## The 4th Down Bot Returns

The 4th Down Bot is returning to the New York Times this season. You might recall we booted him up late last season, but this year he'll be around starting week one. At its heart, the bot is a fun application of the 4th Down Calculator feature here at AFA. It uses both the Expected Points model and the Win Probability model to estimate the best option for every 4th down as a game is in progress.

As I mentioned last year, although the 4th down issue is growing mold with smarter fans, it remains the lowest hanging fruit on the football analytics tree. So it's nice to be able to automate things and not have to do the analysis myself. But on the other hand, we can add 'football analyst' to the list of jobs being taken over by robots.

The Bot will be faster, more accurate, and come with some new features this season. Here is a brief introduction. Here's are a few notes on how it works. And here is his Twitter feed.

### 6 Responses to “The 4th Down Bot Returns”

1. David Kravitz says:

Brian, on the 4th and goal from the 3 you show a punt as +0.2 points. How did the bot come up with that?

2. Brian Burke says:

Because that's the "expected" value of a punt anywhere inside the 40 or so. Basically it's probably going to be a touch back. I didn't bother trying to model silly situations like that with any more fidelity. Waste of time.

3. Anonymous says:

It'd be interesting to see how many of these 4th decisions actually occur in a season, and how often they actually make a difference. The article shows as an example with a score of 34-7, so it truly didn't matter.

i'd also like to see discussion of changes not only in the mean WP, but also the stdev of these decisions. Increasing your WP by 0.2 games per season in the limit, doesn't mean anything if it turns out you actually lost a game because of the decision to go for it - you get fired.

4. Anonymous says:

one other point on these 4th down decision (especially in the redzone, where the choice is high percentage FG or TD attempt).

The 32 teams in the nfl have red zone (td) percentages that range from 44% up to 73%. With such a huge variance, does it even make sense to have a league wide 10 year estimate and conclude that all teams must go for it in a particular situation?

5. Brian Burke says:

Those are observed variances, not true probabilities. If 32 people flip a penny 10 times, some will get 3 heads, some 8, most around 5. But the true probability of the next flip is .5.

6. Tim says:

Brian -- that is true, but surely there is something consistent, at least for the subset of very strong offenses we all know and love (Denver, New Orleans, New England, Green Bay with Rodgers, ...).

I don't have 2013 data, but comparing 2011 and 2012 shows that red zone TD percentage across the 32 teams had an r=0.47 correlation between those two years. I doubt there is much to do about this in the model, though -- we probably don't get a very good estimate of strength of offense until too far into the season, given the small data.