Boy, there’s nothing like a combination of the HHOF induction ceremony and a sports analytics conference in Toronto to produce a lot of quotes from general managers about analytics in sports.
Chuck Fletcher on discrepancies in NHL stats: "In Det. you're going to get 10 more shots on goal in a given night than you will in Minn."
— Craig Custance (@CraigCustance) November 12, 2013
In Chris Johnston’s piece about the conference and the HHOF, he had an expanded quote from Fletcher on the point:
Even Minnesota Wild GM Chuck Fletcher, who is a proponent of advanced stats, raised the issue on the red carpet at the Hockey Hall of Fame’s induction ceremony.
“There are a lot of inconsistencies,” said Fletcher. “For example, in Detroit you’re going to get 10 more shots on goal in a given night than you will in Minnesota. If a puck (barely) misses the net in Minnesota, it’s not a shot on goal.
“In Detroit, they might give you two.”
I gathered the home and road 5v5 shot data for Detroit and Minnesota to look into this.
I’m inclined to think that Fletcher’s basic point is true and that there are probably more official shots on goal in a Minnesota game in Detroit than a Detroit game in Minnesota. I’m not sure that this is a Detroit problem though – Detroit’s scorers record, on average, about a shot a game more per 60 minutes of 5v5 in Detroit than is recorded in Red Wings games on the road. It seems entirely possible to me that there actually are slightly more shots in Red Wing home games than there are in Red Wing road games.
Minnesota’s home/road splits are a bit more suspicious looking but again, don’t seem outside the realm of the possible to me. The home team gets to dictate play more at home because of the last change, which makes it easier to impose fun hockey (Detroit) or boring hockey (Minnesota, historically).
What if we look at every team in the NHL from 2007-13 and compare their home results to the their home results, in terms of how many shots/60 at 5v5 the scorers see.
As you can see from the graph, 22 teams are inside of +/- 1.5 shots/60 between their home and road scorers. Carolina, the least extreme home team in that grouping, sees 2.3% more shots in their games on the road. Columbus’ road games featured 2.6% fewer shots than their home games over the period in question. We are talking about tiny differences for most teams in the NHL.
Minnesota actually stands out as the team with the biggest separation between home shots counted and road shots counted. Wild home games between 2007-13 featured the 29th fewest shots in the NHL. Detroit home games featured the sixth most. You can see where Fletcher gets the idea that Detroit is a shot counting fantasyland. The problem is this: road scorers agree with the home scorer in Detroit. There are a lot of shots in Red Wing games. Meanwhile, the scorers leaguewide say that Minnesota either can’t count shots or that the Wild play a much more stingy game at home.
Again though, the gap is awfully small. Minnesota’s scorers counted shots at 92.9% of the rate of scorers in road games involving the Wild. It’s significant, but it’s not huge. Moreover, assuming that the scorers make mistakes each way, the percentage isn’t really going to be affected – the average team did 2.3 percentage points better in terms of 5v5 shots in home games and Minnesota’s scorers were close to that.
All of this is to emphasize: attacks on the league’s data quality are grossly overblown. Whether the league properly records Johnny Bumbler whacking away at a given rebound right or not doesn’t really matter. Better data would be better but what exists is good enough to start thinking about things and challenging our beliefs.
Pretending otherwise is just that: pretending.Email Tyler Dellow at firstname.lastname@example.org