Filed under:

# WOWY Is Worthless: Part 2

WOWY is worthless. It's things we know from Corsi plus a big dose of statistical noise. In Part 2, I use WOWG (With or Without Goalie) to look at WOWY.

Before I get into the comparison of WOWY to WOWG, let me first review WOWY a little. In part 1, I defined WDiff as (Corsi With) – (Corsi Without). Across the league, there are 24,875 valid WOWY pairs. In part 2 and part 3, for convenience I'm going to use Corsi%, so I threw out any pair that had either 0 events With or 0 events Without. Since small sample sizes are a problem, I also looked at the pairs that had at least 100 events With and 100 events Without (called "100/100") , the pairs that had 200 events With and 200 events Without ( "200/200"), the pairs that had 300 events With and 300 events Without ( "300/300"), and the pairs that had 600 events With and 600 events Without ( "600/600"). Looking at the range of observed WDiff we see

 group mean sd 0% 2.5% 50% 97.5% 100% n All -1.1 16.6 -100.0 -44.5 -0.6 43.2 100.0 24875 100/100 0.0 5.5 -23.7 -10.9 0.0 10.7 19.9 10623 200/200 0.3 4.8 -22.1 -9.1 0.3 9.8 19.6 6947 300/300 0.4 4.5 -16.4 -8.4 0.4 9.4 19.0 4893 600/600 1.0 4.0 -11.7 -6.7 1.1 9.3 15.1 1335

Going from all pairs to 100/100 things narrow down a lot. Going on to 600/600 narrows things only a little more.

WOWG

So I went through the 2013-14 season and computed WOWY for every skater with each of their team's goaltenders (WOWG). Goaltenders have no effect on Corsi production rates so they serve as a form of placebo (I see some of you with your hands raised. I will answer your question in a minute). So what does the distribution of WOWG WDiff look like? Once again I looked at all pairs, 100/100, 200/200, 300/300, and 600/600.

 group mean sd 0% 2.5% 50% 97.5% 100% n all -19.0 40.8 -100.0 -100.0 -1.6 19.4 100.0 3537 100/100 0.0 4.6 -17.1 -9.0 -0.1 9.3 17.1 1772 200/200 -0.1 3.9 -12.1 -7.6 -0.2 7.6 13.8 1377 300/300 -0.1 3.6 -11.2 -7.1 -0.2 7.1 12.0 1120 600/600 -0.1 3.0 -10.1 -6.1 -0.1 5.9 8.5 414

Comparing the 100/100 groups, the standard deviation of WOWG is 84% of the standard deviation of WOWY. For 600/600, it's 75%.

In the 300/300 group we see goalies with "effects" of -11% and +12%. There are only 47 WOWY pairs above 11% and many of those include pairs like Player/Patrice Bergeron, Player/Sidney Crosby, and Player/Daniel Sedin. Interestingly, Player/Mark Giordano appears 4 times and Player/TJ Brodie 3. On the low end, only 28 WOWY pairs are below -11%, and many involve Player/Manny Malhotra, Player/Tanner Glass, Player/Andrew MacDonald, and Player/Max Lapierre. You don't need WOWY to know that Patrice Bergeron is going to make people better or that Manny Malhotra is going to make people worse.

Lack of persistence

If we look at the players who aren't paired up with stars who had a WDiff above 11%, do any of the pairs show persistence from season to season?

 Player With 2013-14 WDiff 2012-13 WDiff 64 MIKAEL GRANLUND 29 JASON POMINVILLE 19.0 -27.4 27 CRAIG ADAMS 3 OLLI MAATTA 14.3 NA 42 TYLER BOZAK 21 JAMES VAN RIEMSDYK 13.9 -0.6 44 MORGAN RIELLY 51 JAKE GARDINER 13.4 NA 55 MARK LETESTU 21 JAMES WISNIEWSKI 13.1 6.1 44 NATE THOMPSON 77 VICTOR HEDMAN 12.9 -4.6 40 TROY BODIE 12 MASON RAYMOND 12.8 NA 5 JASON GARRISON 14 ALEXANDRE BURROWS 12.3 8.2 16 ELIAS LINDHOLM 11 JORDAN STAAL 12.1 NA 55 ED JOVANOVSKI 44 ERIK GUDBRANSON 12.1 -28.8 44 BROOKS ORPIK 36 JUSSI JOKINEN 12.0 -6.2 47 TOREY KRUG 63 BRAD MARCHAND 11.9 NA 28 JARRET STOLL 6 JAKE MUZZIN 11.9 -4.5 4 ROB SCUDERI 2 MATT NISKANEN 11.8 NA 48 TOMAS HERTL 88 BRENT BURNS 11.7 NA 84 MIKHAIL GRABOVSKI 81 DMITRY ORLOV 11.5 NA 5 MARK STUART 18 BRYAN LITTLE 11.4 -2.4 74 ALEXEI EMELIN 11 BRENDAN GALLAGHER 11.4 4.1 81 DMITRY ORLOV 52 MIKE GREEN 11.3 -24.9 59 ROMAN JOSI 57 GABRIEL BOURQUE 11.2 -5.1 33 HENRIK SEDIN 18 RYAN STANTON 11.1 NA 6 BEN LOVEJOY 10 COREY PERRY 11.1 5.9 23 VILLE LEINO 57 TYLER MYERS 11.1 1.6

You don't need to look beyond the first line of the table to see the problem.

Are Goaltenders really placebo?

Yes. Presumably, if a goalie is going to have an effect on Corsi, he's a good puck handler. Two of the best puck handling goalies are Martin Brodeur and Mike Smith. In the 300/300 data, Brodeur's average WDiff is -0.9%. Smith is near the bottom of the list at -3.0%. The goaltender with the biggest impact in this group is Anders Lindback with a WDiff of +4.7% Please don't tell me you think Lindback is a good puck handler.

Goaltenders have no effect on Corsi production but the overall distribution of WOWG WDiff is similar to the distribution of WOWY WDiff.  There are a number of skater-goalie WOWG pairs where the goaltender appears to have a substantial effect on the player. That's not effect. That is simply randomness.

I suppose some people still don't feel like goalies are placebos. So, in part 3, I bring you a new and improved placebo, now with even less active ingredient!