In https://fantasyortheory.blogspot.com/2019/09/fantasy-theory-play-your-studs-or.html we looked at the performance of Wide Receivers vs the passing defense. How does this work for Quarterbacks?
Receivers have much more variation in their performance than quarterbacks, partially because they only get a limited number of catches. However, most top QBs throw 30-40 passes a game, for 200-300 yards, so there is more consistency - and more consistency generally means less "noise" in the data, and more correlation.
So I did the same exercise:
- I collected the gamelogs of the top 10 QBs who played 16 games in 2018
- I matched the QB yardage in each game to the season stats of the opposing defense
- I converted those 2 numbers to a percentage; ie if the defense allowed more yards over the seasons you see a higher percentage on the X-axis; if the QB threw more yards in a game compared to his season average, you see a higher value on the y-axis.
Let's look at the results:
That's a much more promising chart. The correlation for this data is .38 (note that R^2 and correlation is not the same number), which means there is statistically a clear positive correlation (for WRs it's only .13).
So how strong is this effect? If a QB plays against the top-5 or bottom-5, can we see a clear effect? First let's see how the passing defense is distributed:
Most teams allowed between 3400-4100 yards, with the worst and best teams between 10-25% more or less than the average. So let's see what we get if we focus on 3 groups:
- 90% or less ("bottom 5")
- between 90-110% ("average")
- over 110% ("top 5")
Now let's do something similar with their results, and we get something like this:
So roughly, a QB will score on average 15% more yards against a bottom defense, and 15% less against a top defense. If we are talking about a QB who would score on average 20 points, that's 3 points higher or lower. That's nice, but that would generally still not be a reason to bench him, unless you have a second QB who is in the same class. If your backup on average scores 3-5 points less, then most likely your A guy is still the best play.
Receivers have much more variation in their performance than quarterbacks, partially because they only get a limited number of catches. However, most top QBs throw 30-40 passes a game, for 200-300 yards, so there is more consistency - and more consistency generally means less "noise" in the data, and more correlation.
So I did the same exercise:
- I collected the gamelogs of the top 10 QBs who played 16 games in 2018
- I matched the QB yardage in each game to the season stats of the opposing defense
- I converted those 2 numbers to a percentage; ie if the defense allowed more yards over the seasons you see a higher percentage on the X-axis; if the QB threw more yards in a game compared to his season average, you see a higher value on the y-axis.
Let's look at the results:
That's a much more promising chart. The correlation for this data is .38 (note that R^2 and correlation is not the same number), which means there is statistically a clear positive correlation (for WRs it's only .13).
So how strong is this effect? If a QB plays against the top-5 or bottom-5, can we see a clear effect? First let's see how the passing defense is distributed:
Most teams allowed between 3400-4100 yards, with the worst and best teams between 10-25% more or less than the average. So let's see what we get if we focus on 3 groups:
- 90% or less ("bottom 5")
- between 90-110% ("average")
- over 110% ("top 5")
Now let's do something similar with their results, and we get something like this:
So roughly, a QB will score on average 15% more yards against a bottom defense, and 15% less against a top defense. If we are talking about a QB who would score on average 20 points, that's 3 points higher or lower. That's nice, but that would generally still not be a reason to bench him, unless you have a second QB who is in the same class. If your backup on average scores 3-5 points less, then most likely your A guy is still the best play.
Comments
Post a Comment