What are the most predictive football stats?
When analysing football statistics, two key terms stand out – predictability and predictiveness. These terms help us understand which stats can be reliably predicted from one season to the next and which stats can be used to predict a team's likelihood of
winning. By examining these aspects, we can identify the most valuable statistics for evaluating team performance and making predictions.
Predictability refers to how easily a statistic can be forecasted from one year to the next. It is measured by calculating the correlation between a team's performance in a specific
statistic in a particular season and its performance in the same statistic the following season. The higher the correlation, the more predictable the statistic is, making it a valuable tool for predicting future performance.
Predictiveness, on the other hand, refers to a statistic's ability to predict a team's chances of winning, which should be highly beneficial to fans checking out the vast range of
football betting markets available. This can be measured through same-season win predictiveness and next season's win predictiveness. The former examines the correlation
between a specific statistic and a team's winning percentage within the same season, while the latter analyses the correlation between the statistic and the team's winning
percentage in the subsequent season.
For instance, let's take a look at the statistic "turnover margin" over the past 10 seasons (2011-2020):
Year-to-Year Predictability: 11.26%
Same-Season Winning % Predictiveness: 69.67%
Next Year Winning % Predictiveness: 23.73%
The turnover margin shows a reasonably high correlation with the same-season winning percentage, indicating that it can be predictive within the same season. However, its predictability from one year to the next is relatively low, making it less valuable for future predictions.
Another example is "points per play differential":
Year-to-Year Predictability: 35.25%
Same-Season Winning % Predictiveness: 89.95%
Next Year Winning % Predictiveness: 38.39%
Points per play differential demonstrates a strong correlation with same-season winning percentage, as expected, since points are directly related to winning games. While it is somewhat predictable on a year-to-year basis, it falls short of being an ideal predictor for future winning percentages.
An interesting statistic to consider is "first down rate differential":
Year-to-Year Predictability: 48.42%
Same-Season Winning % Predictiveness: 54.65%
Next Year Winning % Predictiveness: 29.01%
First down rate differential stands out as the most predictable statistic from year to year. It effectively separates offensive and defensive performance, showing a significant gap between the predictiveness of offence and defence. However, despite its higher
predictability, it is not as effective as some other statistics in predicting future winning percentages.
Considering the impact of special teams, the addition of "special teams DVOA" to first down rate differential results in a more predictive statistic, surpassing any other statistic examined so far:
Year-to-Year Predictability: 41.36%
Same-Season Winning % Predictiveness: 87.54%
Next Year Winning % Predictiveness: 38.67%
This hybrid statistic with special teams DVOA factored in becomes a powerful tool for handicapping and predicting future winning percentages, outperforming the win/loss record
itself.
In summary, while no statistic can perfectly predict future performance in football due to various factors influencing each season, a combination of predictive statistics such as first
down rate differential and special teams DVOA provides a more reliable basis for making informed predictions and evaluating team performance beyond simply looking at win/loss
records. These insights can enhance the understanding of team dynamics and ultimately improve handicapping accuracy in football analysis.
Follow Us on Facebook – @LadunLiadi; Instagram – @LadunLiadi; Twitter – @LadunLiadi; Youtube – @LadunLiadiTV for updates
