Josh Allen: Throwing velocity analysis

Not too long ago, we discussed throwing velocity on sideline passes. Understandably, there was some disappointment that no actual velocities were calculated. I promised to rectify that issue when All-22 angles were available. As was hopefully demonstrated before, the broadcast angles make it very difficult to determine the exact spot of the ball in many cases. Good news everyone! The All-22 is out and we can set the stage for accurate velocity determinations. Let’s select a Josh Allen play and check out the broadcast angle.

This isn’t his fastest pass, but it’s a decently quick toss over the middle. This play will set the stage for how this works and give a window into how we can trust the accuracy of these calculations. Let’s use some pauses and make some guesses on the broadcast angle.

On the first pause we have the release from Josh Allen. There’s a good degree of confidence of the yard line being at the 28. There’s less confidence on whether it’s over the hash or not. It appears to be near that, but it’s tough to tell exactly. To give an idea of how much error this can introduce, the hash marks alone are two feet wide. Basically “over the hash mark” is already kinda vague. When the ball hits Logan Thomas’s hands, we can safely say the ball is to the left of the other hash, but how far has pretty low reliability on this angle. The ball looks to be over the 48. Let’s introduce the All-22 feed.

The GIF shows the most important bits. Allen’s hand and the ball is actually about a foot to the right of the hash. At the point of impact, Logan Thomas is closer to the hash mark than I would have guessed based on the broadcast angle. The end result is a much better look at travel distance than any one angle would have provided.

For the other variable, time: The hardware/software combination I’ve been using was tested for reliability. Essentially, a game clock was recorded over several trials to determine if the selected frame rates represented their real time. There was a minor discrepancy and the data was rectified based on this. As a result of the above, the velocity of 49 mph should be very close to the truth.

What’s impressive isn’t the flat out speed, it’s more that Allen looks pretty casual about it. Most importantly, now that the methodology has been set and accuracy increased, we can look at throws as they come up. Next time a pass is coming in hot and you’re curious how fast it was, give me a shout and maybe you’ll get your answer.