Monday, June 7, 2010

Shot Recording in the NHL

Shot recording bias is a topic that I have always wanted to look at but haven't yet got around to, until my latest article over at Puck Prospectus. I think it is something that has a real impact on goalies around the league and that we need to avoid rating goalies based on rink effects.

I use the zone time metric in my analysis, which unfortunately limits me to a three year period almost a decade ago. It would have been nice to have a larger sample size, and to try to look at shot effects in today's NHL, but alas the NHL decided to discontinue tracking zone time after the 2001-02 season.


Bruce said...

Interesting stuff, CG. Recording bias is a big concern, seems to rear its ugly head everywhere.

League-wide there was an average of 70.8 shots per 60 minutes of time spent in either attacking zone. Given the numbers above that indicate the puck was in either attacking zone 78.6% of the time in St. Louis home games and 74.8% of the time in St. Louis road games, it follows that an average team playing against average opposition would have an expected shot total of 55.6 total shots per game at home and 53.0 shots per game on the road. St. Louis’ actual totals were 52.1 total shots per game at home and 53.1 shots per game on the road, which indicates that the official scorer there may have been undercounting.

He may have been undercounting shots, or overcounting offensive zone time. No real way of knowing which recorder was off the mark.

But if your conclusion is the right one, maybe Roman Turek had a better Sv% than the one he's credited with. ;)

Vic Ferrari said...

If I may make a suggestion.

Ignore neutral zone time (if there is an area affected by scorer bias .. it will be that imo)

Use Off.Zone.Time / (Off.Zone.Time + Def.Zone.Time).

Call that zone% or some such.

compare that to shots, fenwick and corsi over the same period.

Ideally, just road numbers, but either way it's good.

I'd do it myself, but you seem to have the data handy, Phil.