Over/Under bias in NHL Ice Hockey

Many people know about the favourite/longshot bias that occurs in most sports, but little is talked about the overs/unders bias. Is there in fact a bias at all? We shall investigate.

Using data going back to the 2005/06 season, we have calculated the returns should one have simply bet $100 on either of the over or under in the total goals for the match. Which one will come up better? The results are shown below with the first table being the over bets, the second being the unders:

Clearly, this shows that had one bet $100 on each bet on the over, you would have lost around 4.9% ROI from nearly 5,500 bets. Alternatively, had you bet $100 on the unders, you would have made over $5,000 from nearly 5,500 bets. The bias is definetly there, that unders betting is if anything, the way to go.

Is it possible that this difference is purely due to random variation? Quite possibly, however there is a reason why the bias is there. No-one wants to see a low scoring game, everyone likes plenty of goals and will generally bet with what they want to see. Also, punters will generally remember the high scoring games, the great goals of the year and the best goalscoring players. This in turn biases them towards betting the overs. The fact that in NHL, if the scores are level at end time, we have a playoff, then this I believe, adds more fuel to the fire to say that the bias is there. Because even if the scores are level, there still will be one more goal scored.

For those statistically inclined, the difference is in fact statistically significant. A simple one sample p test shows that the probability of such a result happening due to random variation is less than 1 in a thousand.

So the bias is there, and we have good reason to suggest why it occurs.

Equally interesting, is how the bias works amongst different goal lines. Shown below is the results should one have bet $100 on the overs and unders for different goal lines. Once again, the top table is the over bets, the bottom table is the unders (that makes sense!)

Most of the bets have a goal line of 5.5 or 6, but the overs for a run line of 5 and below made a nice profit. 736 bets were made for a 4% ROI. I have little doubt, that this is because of an over adjustment by the bookies to a smaller runline due to perhaps several previous low scoring games by the teams involved. Once again, this difference is statistically significant and not due to random variation. Similarly, betting unders with run lines of 7 or above resulted in a 16.7% ROI profit from 84 bets. And once again, this is statistically significant and not due to random variation.

Such over-exaggeration from recent matches has forced the lines to move, and this is where we as punters can pick up on and make several value bets. In this short article, we haven’t even studied the teams and we are already picking up value bets that are statistically significant. I see no reason why this trend should not continue.

What will be interesting, is how the overs and unders work with soccer, basketball and many other sports. Will there constantly be value in all these sports? Let me know with you comments below.