Search the Site

Will betting underdogs land you in the Loser Bowl?

I got an interesting email from blog reader William F. Barkley the other day.
I reproduce a condensed version of it below:

This time each year I find myself mired in a College Football Bowl Pool with five college friends (we’re all around your age) from Babson College. We try to pick the winner (against the spread) for each of the 25-30 bowl games. The picks for all games are made before the first game is played. There is no winner in this Pool, only one loser. The person who finishes last in the Pool hosts (pays for) dinner for all the other participants, their spouses, girlfriends, other college friends, neighbors, etc. The financial loss is not as devastating as the public ridicule. The loser also keeps a prized ceramic pump for the year as a sign of their handicapping futility.
So, in this Pool the strategy is simply not to finish last — finishing around .500 would suffice. Having been the loser last year, I thought I’d employ a little statistical strategy for this year. I took the past 7 years of picks and results to look for any trends (the Pool is now in its 13th year).

My thought going in was to simply pick the underdog in each game (being that there are no real home/away teams).

Here are some observations from the 7 years of data:

  1. Over the past 7 years, the underdog covered the spread 59% of the time (100 of 170 games)
  2. Across the 6 participants, the underdog is picked only 40% of the time (414 out of 1020)
  3. The overall win/loss percentage for all participants is 51% (520 out of 1020)
  4. One participant has never lost, one has lost 3 times, the others have each lost once. Interestingly, the participant who has never lost picks the least number of underdogs (34%).
  5. A participant picking all underdogs would never have lost over the 7-year period. A participant picking all favorites would have had the worst record (or tied for worst) 5 out of 7 years.

Armed with this small amount of information, I have, indeed, picked each of the underdogs in this year’s Bowl Pool of 30 games.
I would much rather donate my hard-earned funds to The Smile Train rather than pay for dinner and drinks for 20.

Is my strategy here flawed? Or, with no other outside information, is simply picking the underdog the best strategy to employ in order to avoid the worst record?


You have to admire William F. Barkley’s data-driven approach to the problem. So what do you think — did he pick a good strategy?

It is probably right that bowl underdogs cover the spread more than 50% of the time, but my guess is that the true likelihood is no more than 52%. If there are 25 bowl games and underdogs win 52%, then by betting all underdogs William will win 13 games in expectation, whereas someone picking all favorites will win 12 games.

The weakness in the underdog strategy is that anything that makes William look different from the other players works against him. Remember, he just wants to avoid the disaster scenario where he is worst of all — he doesn’t really care much about the average number of games he wins. Take the extreme case. If every other player picked all favorites, William would almost certainly want to pick all favorites himself. That guarantees him a tie. If he deviates at all, he has a pretty good chance he will end up being the worst because it is as if he is only up against one other player (since all the other players chose identically).

Even though the other players prefer favorites, they still are taking a healthy dose of underdogs. Given that fact, I don’t see a strategy that is obviously better than what William went with.

Unfortunately, though, good logic doesn’t always win the day in small samples. As of the time William sent his email, he had won only 3 of 11 games betting on the underdogs and was tied for last.

Looks like he might be hosting the Loser Bowl this year.