Skill, Luck, and the NHL's Shortened Season

[Renew Deal]

<snip>
These situations seem similar on the surface. The leagues play an identical number of games during the same time of year and the sports have a like number of players in the game at a time. Basketball has more scoring, but hockey games are longer. Playing a slate of games that is 60% or more of the full season seems to be a comparable solution.

But if the objective of the regular season is to determine which teams are best, there is a huge difference between the NBA and the NHL. The key is the relative contribution of skill and luck in determining results for a season. Of the professional sports leagues — which also includes include the National Football League, the Premier League (soccer), and Major League Baseball — skill plays the largest role in the NBA (PDF) and the smallest role in the NHL.

Here's the way to think about it. When a sport has lots of skill and little luck, you don't need a large sample size to determine which team or individual is more skillful. Less than 1/6th of a minute of running against Usain Bolt, for example, would do the job completely. But when there's lots of luck, you need a large sample size to ensure that luck evens out and skill shines through. Baseball has a lot of luck, but you can be reasonably confident that the better teams make the playoffs because the regular season includes 162 games (over 27,500 minutes).

To make this concrete, we can calculate the number of games you need in the NBA and the NHL to get an equivalent contribution of skill and luck. Because basketball has lots of skill, you only need 10-15 games, or about 15% of the season, to obtain the same signal that you get from 70-75 hockey games, or nearly 90% of the season. Sixty-six games for the NBA is plenty to allow the top teams to climb to the surface. But 48 games in the NHL is way too few. The NHL season is simply too short to be confident that the more skillful teams will rise to the top. If there's a year for a mediocre team to do unexpectedly well, this is it.
<snip>

http://blogs.hbr.org/cs/2013/01/skill_luck_and_the_nhls_shorte.html

[caraher]

1. What a load of crap - it isn't about skill vs. luck at all

First, where's he's obviously right: in a shortened season, there's a larger chance of weaker teams edging stronger teams out for playoff spots, in any sport. This is a particularly acute problem in pro football, given the small number of games. Some stats freaks have done a decent job of figuring out how long a season it takes to reduce the odds of the standings being "wrong" to some basically arbitrary level. That's kind of interesting, I guess, but nobody involved in deciding whether to cancel the season is going to have this as their chief concern.

Where he's wrong:

1. The basic premise that "the objective of the regular season is to determine which teams are best." Sure, that's one outcome one would hope for, but professional team sports exist for the entertainment of their fans. (OK, I'll concede to cynics who say they exist to make money for their owners (I supposed excepting the Packers), but in principle they earn that money by providing something of value to fans, and whether its fodder for water cooling conversation, stats for a fantasy league or just burning hours staring at the TV, ultimately it all boils down to entertainment.) Even if one accepts all the statistical arguments, nothing in them suggests that 48 games do not make the season worthwhile to fans or players - only that there's a somewhat greater chance that the 16 "truly best" teams don't make the playoffs. These stats-obsessed people seem almost offended that the NBA season is so long...

2. His sources don't really back up his language. In particular, the word "skill" is used inappropriately at every turn. What he refers to as "skill" vs. "luck" really boils down to competitive imbalance at the level of single games. In other words, because "upsets" are less likely in an NBA game than an NHL game, that somehow means the outcome of NBA games reflect "skill" more than NHL games. His chief source, Tom Tango, actually says, "The question on the table is how many games do you need to get the truly better team to have the better record." "Truly better" implies nothing at all about skill, and this is only driven home further by the next source cited, an article by some economists whose title is, "The Short Supply of Tall People: Competitive Imbalance and the National Basketball Association." Their speculative hypothesis is that the performance of pro basketball teams naturally tends to be more stratified because height is so important in player performance, and the results of games are dominated by a small handful of exceptionally tall and athletic players. How that translates into basketball being more a game of "skill" than hockey, baseball or football is beyond me! Rather than framing the situation in terms of "skill vs. luck," "competitive balance" is a much more accurate way of talking about the topic.

Reading through his sources, my main take-away is that the level of competition for the fan attending a single game is the best in the NFL, similar in the NHL and MLB, and worst in the NBA.

The other thing he mentions but glosses over, that's a huge difference between hoops and hockey, is that while (leaving aside goalies) both games involve 5 players at a time during play, basketball teams have smaller rosters and don't usually go nearly as far down the bench as a hockey team. You'll typically see at least twice as many players see significant game action on a hockey team compared to a basketball team, and that's probably a real source of competitive balance: an NBA star regularly plays 40+ minutes of a 48 minute game, while an NHL star logging more than 30 minutes of ice time is really pushing it. As a result, the guys on the second and third lines are much more important than NBA benchwarmers, and it's harder to build a team with a significant edge all the way down a lineup that deep.

Along with measures the league has taken to ensure competitive balance, the roster size effect might also be a big part of the competitiveness of NFL games. Just looking at starters you've got 22 players plus kickers, and at most positions you need a good stable of backups to cover for the inevitable injuries.