Do Google Searches Have any Impact on a Team’s Record Against the Spread?

If there is any edge to be gained in the world of betting on NBA games, there has to be a population of bettors on which we are actually gaining this edge. I’m going to go ahead and generalize by saying that people make “bad” bets for one of two reasons: One, they are basing their bets on incorrect or outdated information (I touched on this a little bit in a previous post), and two, they are basing their bets on emotional reasons rather than data-driven ones. Since I have already partially looked at the former scenario, I am going to base this post on the latter.

Imagine, if you will, a die-hard fan blindly betting on their team to beat the spread, considering only their loyalty when deciding to place a wager (I’m reminded of a story about old Italian couples who would supplement their monthly incomes with bets on Rocky Marciano, although in that case their strategy seemed to work out, as Marciano ended his career undefeated): This is exactly the type of bettor who we want to bet against, as if they are great enough in number, their action will skew the odds favorably towards the opponent of their beloved team. Once we have determined that it might be advantageous to bet against teams with a large and rabid support base, the hard part becomes quantifying a team’s fan total. I actually came across an interesting blog post that looks to this question for Major League Baseball, although both data sources the author looked at had a number of holes in them (I discovered a similar problem during my own research).

The easiest way to quantify a team’s fan total would be to simply look at attendances across the leagues, although this clearly does not tell the full story, as teams always have fans in other parts of the country or even the world who cannot attend games. Another option might be to look at away attendance, although I would tend to think that this would be skewed towards teams who are performing well in addition to teams with a large fan base. My next thought was to look at merchandise sales, but unfortunately this is a very difficult number to get a hold of, as there are countless unofficial and bootleg merchandisers to account for.

Facebook likes would seem to be a good metric to use in this case, but the data is too polarized to be usable, with teams such as the Lakers having upwards of 27 million likes, and teams like the Jazz struggling to garner even a million.  I finally concluded that I would use Google Trends as a measuring stick for fan bases, which measures how many search queries have been performed for a given term in relation to the total number of queries for that period. I realize this is not the most exact metric to use, as it more measures how salient a team is in the public conscious than how many fans it has. However, it was the best I felt I could find for the sake of this analysis. As a result, I went into this analysis with the hypothesis that a team’s average Google Trends number would be negatively correlated with the number of wins against the spread (WATS) they achieved in the 2014-2015 NBA season.

 Results: I ended up taking each team’s average Google Trends number from November 2014-March 2015 (roughly the length of the NBA’s regular season), and running a regression to see if these numbers had a positive correlation with the number of times each team succeeded in beating the spread. Since I had previously observed that outperforming (or underperforming) preseason predictions had a significant impact on a team’s success against the spread, I decided to make this a multiple regression include Wins Over/Under Preseason Prediction as a mitigating factor. In the end, I found that there was not a significant effect of a team’s average Google Trends number on the number of WATS they had in 2014-15: Average Google Trends Number had a coefficient of -.04, which was significant at the .353 level. In other words, each increase in a team’s Google Trends number was correlated with a .04 decrease in the number of WATS the achieved, and there was only a 64.7% chance that this was not due to chance.

Analysis: It’s tough to say whether there was simply no correlation to be found here, or whether my metric for measuring each team’s fan base was inaccurate. On top of that, there may be differences within each team’s fan bases that identify as bettors, or that bet solely based on their fandom. In any case, I would definitely be interested to revisit this topic in the future if I feel that I have a more accurate metrics at my disposal.

 

Leave a Reply

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