## Extra Points

News and commentary from around the Web

# Next Gen Stats Introduction to Completion Probability

We've all seen the Next Gen Stats as part of NFL broadcasts -- player speed, distance traveled, etc. -- but this story on NFL.com is the first time I've seen anything really useful done with the data. For each pass thrown in the NFL, they've calculated the likelihood of a completion based on six factors:

• Air distance, the actual distance between the passer and receiver, both horizontally and vertically.
• Target separation, the distance between the receiver and the nearest defender in coverage.
• Sideline separation, the distance between the receiver and the sidelines.
• Pass rush separation, the distance between the passer and the nearest pass-rusher.
• Passer speed, the foot speed of the quarterback at the time of the pass.
• Time to throw, the time between the snap of the ball and the pass.

They have not listed full-season numbers yet (which would be pretty meaningless after two games anyway), but they have listed the three most unlikely completions of Week 1: Aaron Rodgers' 39-yard touchdown to Geronimo Allison against Chicago, Case Keenum's 4-yard touchdown to Demaryius Thomas against Seattle, and Tom Brady's 21-yard touchdown to Rob Gronkowski against Houston. It will be interesting to see the full-season offensive numbers -- and the defensive numbers, which theoretically would reveal which teams were "really" playing good or bad defense, and which were lucky or unlucky in terms of the quality of plays made by their opponents.

7 comments, Last at 03 Oct 2018, 5:15pm

### 1Re: Next Gen Stats Introduction to Completion Probability

by Dan // Sep 21, 2018 - 6:40pm

This won't give us straightforward information for dividing up responsibility for successes & failures, because many factors (players & scheme) influence both the completion probability, and the result after controlling for completion probability, and many players can influence both. Thinking about what it might shed some light on:

For quarterbacks this tells us something about arm talent vs. the rest of his skills. A QB with good arm talent (include accuracy, placement, velocity) will complete more passes than the probability numbers say he should, whereas most other skills (pocket presence, decision making, ...) mostly just influence the completion probability.

It seems like pass rush, protection, and offensive & defensive schemes mostly influence the completion probability and not the result. So this could provide a less noisy way of measuring those factors (although completion probability still blends them all together, along with various other factors).

For WRs & coverage defenders this should tell us something about ball skills. Ball skills mostly influence the actual result after controlling for completion probability. So we can learn (e.g.) about which CBs tend to have good position when the ball is thrown, and which tend to fail to get their head around and make plays even when they do have good position.

### 2Re: Next Gen Stats Introduction to Completion Probability

by Mountain Time ---- formerly Ninjalectual // Sep 21, 2018 - 10:03pm

Sounds promising!

### 3Re: Next Gen Stats Introduction to Completion Probability

by ChrisS // Sep 24, 2018 - 1:48pm

I am confused by the "both horizontally and vertically" in the criteria "Air distance, the actual distance between the passer and receiver, both horizontally and vertically". I agree it is easier to complete a pass thrown 10 yards vertically than a pass thrown 10 yards downfield but that seems to be due to the other factors mentioned. If all the other factors are the same is throwing the ball 21 yards straight downfield that much harder than throwing it 15 yards downfield and 15 yards towards the sideline (also 21 air yards)?

### 4Re: Next Gen Stats Introduction to Completion Probability

by Dan // Sep 25, 2018 - 3:02pm

I assume they look at them separately. An extra yard of horizontal distance makes the pass harder to complete, though not by as much as an extra yard of vertical distance.

### 5Re: Next Gen Stats Introduction to Completion Probability

by Vincent Verhei // Sep 27, 2018 - 4:49pm

The key is that they are measuring the actual real-life distance between the quarterback and the receiver. All we have had to this point is the distance from the target to the line of scrimmage. A 5-yard pass over the middle was treated equally to a 5-yard pass to the sideline, even though the ball has to travel much farther in the latter (assuming the quarterback stays in the pocket). If you click the link for this, they have a chart showing this is a significant factor.

### 6Re: Next Gen Stats Introduction to Completion Probability

by ChrisS // Oct 03, 2018 - 11:52am

Yeah I agree that total air yards is most important for completions. I just did not see how breaking it down by horizontal and vertical would matter, which is how I interpreted the last clause in the first bullet "Air distance, the actual distance between the passer and receiver, both horizontally and vertically."