
2009-10 Roy 0.913 0.876 4.2%
Connolly 1.318 1.01 30.5%
2008-09 Roy 0.965 0.809 19.3%
Connolly 1.384 1.195 15.8%
2007-08 Roy 1.415 1.235 14.6%

Ryan Clowe

Clowe Teammate With Without % Impr.
2011-12 Couture 0.984 0.722 36.3%
2010-11 Couture 0.933 0.733 27.3%
Pavelski 1.062 0.791 34.3%
2009-10 Pavelski 0.863 1.24 -30.4%
2008-09 Pavelski 0.922 0.557 65.5%

Scott Hartnell

Hartnell Teammate With Without % Impr.
2011-12 Giroux 1.286 0.812 58.4%
Jagr 1.309 0.683 91.7%
2010-11 Briere 1.169 1.064 9.9%
Leino 1.053 0.987 6.7%
Giroux 1.205 1.038 16.1%
2009-10 Briere 0.918 0.931 -1.4%
Carter 1.04 0.606 71.6%
M. Richards 0.719 0.707 1.7%
2008-09 Carter 1.299 0.699 85.8%
Lupul 1.272 0.327 289.0%
M. Richards 1.016 0.911 11.5%
2007-08 M. Richards 0.696 1.214 -42.7%
Briere 0.883 0.581 52.0%
Carter 0.936 0.754 24.1%
Lupul 1.305 0.923 41.4%

Dustin Penner

Penner Teammate With Without % Impr.
2011-12 Stoll 0.644 0.408 57.8%
M. Richards 0.577 0.494 16.8%
2910-11 Hemsky 1.004 0.714 40.6%
Cogliano 0.856 0.545 57.1%
2009-10 Gagner 1.063 0.653 62.8%
Brule 1.505 0.782 92.5%
Cogliano 1.171 0.669 75.0%
2008-09 Horcoff 1.123 0.724 55.1%
Hemsky 1.32 0.658 100.6%
Cogliano 1.09 0.671 62.4%
2007-08 Hemsky 0.974 0.741 31.4%
Horcoff 1.015 0.696 45.8%

Brenden Morrow

Morrow Teammate With Without % Impr.
2011-12 Ribeiro 0.887 0.642 38.2%
Eriksson 1.055 0.935 12.8%
Ott 1.069 0.880 21.5%
2010-11 Ribeiro 1.104 0.365 202.5%
Benn 1.040 0.900 15.6%
2009-10 Ribeiro 0.911 0.507 79.7%
Benn 0.923 0.692 33.4%
2007-08 Ribeiro 1.196 0.384 211.5%
Mietinen 1.087 0.808 34.5%

Nathan Horton

Horton Teammate With Without % Impr.
2011-12 Krejci 1.298 0.455 185.3%
2010-11 Krejci 1.276 0.632 101.9%
2009-10 Weiss 1.153 0.525 119.6%
2008-09 Weiss 1.141 1.094 4.3%
2007-08 Weiss 1.191 0.543 119.3%

Save for a small number of player combo seasons the big strong wingers made their smaller skilled linemates (particularly the centermen) better offensive performers and while most of the players I looked at above are quality players no one will really call them elite offensive stars that can carry an offense.  They are at best secondary top line players or second liners.  Now it could be that when some these guys are not on the top line they are replaced with a third line player who brings down the production of the top line but there does seem something happening here that makes me think if you have a small, skilled center you should really look to pair him up with a big, strong, winger even if that winger is less talented.

 

 

Jan 302011
 

Yesterday there was a post on the Behind the Net Blog which discussed the Washington Capital’s 2009-10 even strength shooting percentage of 11.0% and the conclusion was that it must be mostly luck which resulted in a shooting percentage that high.  But was it?  It was noted in the article that in 2007-08 the Capitals shot at 8.1%, in 2008-09 they shot at 8.2% and this season they are shooting at 8.2% again.  So clearly 2009-10 appears to be an anomaly, but was it a luck driven anomaly or something else?

Most people in the hockey analysis world have been using a simple binomial distribution to simulate luck so I’ll do that here too.  The thing is, if the Washington Capitals were really a 8.2% shooting team last year, the chances of them shooting 11.0% or better on 2045 shots is a mere 0.0042%.  That kind of luck we should expect once every 8000 NHL seasons.  In short, we can be pretty confident that the Capitals 11.0% shooting percentage wasn’t all luck driven.

So the next question is, how much of it is luck, and how much can we attribute to other factors?  Well, let’s assume that their good luck was significant to the point where there would only be a 5% chance they could have experienced even more luck.  We can do this by constructing a binomial distribution using centered on a shooting percentage where the chance of producing a shooting percentage of >11.0% is 5%.  The result is shown in the following chart:

The far left vertical line is the number of goals that Washington would produce if they had an 8.2% shooting percentage and the far right line is their actual shooting percentage.  The center vertical line is the theoretical shooting percentage we would need to meet the 5% luck conditions outlined above.  Under this scenario one could suggest of the extra 57 goals that Washington scored above what they would get if they shot at 8.2%, 22 of those goals can be attributed to luck and 35 can be attributed to skill.

But what if we assumed the Capitals were extremely lucky and there was only 1% chance of having greater luck.  Under that scenario their true talent level would be 9.49% shooting percentage and 26 goals would be due to skill and 31 would be due to luck.

Regardless of how you want to look at it, a significant portion of the Capitals elevated shooting percentage was likely due to non-luck factors, be they actual talent, playing style, score effects, etc.