# Fumble luck, Part II

In my previous post, I explored a decade’s worth of NFL fumble recovery outcomes to find that, for the most part, a team’s fumble recovery percentages (FR %) are independent from one year to the next. Here, I’ll check how variations in FR % impact game outcomes, as measured by wins and losses.

## What happens when fumble luck regresses to the mean?

Under the assumptions that teams with low FR % are somewhat unlucky and that teams with high FR % are somewhat lucky, we can assume that, eventually, such luck will even out, and recovery percentages will regress towards an overall mean of 50%. If so, it also makes sense that teams with poor fumble luck in one season might do better, as judged by wins and losses, in the following season simply by recovering a higher percentage of fumbles.

To check this, I split up a team’s FR % into evenly spaced quintiles, and measured the difference between each team’s wins during that season with the number of wins recorded in the following season.  A positive win differential (Year 2 wins – Year 1 wins) will indicate that a team is improving. For 2013, I scaled each team’s current win total based on the fact that all teams have only played 14 games.

Since 2003, teams in the bottom FR % quintile have improved, on average, 0.30 wins, while teams in the top FR % quintile have lost, on average, half of a win. While that might not seem like an overwhelming effect, keep in mind each NFL team plays only 16 games.

Using a regression of win difference on FR % indicated that this association was marginally significant (p-value = 0.07). Below, the line of best fit is overlaid on a scatter plot of FR % in Year 1, and the Year 2 win difference.

The estimated slope of this line (-0.05) indicates that, on average, teams with higher FR % in Year 1 may have a more difficult time replicating their win total in Year 2, and that teams with a lower FR % in Year 1 can expect some sort of boost in Year 2. On average, it seems, wide differences in fumble luck alone could be responsible for upwards of a win or two between one season and the next.

## How about fumble luck and a team’s likelihood of covering the spread?

Thus far, we’ve suggested that fumble recovery percentages are mostly independent from one year to the next, but that changes in fumble luck might coincide with moderate changes to a team’s success.

To explore the in-season effect of fumble luck on a team’s performance against the Las Vegas spread, I extracted each team’s record against the spread (counting ties as half-wins) since 2003, using Goldsheet.com, and scaling 2013 wins to account for the fact that there are still two games remaining.

The plot below shows a team’s total ATS wins (jittered to show overlapping points) versus its overall FR %. The trend line (p-value < 0.01) does appear to have an upwards slope; as FR % rises, so too, it appears, does the number of times each team covered.

For those interested, the correlation coefficient between a team’s ATS wins and FR % was 0.14, which is likewise significantly different from 0 (p-value < 0.01). Alternatively, we could use these metrics to suggest that roughly 37% of a team’s variation in ATS wins can be accounted for using FR %. That’s a relatively strong impact, especially for something that’s mostly due to chance. Perhaps these results are something to keep in mind the next time you place a bet on a team led by Mike Glennon.

For those interested, there’s plenty of work left to do. Exploring offensive fumble types over time, or forced fumbles by defensive coordinator/head coach, might indicate some aspects of fumble recoveries which can be controlled. Further, a player-by-player analysis would be enlightening; for example, which running backs fumble most often, and which ones lose the most fumbles?

## Lastly what teams have been fumble-lucky or unlucky in 2013?

Dallas (70% recovery percentage) and Kansas City (63%) have been the luckiest, as judged by total fumble recovery percentages, while the Jets (31%) have been the unluckiest. In fact, the Jets have recovered just 2 of their 17 opponent fumbles. Expect a bounce-back from at least one of these teams in 2014.