I attended the 2012 MIT Sloan Sports Analytics Conference last weekend in Boston. Additionally, I gave a poster and was invited to be part of the Hockey Analytics panel with Mike Milbury, Tony Amonte, Peter Chiarell and Brian Burke. NHL.com had a write-up of the panel here. I’m going to put together some posts/comments on that panel and some of the topics that came up.

The first of these is the topic of faceoffs. I said in the panel that it takes an additional 100 faceoffs wins to get an extra goal.  As it turns out at 5v5, on average,  it takes an about 165 faceoffs to get your team a goal but as we have a zero sum game, an extra 82 faceoffs gets you a goal differential.  This is based upon just over 100,000 faceoffs taken at 5v5 over the past two completed regular seasons.  What we (Schuckers and a student at St. Lawrence University, Jim Curro) did was calculate the probability that the home team scores – the probability that the away team scores for 20 seconds after each faceoff.  We called this the NP20 for net probability at 20 seconds.    The table below gives you the breakdown of NP20 by winner and by location for 5v5 faceoffs.

If we take a weighted average of the NP20 (weighted to reflect the distribution across locations) we find that the average NP20 gained by a faceoff win is 0.006094.  Thus, it takes about 1/0.006094 = 165 faceoff wins to get a goal.  That’s an average number but it gives a good guide as to the worth of a faceoff.

Winner
Location
(Relative to Winner)
NP20
Count
(n=   )
HOME
Offensive Zone
0.01669
16241
AWAY
Offensive Zone
-0.01524
14765
       
HOME
Neutral Zone
0.00259
20085
AWAY
Neutral Zone
-0.00358
19546
       
HOME
Defensive Zone
0.00049
16412
AWAY
Defensive Zone
0.00012
16010

Some further details on the analysis that Jim Curro and I did can be found here.