Monday, August 13, 2007

Weight For It

Next up on my stat kick, I thought I'd take a look at weighted goals. The theory behind weighted goals is that goals scored in a tight game are more important than goals scored when the score isn't that close. There may be a few exceptions to that logic, but for the most part, it makes sense. For more info on it, check out this article from Marc Foster and Chris Apple.

The way that these numbers were calculated were that goal that resulted in a tie game or one-goal lead was worth one point. Any goal that cut a lead to one goal and increased a lead to two goals was worth .75 points. A goal that cut a lead to 2 or increased a lead to 3 was worth .5 points, and all other goals were worth .25 points.(I made a slight modification to the original formula by getting rid of the .1 points for a goal that increased a 5 or more goal lead, or cut a 6 or more goal lead)

Results are presented as such: Goals Scored, Weighted Goals Scored , Percentage of Weighted Goals/Goals Scored.

I thought it'd be interesting to start by looking at some of the top Hobey Baker candidates from last season to see how they did.

Ryan Duncan, North Dakota 31 26.25 85%
Eric Ehn, Air Force 24 19.5 81%
T.J. Hensick, Michigan 24 14.5 60%

Duncan's numbers look pretty impressive there. But how do there compare to some other players? I (unscientifically) picked a few big scorers from last season to see how they compared.

Mike Santorelli, Northern Michigan 30 20.25 68%
Scott Parse, Nebraska-Omaha 24 15.5 65%
Jay Barriball, Minnesota 20 16.25 82%
Travis Morin, Minnesota State 17 12.75 75%
Nathan Davis, Miami 21 15 71%
Andrew Gordon, St. Cloud State 22 16.5 75%
Bryan Lerg, Michigan State 23 17 74%
Jon Kalinski, Minnesota State 17 12.75 75%

That makes the average weighted goals percentage 74%. So in comparison, Duncan's 85% looks even more impressive, while Hensick's was definitely on the low end. At the same time, you could argue that the numbers are a bit biased by league. The CCHA tended to be higher-scoring and have more lop-sided games last year, both of which hurt their players in this measure.

And because I know you're curious, I liked to make fun of Phil Kessel during his short tenure at Minnesota for scoring a lot of meaningless goals. So how did Phil do in this measure? Not too bad actually. Kessel had 18 goals and 13.25 weighted goals for a percentage of 74%. So my apologies go out to you, Phil Kessel.


Eric Carlson said...

Interesting post. Kyle Greentree scored 21 goals for Alaska on which he compiled 18.5 points for an 88.1%. All but two of his goals were for .75 or more. Just for comparison Curtis Fraser scored 19 goals on which he had 14.5 points for a 76.3%. Nobody would probably argue here that Greentree didn't score more big goals. They just wanted him to score more of them in the second half of the season! I still don't know exactly what this shows though.

Donald said...

I tend to think this is less than significant. An "important" goal is often only made so by the goals that preceeded it.

Your team is down 3-0 ...

Which goal is more important? The first one that gets your team into the game? The second one that makes it a one goal game? The goal that ties the game? Neither the 2nd or the 3rd would be important without the 1st would they?

Assigning an arbitrary significance to which goal is more important doesn't make sense to me. Isn't it the culmination of all the goals that counts?

I didn't read any of Chris' links in this article but I'd imagine this is somehow derived from hideousness that is baseball.

Chris said...

Stats are only good up to a certain point, and with this, like most, that point is much closer to "That's kind of interesting" than "OMGUniversal Truth about Hockey".

I think it does provide a nice glimpse of who is scoring goals in close games, and who is scoring a lot with the game mostly decided, especially on the extremes. In that regard, Kyle Greentree's numbers are pretty phenomenal.

Marc Foster said...

Donald, not all the metrics in my articles are direct derivatives of, or otherwise inspired by sabermetric methodologies. Disciplined Aggression is one such. I've also been spending some time on new analytics based on the NHL's RTSS data. Unfortunately, I don't have access to the RTSS raw streams (the decoder sheets for which are 6-7 pages long), so I don't know when I'll be able to test them.

The data only gets you so far, determining what the data means gets you much further down the road.