## Washington Post: What Analytics Can and Can't Do in Football

This week's article at the Post takes advantage of the Redskins bye week to take a step back and discuss what analytics can and can't do in football.

...We can look at which teams succeed more often in various situations and rank them. Unfortunately, that’s almost all we can do with this type of approach. It’s great fodder for Internet message board arguments, but it doesn’t help teams gain an edge.

To really help coaches, a deep understanding of how teams come to win games is required. Advanced models such as Win Probability and Expected Points are very powerful tools, because of what an economist would call “linear utility functions.” The term linear refers to the quality of proportional desirability. For example, it’s exactly twice as good to have a 40% chance of winning a game as a 20% chance of winning a game. You can’t say that about total yards or simple points...

### 4 Responses to “Washington Post: What Analytics Can and Can't Do in Football”

1. Anonymous says:

Interesting article, but the conclusion and the importance of the analysis is way way overblown.

Exactly how do you generate these ~2 more wins per year? Take a team, this season, and tell us exactly what would change.

In performing such an exercise with careful error propagation, it would show just how inapplicable this analysis is.

If you take a team, cut its running back, pass 75% of the time, do an onside kick once a game and go for 4th downs for less than 5 yards, it seems highly doubtful that one could prove it would result in any improvement. Especially after about 2 games, when teams adjust to the passing, the 4th downs, and the onside kicks.

2. James says:

How does a team adjust for 4th downs?

3. Nate says:

> If you take a team, cut its running back, pass 75% of the
> time...Especially after about 2 games, when teams adjust
> to the passing, the 4th downs, and the onside kicks.

I think that run/pass analysis should really be more sophisticated than the "run or pass" we usually see before it's used to inform general play calling. Not only are there are teams that seem to do better with running than passing, but there's a variety of different styles of both passing and running plays that should have different properties. Moreover deceptive plays like draws and play action passes can contribute to EPA/WPA in significant ways that numerical methods may miss.

Some coaches are better than others, but there's definitely some low-hanging fruit on 4th down decision making. I think actual returns would be pretty close to expected ones here, since defenses can't do much to adjust.

According to the zeus guys, the difference in play calling returns between the best and worst coaches in the league seem to be around 1 expected game per season, and most of that is in 4th down decisions.

Surprise on-side kicks really only make sense when the game is close, or the kicking team is behind, so 'once a game' is probably not the right way to go but we probably ought to see more of them. Regarding the adjustment, I wonder if receiving teams that expect onside kicks have worse starting field position.

4. Jared says:

Really liked the post Brian, especially the point about how the usefulness of analytic techniques will decrease as more teams begin to use them. There was only one point in the whole thing that I disagreed with. It is possible to mathematically separate a player's value from his teammates. It only works for certain positions, and the techniques for doing it are not well known and even less well understood. Never-the-less, one can do it if they have the time, the data, and the energy. :)

I look forward to reading more great posts.

### Leave a Reply

Note: Only a member of this blog may post a comment.