Early in the fourth quarter of last night’s game against the Rams, the 49ers faced fourth and goal just inside the L.A. two, while leading 14-9. ESPN’s Joe Buck pointed out that the analytics would advocate going for the touchdown.
49ers coach Kyle Shanahan opted to kick the field goal after taking (apparently on purpose) a five-yard delay-of-game penalty. Following the game, Shanahan provided reporters with a brief explanation of the decision.
“We were close, we discussed it,” Shanahan said. “If we were on the one, it might have been a little more tempting, but I thought our defense was playing well and stuff and felt, it wasn’t that tough.”
Here’s why it wasn’t that tough. His defense had been suffocating the Rams. So he took the three, went up by eight, and trusted the defense to keep doing what it had been doing all night.
And it worked. On the next drive, the 49ers generated a defensive touchdown after an interception of quarterback Matthew Stafford, and 17-9 became 24-9 became game over.
That’s why the percentages that routinely are espoused in such situations (X win percentage if you kick the field goal, Y win percentage if you go for it) must be considered in concert with a broad range of other factors that influence the final decision the coach must make. For example:
How do I feel about my available menu of plays for fourth and goal from just inside the two?
How have similar plays worked so far tonight?
Is there a play I’ve already used in a situation like this that I should actually use again?
Are there aspects of the opponent’s defense that are working differently than expected?
Are certain offensive players performing better or worse than expected?
Have any injuries impacted the expected effectiveness of the play that may be called?
Veteran defensive coach Dean Pees added more basic wrinkles to the broader analysis earlier this year.
“The computer told you that,” Pees said of decisions made during games. “When did the computer know what the weather was? Whether it was raining, whether the wind was blowing. Whether you were playing good on defense. OK, they say, ‘Well, it’s a two-point game. Should you go for it on fourth down?’ Well, I don’t know. Is the score 42-40 or 6-3? It makes a difference. . . . If they’re playing great defense, don’t. If they’re playing lousy defense, yes. The computer doesn’t tell you that.”
He’s right. Analytics takes football decisions and tries to make them quantitative and automatic and ideally devoid of discretion. They’re not, and they can never be. There will always be a qualitative element, no matter how loudly anyone claims that the “model” takes certain factors into account. At the end of the day, it remains a human being’s decision regarding the manner in which 11 human beings will, or won’t, successfully impose their will on 11 other human beings who have a vastly different agenda.
That’s not to say the quantitative aspect of the situation should be ignored. However, it never, ever should supplant the decision made by the coach, based on every relevant factor and honed by his full lifetime of football knowledge and experience.
For more thoughts on the right way to think about analytics, have a listen to the segment from PFT Live that accompanies this thing you’ve just read.