Do Preseason Predictions Affect a Team’s Success Against the Spread?

When someone bets on sports, in many cases they are doing so because they feel their predictions about a certain game are more accurate then the Vegas bookies. They might feel that they know something others don’t, or that Vegas is underestimating or overestimating the impact of a particular event, or even that they have an overwhelming gut feeling. In many ways, this is an ambitious asumption. After all, what gives someone the confidence to believe that they know better than the aggregated opinion of thousands of bettors (which is what Vegas ultimately bases their lines on)?

If I had to briefly summarize the reason that I feel it is possible to beat Vegas, I put it as such: People have preconceived notions, and these notions are difficult to change, even in the face of contradictory information. There are a number of studies and papers on this phenomenon, some even indicating that people become more sure of their prejudices when presented with facts suggesting they are incorrect. One of the ways of gauging these preconceived notions is to look at bookies’ and experts’ preseason predictions for a team’s wins over the course of the year. Hypothesizing that bettors are not responsive enough in updating their opinions of how good or bad a particular team is, a team’s record when measured against their preseason wins prediction should have some bearing on the number of times they are able to beat the spread.

In order to test this hypothesis, I set out by getting an average of each teams predicted win total before the 2014-2015 season started. To find this average, I took the predictions from two betting sites (Bovada and Westgate Superbook) and one predictions/data analysis site (TeamRankings).

Next, I compared each team’s record to their projected number of wins at a quarter-way through the season, halfway through the season, three-quarters of the way through the season, and the end of the season. For example, if a team was projected to win 40 games on the season, and I was looking at them 3/4 of the way through the year, then they would be expected to have 30 wins at that point. I would then take their actual number of wins, and subtract the projected amount from this total. So if the team above had actually achieved 43 wins at this point, then they would be +3.0 at the 3/4 mark. Once I had the amount each team had over- or under- achieved their win totals by at these points, I ran a correlation with their WATS (Wins Against the Spread) at the same point to determine if there was a predictive effect.

Results: At every point during this past season, I found an extremely strong (at the .000 level) correlation between a teams performance against their predicted win total, and their number of WATS. The correlation at each point is as follows: 1/4 Season – .75, 1/2 Season – .82, 3/4 Season – .78, Full Season – .75.

Analysis: It seems that there is a very strong correlation between how much a team under- or over-performs their preseason predictions, and WATS. The results indicate that this correlation increased in strength up until the halfway point of the season, and then decreased in strength thereafter (but was still extremely strong, relatively speaking). To get more accurate information about this, ideally I would want to run a follow-up correlation on each section of the quarter-season individually, without each previous quarter’s data muddying things.

As for what sort of conclusions we can draw from this, I think the data supports my hypothesis that bettor’s habits do not react quickly enough to new information, and thus the spreads do not either. What this means is that teams that were expected to have a great season and then fail to deliver will have see spreads that do not give them enough of a handicap, and likewise, teams that outperform expectations will encounter spreads that give them too much of a handicap. To illustrate this point, you can have teams with a losing record (Utah Jazz, 38-44), yet that outperformed expectations (25.13 predicted wins), and therefore ended up with a winning record against the spread (44-39). Similarly, you can have a team with a winning record (Oklahoma City Thunder, 45-37), that did not live up to their expectations (56.7 predicted wins), and therefore ended up with a losing record against the spread (39-44). The Oklahoma City case is especially telling, as bettors were unable to accurately adjust their high hopes for the team even after reigning MVP Kevin Durant was lost for the season after 27 games.

The final (and ultimately most important) question to consider in this analysis is how to use this information to help us as bettors. The simple answer would seem to be that at the very least, a gambler can make a profit by figuring out which teams are over-performing against their preseason predictions in the first quarter of the season, and then betting on them in the second quarter of the season and beyond. Of course, there are all sorts of mitigating factors to take into account, such as the health of key players and strength of schedule. These factors are ones I will make a point to look at in refining this potential strategy down the road.

Leave a Reply

Your email address will not be published. Required fields are marked *