By Max Kaplan
So the bowl season has come and gone. My winter wishes came true in the form of an NHL collective bargaining agreement. While December may have been filled with “Best of” articles, there seems to be a lack of accountability in the world of sports statistics. For example, we are already seeing preliminary Top 25 rankings for NEXT season. No one evaluates how well their rankings performed last year because (a) it isn’t interesting or (b) they were not accurate. Probably both.
The purpose of this post is to look back at the college football bowl results to see if we may gain any insights on how to win (or at least gain the upper hand) in future bowl confidence pools. This will be the first of many articles that look retrospectively to get an edge in fantasy, pick-em, or pools.
Let’s get to it. (The numbers used here are calculated from the raw data from Yahoo Sports)
First and foremost, every set of bowl confidence picks from Princeton Sports Analytics performed in the top third in the nation. While it is subject to obvious sample-size caveats, it suggests that knowledge of football lends an immediate advantage. It is not completely random.
However, we all were successful in predicting the winner in only 20 to 24 games out of 35 total. The low variance suggests that there is randomness in choosing the winners.
However, the goal is not to pick every game correctly. It is to beat your peers.
In poker, you play the opponent, not the hand. Everyone has the same chances of winning a hand before the cards are dealt.
So too in bowl confidence pools should you play against the nation (against the aggregate provided by Yahoo), as opposed to trying to predict every game. This year, according to my calculations, the nationwide average was 20.7 of 35 games.
Damir Golac, winner of the Princeton Sports Analytics bowl confidence pool, only needed 22 correct to reach the 96th percentile. He chose the favorite in all but one game. This year, choosing the favorite in every game would get you to 23 of 35 (12 underdogs won), already much better than average.
In only two games did the nation choose the underdog: Rutgers over Virginia Tech and Kent State over Arkansas State. They were wrong both times. This leads to my first piece of advice.
Advice #1: Pick The Favorite (More Than You Think)
The reasoning is simple. 23 favorites won. 12 Underdogs won. Double your chances by picking a favorite. It sounds obvious, but the nation does not currently follow it to the extent they should. You may be inclined to make your picks different by choosing underdogs, but as long as you are in a smallish bowl pool (30ish people or fewer), your best bet is to just pick the favorite. To win larger pools, one must take risks by choosing underdogs and leave the outcome to chance.
Advice #2: Pick The Underdog Only When The Point Spread Is Close
I know I said not to choose the underdog. You should be at least 66% (23/35 is the opportunity cost) sure that they will win before choosing the underdog. This piece of advice is also intuitive. A 3.5-point underdog is more likely to win than a 14-point underdog (Louisville over Florida was an anomaly). In the 23 favorite wins, the average spread was -7.5. In the 12 underdog wins, the spread was +5.7. If you are going to choose an underdog, choose the close underdog AND be 66% sure.
Wisdom of Crowds
The wisdom of the crowd is the theory that the aggregate information collected by many people is more useful than the information gathered by a person on his own. To help people make their picks, Yahoo published the percent of users who chose either team.
If you put the consensus blowout as your most confident pick (ironically, this was Florida over Louisville) and put the most even choice as your least confident, you would end up with 432 points. This would be in the top 15 percent in the country. Average was 385.
As mentioned previously, the consensus picks included two underdogs. If you followed my first piece of advice and only chose favorites, you would score 445 (top decile).
In the same way that the wisdom of crowds assumes that some other force is smarter than you are, you may assume that the point spread predicts the result of the game. Point spreads have been remarkably accurate in the past, which is part of the reason why it is hard to make money by gambling.
If you pick the favorites with the same confidence as the point spread, you would score 446 (average was 385).
YOU COULD KNOW NOTHING ABOUT FOOTBALL AND STILL BE TOP 10% IN THE NATION BY FOLLOWING EITHER OF THESE 2 STRATEGIES.
If top 10% is not enough for you, further changes subject yourself to chance. However, you have a 40% advantage over everyone else.
Note: The other statistic that Yahoo provided was average confidence for either pick in each game. This was highly unreliable (compared to % picked and betting line) because people would be more confident than they should for their own team (self-serving bias) and in bigger games (availability heuristic). Even this biased statistic performed better than average (407).