NHL Preseason: Perception vs. Reality

Updated: January 11, 2018 at 12:23 am by Jayson Spikes


Players cannot possibly care as much about the preseason as the regular season. It’s simply human nature, and I think, as fans, we get that, usually.

Until, y’know, the Oilers start 5-0, the Flames start 2-2, Leafs start 2-2, the Canucks start 2-3-1 and the Golden Knights almost score double digits in their franchise debut. Then it can be hard as a fan to not get excited (if you’re an Oilers fan), anxious (if you’re a Flames, Leafs or Canucks fan) or bewildered (if you understand the phrase “expansion team”).

So today I wanted to blog about the possibility that the public systematically underestimates the impact low or zero motivation may have on the outcome of games. The was explored in the gambling context by Preseason bias in the NFL and NBA betting markets, where the authors found compelling evidence to support the case.


Gamblers need to win, on average, 52.38% of evenly denominated point spread bets to break even long term (it isn’t 50/50 because the sportsbook takes a commission on every bet).

However, despite this, the efficient market hypothesis suggests that, over a sufficiently large sample size, through the interaction of sports books and the gambling public, on average, the two-goal favourites win by an amount very close to two goals and that there is no obvious strategy to win enough to make consistent profits, other than simply knowing more than the general public.

And the data backs this up. In the NFL (1995-2014) and NBA (2005-2014) the underdog wins (with the point adjustment for being an underdog) 50.33% and 50.10% of the time, respectively. These numbers are very close to the 50/50 the efficient market hypothesis would suggest.


But the above numbers apply to the regular season. The authors of the paper started with the hypothesis that bettors over complicate the meaning and motivation of preseason games. That is, the odds of two unmotivated yet still competent teams (i.e. composed of professional, highly trained athletes) is probably closer to a coin toss than the market is willing to accept.

The argument is compelling and intuitive. There are several reasons for increased uncertainty in the preseason over the regular season:

  • Lack of motivation to win because of minimal benefit
  • Fear of injury
  • Partial or Non-participation of key players
  • Actively playing unproven players to evaluate them
  • Variable motivation of players playing (e.g. Getting into game shape vs trying to make the roster)
  • New rules being tested out by leagues (I’m looking at you, parade to the penalty box caused by the new face off rules and slashing crackdown)


Taking the data from those 1995-2014 (NFL) and 2005-2014 (NBA) seasons, the researchers identified 5 different groups of games and tested three generic gambling strategies for each.

The different pools of games were:

  1. All NFL regular season games
  2. All NFL preseason games
  3. All NFL preseason games excluding week 3 (when teams tend to play their starters and games are closer to “regular season” than normal)
  4. All NBA regular season games
  5. All NBA preseason games

Within each one of these groups of games, the authors looked at the results of underdogs covering the point spreads in games of different levels of perceived competitiveness:

  1. All games
  2. Games with a point spread greater than 3 points
  3. Games with a point spread greater than 5 points

The results are shown in the following table:

Strategy 1

(All underdogs)

Strategy 2

(>3pt underdogs)

Strategy 3

(>5pt underdogs)

All NFL regular season




All NFL preseason




NFL preseason excl. week 3




All NBA regular season




All NBA preseason




These results are quite fascinating. Looking at the regular seasons for the NFL and NBA in the time period examined, the results were pretty close to 50/50, and nowhere near the 52.38% required for gamblers to make a consistent profit. Which means the market is very good at predicting these outcomes over sufficiently large sample sizes.

However, in the preseason, it looks like (for the periods and leagues studied) the market seems to think there is a bigger skill gap than there actually is in the preseason. Interestingly, this market error seems to get bigger with match-ups between teams that would probably be more lopsided in the regular season and all but one of them being larger than the 52.38% needed to earn a consistent profit and most of them being nearing of passing statistical relevancy tests.


At least in the NFL and NBA (in the years studied) there is good evidence that gambling markets are good at pricing regular season games, but that there is information not being accounted for in the preseason making it more likely that underdogs win than the market thinks should be the case (with the result increasing for larger underdogs).

The authors suggest that there are more random variables in the preseason pushing results closer to a coin toss (low motivation, resting players, new league rules, etc.) that the betting public isn’t accounting for

While the same analysis was not done on NHL preseason games, I don’t know why the same analysis wouldn’t apply. Which means that Flames, Leafs and Canucks fans probably shouldn’t start worrying yet and Oilers fans should make sure to keep their expectations reasonable.

And all three fan bases can still probably look forward to beating up on the Golden Knights this coming season.