Why Brandon Staley was Wrong to Play Conservative
Staley had the opportunities to tilt win probability in his favor. Let’s look at the flaws in his decision making, why win probability is such an important tool, and how teams can influence it.
Brandon Staley made a name for himself as a rookie head coach by challenging the status quo. The analytics community fell in love with Staley because of his willingness to challenge “the way it’s always been done,” and embrace the “nerds” and their numbers.
The discussion around analytics tends to always drift back to 4th downs because those decisions are the most visible, most tangible, and feel like they have the most immediate impact. Staley has generated a lot of discussion specifically around 4th down decision making because he’s been so willing to go for it on 4th down when other coaches wouldn’t. His decisions align with the analytics that say coaches should go for it on 4th down way more often than they do.
Last night, Staley reverted a bit in his decision making and became far less aggressive than he was early in 2021. In multiple situations, Staley had multiple “go for it” suggestions based on the numbers by Next Gen Stats where he chose to punt.
After the game, Staley explained: "Just wanted to give our defense a chance to compete. I really loved the way we were playing. I felt like that was the formula to flip the field. I felt like we were aggressive when we needed to be tonight -- we converted all four of our fourth downs. I just felt like, with who is over there and the way our defense is playing, I felt like the field position would be a big edge for our defense to be able to put them back there. I like the way our defense competed tonight."
Staley’s decision was wrong in each instance and the logic he provided doesn’t justify any of the decisions. Let’s look at 3 reasons why Staley’s decisions to go for it were might have cost his team a key divisional win.
Probabilities falter if decision making is inconsistent.
In Blackjack, the standard methodology of playing is to “go by the book.” The book, in this sense, is the directional guidance of decision making based on probabilities of winning. For example, if you have 11 and the dealer is showing a 9, “the book” says you should double down every single time. The math behind it says that if you double down every time in this situation, you’ll win 52% of the time, push (tie) 7.3%, and lose 40%. What “the book” doesn’t say is that you will always win if you bet this way. Sometimes you will lose. But over a long enough timeline (with a large enough sample size, as the analysts would say), you will profit.
That’s the nature of probabilities. They don’t predict individual outcomes, but are predictive over a large enough sample size.
Now imagine the same situation – you have 11, dealer showing a 9 – and sometimes you double down, and sometimes you don’t. Now what’s your probability of winning? There’s no way to know because you created variance in the probabilities by not following the book. Instead maintaining your advantage over the casino, you made a decision that went against your favorability in the situation. If you have a 52% chance to win and a 7% chance to tie, why would you not take those odds?
Staley did this exact thing versus the Chiefs on Thursday night. The numbers favored him in three 4th down situations in the first half and he chose to go against the book in 2 of them. In those specific situations he may have converted on 4th down or he may not have. But he didn’t maximize his advantage over the Chiefs in those situations because he chose to punt. Not only did he negate his advantage in those situations, he provided an advantage to the Chiefs by punting.
Note: There are numerous models for Win Probability. Generally, they all take into account score, time of game, field position, down and distance, conversion rates across the league for those situations, and very importantly, they include the other team’s projected starting field position based on the decision made and potential outcomes. It’s not just about the immediate impact of the decision to go or not go, but also the downstream effects of that decision as well. Every win probability model I saw said that Staley should go for it and had it rated as a “go” or “strong go.”
Any way you look at the numbers, Staley created a disadvantage by not consistently following the book.
You should be more aggressive with a stronger defense, not more conservative.
Analytics detractors always like to argue that context is important. Staley fell into the same trap. He tried to contextualize the situation.
Staley argued that it was important to “flip the field.” This is a term conservative coaches utilize often to effectively say they were scared of getting scored on. Staley stated that he wanted to flip the field because he liked the way his defense was playing, the opposite of how he should have thought about it.
Staley’s logic is problematic because a better defense has a higher likelihood of stopping an opposing offense quickly. A poor defense needs a longer field to increase their chances of stopping an opposing offense. Simplified, it’s harder to score on a 99 yard drive than a 40 yard drive. It’s more important to play for field position when you need the extra buffer for your bad defense.
The win probability models factor in the likelihood of the opponent scoring based on field position if a 4th down conversion is failed (or a FG is missed). If Staley felt like his defense was playing well, he should have trusted them instead of coddled them.
For the most part, win probability models have a large enough sample size to include most context. If there is something specific that doesn’t occur in most football games – most notably weather or injuries to important players – that’s when a team should consider context the model can’t account for.
Staley made decisions like he was the favorite, not the underdog.
The Chargers came into the game as 4-point underdogs. As the (arguably) worse of the two teams, Staley needed to not only play the probabilities to maximize his chances of winning, but also find a way to flip the win probability in his team’s favor.
Theoretically, every team should make decisions by the book as it provides optimal decision for any given situation. One example of how well these models account for context, many win probability models include pregame odds in their calculations and de-weight it as the game goes on. Technically, when the Chiefs play, say, an Urban Meyer-led Jaguars team, the score starts out zero-zero and the win probability is 50%. In reality, we all know the Chiefs are more likely to win and models use pregame odds as an effective way to account for that.
It that situation, the Jags may start out the game at a 40% win probability. If they only go by the book, they may never break 50% win probability and have a real shot at winning the game.
According to data from nflfastR, the Chiefs started the game with about a 62% win probability. ESPN analytics had it at about 55% in favor of the Chiefs to start. That means that if the Chiefs just played the numbers, they should have a strong chance of winning. If the Chargers just played the numbers, they had a strong chance of losing.
The question then becomes, “How can an underdog increase their odds of winning?”
The answer: introduce variance.
The Chargers needed to do something to flip the numbers in their favor. They needed to introduce variance.
Variance is the result of anomaly events. In effect, variance is the result of big plays that swing the win probability quickly one way or another. This is why you always hear coaches talk about creating turnovers, hitting big plays, and avoiding the other team creating big plays. These plays create large swings in win probability.
As a favorite, maintaining the status quo means maintaining their advantage. An underdog should want to increase variance to break that status quo. You saw this play out last night when Jaylen Watson intercepted the ball on the goal line and returned it 99 yards for a pick six.
That play is the epitome of variance. It may be one of the largest swings in win probability we’ll see all season.
In 4th down decision making, going against the book and being successful is how a team can introduce positive variance. When the book says to punt, teams can greatly increase their odds of winning by going for it and converting. Obviously, if a team goes for it and fails, it creates variance in a negative way. But as an underdog, they were already likely to lose anyway so it stings a little less.
As a favorite, it’s still a bad decision to punt if the book says go for it. However, it’s more understandable as they’re avoiding the big negative variance play (failed conversion) by punting in those situations. Staley was making decisions to maintain the status quo when he should have been maximizing his decision making advantage and even introducing a little variance in an aggressive way.
Staley introduced negative variance for his team by not going for it. Staley gave the Chiefs gifts when he chose to punt and the numbers said go. While these decisions were small doses in negative win probability, they add up over a game.
Would it have mattered if Staley were more aggressive?
It’s impossible to say if the Chargers would have won had Staley made the right decisions. But according to Seth Walder and ESPN Analytics, the Chargers gave up 7.3 percentage points in win probability, in the first half. If a 4 point deficit in pregame odds resulted in a 5% deficit in win probability to start the game, per ESPN’s model, then we can surmise that those 7.3 percentage points would have been enough to win the Chargers the game.
Good story. Surprised you focused on the 1st half tho. You don’t once mentioned the lack of going for it on 4th down late in the 4th quarter when they were down 7 or the running plays they kept calling during those last few scoreless drives.
Good post. Two points: 1) A year ago Staley often said Herbert’s ability drove the logic to going for it on fourth down. Even if he was downplaying other factors, like the historical data that I guess is baked into the probability models the Chargers analytics staff would use, Herbert rewarded Staley’s faith at a pretty high rate. So the question that last night’s decisions raised is this one: if Herbert was worthy of Staley’s faith last year, why would he be worthy of less faith last night? Doesn’t make sense that he would. He had another season under his belt coming into this year and generally performed well. 2) you mentioned that injuries could lead a coach to go against the probability model. I’d imagine that the injury to Corey Linsley may have dissuaded him from going for it. However, he was still playing in the situations you mentioned. QUESTION: Was Reid’s decision to kick the FG from the 1 worse in terms of probability net outcome than Staley’s two punt decisions put together? I think I saw that ESPN said it was. I’d suggest that KC’s terrible performance in short yardage influenced Reid’s decision. They were like 1 for 5. The Chargers D had created a bunch of negative plus, at least four of them. And I think Geoff Schwartz’s video critiques of Chiefs RB point to reason for skepticism they’ll get that yard.