The Return of the Turnover Index
Make money betting on the NFL with this one simple trick!
For the past three NFL seasons, I have been publishing picks against the spread based on a simple betting strategy. That strategy, in a nutshell, is to bet on teams with a much lower defensive turnover total (season-to-date) than their opponent.
In the NFL, turnovers are largely random events. They correlate somewhat with team offense, but hardly correlate at all with team defense. However, the betting public has been slow to recognize this fact. Teams that have amassed strong records by "forcing" defensive turnovers may be overvalued by the market.
That's the theory, anyways. In practice, results have been modest. Over three seasons, my picks have gone 40-34-2 against the spread. That's enough to eke out a slight profit, but nothing spectacular. And it is somewhat below the 59% cover rate implied by my original analysis.
As a reminder though, my actual betting strategy is a bit more sophisticated than straight up picks against the spread. I have built a simple model that predicts the probability of covering the spread, based on the the difference between each team's per game turnover rate. Armed with a probability model, I manage a hypothetical bankroll in which wager size is dictated by the Kelly criterion. See my 2013 introductory post for more detail.
Early in the season, the original model was overconfident, leading to some bad (hypothetical) losses in the initial 2013 rollout of the methodology. I have since tweaked the methodology to dial back the confidence. In the first (and only) year of that revised methodology, results were encouraging, going 9-2 against the spread and generating a 9.4% return on investment. Here is how an initial investment of $1,000 would have fared if this wagering strategy had been in place since 1999:
A $1,000 bankroll in 1999 would have grown to $3,730 by 2014, for an annualized return of 9.2%. Of course, it is easy to find profitable betting strategies by mining the past, and countless have lost their shirts lost doing so. The only proper test of a betting strategy are the predictions it makes ahead of time. With that in mind, here are the week 6 picks (or pick, rather) against the spread (there were no week 5 picks):
The Broncos are undefeated, thanks in part to 14 defensive turnovers over five games. The Browns are 2-3, but may be better than their losing record would indicate, having been the beneficiary of just 5 turnovers on defense. Based on a per game differential of 1.8, our simple model predicts that the Browns have a 54.7% probability of covering the spread. At that probability, the Kelly criterion indicates we should bet 5.0% of our bankroll (i.e. $50).
For the past three NFL seasons, I have been publishing picks against the spread based on a simple betting strategy. That strategy, in a nutshell, is to bet on teams with a much lower defensive turnover total (season-to-date) than their opponent.
In the NFL, turnovers are largely random events. They correlate somewhat with team offense, but hardly correlate at all with team defense. However, the betting public has been slow to recognize this fact. Teams that have amassed strong records by "forcing" defensive turnovers may be overvalued by the market.
That's the theory, anyways. In practice, results have been modest. Over three seasons, my picks have gone 40-34-2 against the spread. That's enough to eke out a slight profit, but nothing spectacular. And it is somewhat below the 59% cover rate implied by my original analysis.
As a reminder though, my actual betting strategy is a bit more sophisticated than straight up picks against the spread. I have built a simple model that predicts the probability of covering the spread, based on the the difference between each team's per game turnover rate. Armed with a probability model, I manage a hypothetical bankroll in which wager size is dictated by the Kelly criterion. See my 2013 introductory post for more detail.
Early in the season, the original model was overconfident, leading to some bad (hypothetical) losses in the initial 2013 rollout of the methodology. I have since tweaked the methodology to dial back the confidence. In the first (and only) year of that revised methodology, results were encouraging, going 9-2 against the spread and generating a 9.4% return on investment. Here is how an initial investment of $1,000 would have fared if this wagering strategy had been in place since 1999:
A $1,000 bankroll in 1999 would have grown to $3,730 by 2014, for an annualized return of 9.2%. Of course, it is easy to find profitable betting strategies by mining the past, and countless have lost their shirts lost doing so. The only proper test of a betting strategy are the predictions it makes ahead of time. With that in mind, here are the week 6 picks (or pick, rather) against the spread (there were no week 5 picks):
| turnovers | |||||||
|---|---|---|---|---|---|---|---|
| game | away | home | diff | per game | pick | prob | bet |
| Broncos @ Browns | 14 | 5 | -9 | -1.8 | Browns | 54.7% | 5.0% |
The Broncos are undefeated, thanks in part to 14 defensive turnovers over five games. The Browns are 2-3, but may be better than their losing record would indicate, having been the beneficiary of just 5 turnovers on defense. Based on a per game differential of 1.8, our simple model predicts that the Browns have a 54.7% probability of covering the spread. At that probability, the Kelly criterion indicates we should bet 5.0% of our bankroll (i.e. $50).
Leave a Comment