Hockey, Sports

“Real” Per Game Hockey Stats

Hockey has been very slow to pay attention to per game numbers as a means of evaluating players compared to, say, a sport like basketball, where they are virtually everything. This has been especially true with the Hall of Fame, where longevity and championships (i.e. luck) are favoured above a player’s ability to perform (with the notable exception of the “50 goals in 50 games” fetish).

This has always been a curious approach since it is obvious that a player who scores, say, 75 points in a 70 game season has done something fairly more impressive than a player who scores 75 points in an 82 game season.

But recently there has been more of a focus on per game numbers (at least outside the Hall). This helps evaluate players who get hurt. For example, look at Crosby’s out of this world per game numbers this season. More recently, some people (like Scott Cullen of TSN) have embraced another basketball concept: the per minutes stat.

The idea in basketball was to create a statistic that was more accurate than per game. For example, a player scores 27ppg during a season. If this player plays 500 more minutes than another player who scores 26ppg, there is an obvious problem with judging based on per game. 27ppg looks superior until we learn that 26ppg accomplished his similar numbers in far less time. This is why PER was developed. But I’m not talking about anything so sophisticated (and flawed).

To fix this, basketball started talking about per 48 minutes. And that is where hockey appears to be now. Cullen and others have started talking about per 60 minutes.

But there is an obvious problem with this, which was recognized within the basketball world when they changed it to per 36 minutes. The idea is the 36 minutes is approximately the amount an average starter on an average team plays during a game. So his per 36 minutes stats are far more accurate than his per game stats or his per 48 minutes stats.

Hockey isn’t there yet. Which is why I am proposing a couple of ideas.

Since 1997, we can calculate the average number of minutes a player plays. (We cannot figure out “real” per games before this since the NHL didn’t care about minutes). So, for example,

  • The average Top 6 forward this year played approximately 18 minutes and 18 seconds each game.
  • The average Top 4 defenceman played approximately 22 minutes and 21 seconds. Given that defencemen nearly almost always play more minutes than forwards, I think we should keep these two separate.
  • But if we wanted to combine them we could say that the average non-role player plays about 20 minutes a game (the reason to focus on non-role players is because the aim is to focus on what a player’s production would be if he were to play as a “starter” rather than as a role player; there needs to be a standard, rather than many standards).
    I think it makes far more sense, if we are to start talking about per minute stats in hockey, to speak about per 18 minutes for forwards, and per 22 minutes for defencemen (alternatively, per 20 minutes for players).

I think this would give us a far more accurate idea of what a player does in an average game than the per game numbers. It only helps us since for the last 15 years give or take. But it’s better than what we have.

So hopefully when evaluating post-1997 players, like Iggy, I will at least mention their per 18/22 minutes stats. If I don’t, remind me.

1 Comment

  1. I had to go to work so I stopped writing.
    Some caveats:
    The idea isn't so much to obssess over "real per game" stats so much as to try to creat "real per 82 games" stats (using seasons of 1476 minutes for forwards and 1804 minutes for D). This then gives us a sense of what a player accomplishes in a season assuming he plays the average amount of ice time for his position.
    Clearly this is pretty much only useful for evaluating offensive effieciency. I don't see how it could be useful for defence. I guess if we combined it with the idea of Goals Against shares that I mentioned in an earlier post, there might be something to that.
    But alone this is only useful for evaluating forwards and offensive defensemen (for example, using this probably would have shown how stupid the McCabe and Souray contracts were, as they signed them). Moreover, any top 6 forwards who kill penaltys in addition to playing on the power play will have their numbers skewed somewhat.
    Still I think using this lets us arrive at somethinb better than just using per game numbers to calculate a per 82 game goal, assist or point total.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.