The bugs continue in the NHL’s enhanced stats

Chris Oddo asks:

I just noticed a massive discrepancy between the NHL site and yours on Offensive zone faceoff percentage.

I mean they have Tavares winning over 70 percent for the last four years and you have him down around 55, which seems more realistic. They have Paul Gaustad winning 212 out of 233 defensive zone faceoffs. That can’t be, right?

I was a bit concerned at first to see that my numbers might be wrong and also that I had somehow missed the superhuman talent of Paul Gaustad, with his 91% defensive zone face-off win percentage. The good news is that my numbers are fine; the bad news is that the NHL still hasn’t fixed their bug-laden enhanced stats.

The short answer to the above discrepancy is that on they are counting offensive zone face-off losses as defensive zone face-off losses and vice versa.

Here is the longer answer. The NHL keeps play-by-play files for each game (which I get my data from, and so do other similar hockey stat sites). Face-offs are recorded in the play-by-play files like this:

FAC NSH won Def. Zone - WSH #19 BACKSTROM vs NSH #28 GAUSTAD

The zone is always from the perspective of whoever won the face-off. So it was a defensive zone draw for the winner, Nashville’s Paul Gaustad, and an offensive zone draw for the loser, Washington’s Nicklas Backstrom.

But if you’re lazy, or don’t really know how hockey works, you might just see the words “Def. zone”, count that as a defensive zone face-off for both players and call it a day. This is what seems to currently be doing.

Compare my numbers for Gaustad with

SourceOff. zone FWOff. zone FLDef. zone FWDef. zone FL 28 21 212 172 28 172 212 21

Note the numbers are the same and the face-offs won are correct; they just have the losing columns mixed up.

Using Gaustad’s numbers as an example makes’s error obvious, because Gaustad is used very heavily for defensive zone draws and almost never in the offensive zone. For a player whose face-offs are more evenly spread out, by just looking at the incorrect numbers you might never even notice something was wrong.