The Factors: Wind's influence on completion percentage
In my previous article, I assigned a wind speed for each outdoor game in recent seasons based on the maximum gust speed I collected from a nearby ISD weather station during the approximate hours of the game. That seemed to produce more reasonable results than average wind speeds did — in particular in its successful identification of Week 11’s Giants vs. Chiefs game as the heaviest wind game so far in 2017. Next, I wanted to use those approximations to research the impact of wind on player performance.
There are several aspects of football that wind could affect, but passing accuracy seemed like the easiest place to start and likely the most important for fantasy purposes. And so, as I’ve done with many different factors in this column over the last few months, I ran some weighted average splits. I chose to split games into three categories: those with maximum winds below 5 mph, those with maximum winds between 5 and 10 mph, and those with maximum winds above 10 mph. That left me with a decent sample size of 5,736 weighted average pass attempts in the comparison of low-wind games to high-wind games, the smaller of the two comparisons I ran, and it also proved to have results that matched up with my intuition.
|Wind Speed’s Impact on Comp% and YPA|
|Wind Speed||Sample Size||Comp%||YPA|
Compared to a game with little wind, a game with moderate wind costs a quarterback 0.7 percent of his completion percentage and 0.13 yards per attempt. High wind in turn costs him 1.8 percent of his completion percentage and 0.30 yards per attempt. Scaled to a typical full game of 36 pass attempts, a quarterback can be expected to lose about a quarter of one completion and 4.5 yards with moderate wind and about two-thirds of one completion and 10.8 yards with high wind.
Neither touchdowns nor interceptions were markedly affected by moderate or high winds. That isn’t a huge shock because the magnitude of change in completions and yards is as small as it is, but I also hypothesize that inaccurate passes create far fewer interceptions than bad decisions do.
In testing the expected changes in completion, touchdown, and interception rates with moderate and high winds, I probably satisfied the major fantasy-specific question I can answer. It becomes tricky when you get too fine with this research because extreme weather is so uncommon. That said, I did want to push one level deeper to see if I could determine what specifically tends to happen to passes in higher winds.
Whenever a quarterback throws an incompletion, Pro Football Focus charters assign a reason for that incompletion. There are a bunch of categories that paint a sophisticated picture, including cases when the receiver catches the ball out of bounds, when a receiver drops the ball, or when a quarterback is hit as the throws. Those examples don’t speak to the passer’s accuracy per se, but there are four categories that definitely do: throws behind a receiver, throws in front of a receiver, overthrows, and underthrows. For my testing, I combined the first two of those into what I’m calling lateral misses and last two into vertical misses.
Because I don’t have the data to know which way each drive faces — teams change their orientation every quarter — I can’t know whether a vertical wind is in front of or behind a passer. However, because I did collect the direction of wind for each game and can determine the orientation of each field — thank you, Google Earth — I could split out the vertical and lateral wind speeds from the overall wind speed with a little trigonometry. Explained more simply, if a game features a headwind of 10 mph that blows directly from end zone to end zone, then it has a 10 mph vertical wind and no lateral wind. If instead a game has a 10 mph crosswind that goes directly from sideline to sideline, it has a 10 mph lateral wind and no vertical wind. Most games have wind somewhere in between those directions, and the cosine of the relative wind direction provides the lateral wind multiplier and the sine of the relative wind direction provides the vertical one. A 10 mph wind that is on a 45-degree angle from the end zone has an equal vertical and lateral component of about 7.1 mph.
I used that approach to calculate changes in vertical and lateral miss rates comparing games with low lateral winds to games with high lateral winds and games with low vertical winds to games with high vertical winds. For those tests, I split my data into just two wind groupings, those below 7.5 mph and those above 7.5 mph. I did that for sample-size purposes because I do not have incompletion description data back as far as I do weather data. I expected to see that lateral winds created lateral misses while vertical winds created vertical misses, but I was surprised to see that winds in either direction created more misses in front of or behind receivers than they did overthrows and underthrows.
|Vertical and Lateral Miss Rates with Vertical and Lateral Winds|
|Wind Dir||Sample Size||Vert Miss%||Lat Miss%|
Vertical winds do create a small uptick in vertical miss rate, so I’m fairly confident I have my data in order for accurate testing. I think this is just an unexpected result, but one that does mesh well with my golf analogy from last week. My personal sports experience suggests it’s much easier to accurately adjust for a wind directly into or behind you by throwing harder or softer. Lateral accuracy is a much more complicated puzzle, and one where atypical elements at least in my own experience more frequently impact one’s mechanics.