48 game season! Anything could happen! Too small of a sample size! Statistical anarchy! The Columbus Bluejackets might make the playoffs!
But are 48 games really too short? What makes 82 games so great? Are there an optimum number of games to play before knowing which 16 teams are truly fit for the playoffs? I decided to take a look at the performance of some teams from last season to see if there was an answer.
In order to do this I graphed out the point percentage of each team as they progressed through the season. This percentage is calculated on a 5 point scale, with 5 points for a regulation win, 4 points for an OT win, 3 points for a SO win, 2 points for a SO loss, 1 point for an OT loss, and 0 points for a regulation win (point percentage data came from Hockeydb.com). The idea is that as the season progresses, the point percentage should level out, converging on a percentage that represents that team's ability to win games. Once this line converges, the proper number of games have been played to evaluate that team.
Here is what the central division looked like at the end of last season:
Fig 1. Columbus gets a pink line for obvious reasons.
At the beginning of the season the sample size is too small, so there is a large amount of variation. About 20 games in, the trends begin to come into focus. Nashville's percentage then begins to improve and plateaus nicely around the 50 game mark. St. Louis stays even for most of the season. Detroit does too, until around the 60 game mark and then begins to decline, while Chicago is all over the place. And Columbus is, well... Columbus.
So at the 48 game mark the standings are St. Louis, Detroit, Chicago, Nashville, and Columbus. At the end of the season the standings are St. Louis, Nashville, Detroit, Chicago, and Columbus. Which set of standings is more valid?
It is hard to say. With injuries, trades, coaching changes, and other factors, a team's point percentage value can change, causing the trend to try to converge on a new value. This is what I believe happens to Detroit at around the 60 game mark, when the team caught a bad case of being old. Nashville's rise in the middle of the season can be attributed to players coming off of injury. And St. Louis's jump at the beginning of the season is probably the work of Ken Hitchcock, although it is early enough that it could be a statistical aberration and Ken could be lucky as hell to land a team that good.
Of course, there were a few teams that looked entirely different from one half of the season to another. Minnesota became the poster child for "regression to the mean." Toronto buckled under the pressure of all the hopes and dreams of the Canadian people. Here is a look of their trends compared with the eventual Stanley Cup finalists:
At the 48 game mark these 4 teams are virtually on top of each other. By the end of the season though, New Jersey and LA had clearly separated themselves from Minnesota and Toronto. Minnesota and Toronto seemed to be the only two teams whose position from the 48 game mark to the end of the season didn't have and obvious change in player personnel or coaching to trigger it.
So are 48 games enough to pick out playoff teams? For the most part, I'd say yes. Sure the standings will be different than at the end of an 82 game season, but the differences can largely be attributed to trades and injuries. However, there could be a couple funky teams that grab the 8 seeds. And once the playoff start, it's a crap shoot anyway.
For the fun of it here is the Atlantic Division as well:
(Would have graphed out more but I got lazy :P)