Andy Reid’s 4th Down Decisions

The latest from the sports data science guy at KC Sports Network, Joseph Hefner

There’s been a lot of angst amongst Chiefs fans this year about Andy Reid’s fourth down decisions. The narrative is that he’s been much more conservative this year than in previous years. Fourth downs are very visible moments that fans latch onto when creating narratives, so I wanted to take a look at the numbers and see if the narrative matches the facts.

So first, let’s take a look at what the options are on fourth downs. You can do one of three things. You can punt the ball, you can try to kick a field goal, or you can go for it. You know that win probability chart that the ESPN app has for all the games that shows how likely a team is to win at any given moment? That chart uses Win Probability (WP), which is a stat we have access to in public data.

The coolest thing about the Win Probability model is that we can input plays that DIDN’T happen, and it will pump out the change in win probability based on whether the fake play was successful or not. That means we can test every fourth down situation with all three options, and see how the WP changes for a punt, field goal, or going for it.

Here’s a concrete example. In the Cincy game this year, KC had the ball in the 4th quarter with 3:24 remaining. Tie game. 4th and 7 from Cincy’s 37 yard line. Chiefs current win probability was 33.3%. Here’s the WP change for all three options KC had:

Punt: -7.2% WP

FG: +0.9% WP

Go: +2.6 WP

The WP model factors in the chance of failure and a turnover on downs when it gives these percentages. That means the Chiefs would have been 1.7% more likely to win the game by going for it than by kicking a field goal, even with the chance of failure. That’s a pretty big swing for a single play. Butker ended up missing the 55 yard field goal, and the Chiefs offense never touched the ball again.

Does Andy Reid typically choose the more conservative option, or was this just a case where he’d seen Mahomes limp a little and chose to protect his QB over being aggressive?

This graphic shows how Andy Reid’s fourth down decisions have compared to the rest of the NFL. The black dashed line is the NFL average. Each dot is one team. This includes just plays where the increase in WP by going for it was more than 1.5% more than the next best option, so just 4th downs where the team definitely should have gone for it.

You can see that Andy was a bit above average for most of Alex Smiths years, and then increased his aggressiveness with Pat. In 2019, he was incredibly aggressive, and then regressed to just aggressive, but not aggressively so.

This year though? This year he’s consistently turtled up. He’s had 20 fourth down situations that were clear “go” situations, and has gone for it only 6 times. That 30% “go” rate is good for 3rd worst in the NFL this year. He’s been as bad this year as he was good in 2019.

And it’s not like he’s been failing in those situations. The Chiefs have converted on 5 of those 6 situations. Overall this season, Reid has gone for it 12 times, and converted 9 of those. That’s a 75% success rate for all 4th down decisions, and an 83% success rate when he’s gone for it in clear “go” situations.

So what gives? My theory coming into this was the mantra we’ve all probably heard quite often about Reid, especially this season. Andy Reid doesn’t call his good plays against bad teams. We’ve played quite a few bad and mid teams this year, so maybe he’s just trusting that his scheme is good enough to get the win.

In this chart, I’ve grouped all the times the Chiefs should have gone for it based on the pre-game win probability. So how likely were the Chiefs to win the game before the game started. It shows the percentage of times we went for it when we should have, and shows the total number of times we should have gone for it.

So first off, notice KC didn’t have any games with a starting win probability of less than 40%. That’s because we’re a very good team. In the 41-60% bracket, where the teams are reasonably close in talent level, we’ve gone for it 60% of the time (3 out of 5 opportunities). That seems promising.

In games with a 61-80% WP for the Chiefs, which are the games where we’re fairly large favorites (+3.5 to +10 point spread), we’ve gone for it just once out of 13 opportunities. This looks a bit like Andy makes his 4th down decisions prior to the game starting. Is it a good team? “I’ll consider going for it.” Is it a bad team? “Let’s just run the scheme and see what happens!”

Maybe that’s a bit unfair to him, though. Maybe he does update his plans based on in-game situations. Let’s take a look at that same chart, only for current in-game win probability rather than the pre-game win probability. The in-game WP takes the pre-game win probability into account, but also factors in the current score, time remaining, etc.

Here we see that although the spread of “go” opportunities is pretty evenly split, the Chiefs are far more likely to actually go for it in higher current WP situations. So when we’re not likely to win based on current WP, Reid doesn’t go for it, but when we are currently favored to win, he goes for it more than twice as often.

To sum all of this up, we’ve seen that this year is quite the outlier year in regards to Andy Reid’s aggressiveness. We’ve seen that he almost never goes for it in games where we are expected to win. And we’ve seen that he’s far more likely to go for it when we are already likely to win that game. Essentially, we go for it against good teams that we are beating.

So what do all these numbers mean? Honestly, your guess is probably as good as mine. Reid’s history shows that he’s more aggressive than average. This year is a massive outlier. Maybe that regresses toward aggression in the playoffs, especially since we should be playing good teams. We should at least see a bit more aggression because of that.

But if you’re looking for 2019 Andy who wasn’t scared of anything? Yeah, he’s probably not showing up this year. Hopefully we can still win the Super Bowl without him.