Ben, I Beg of You - Stop Going for it on Fourth Down, You Maniac

 If you're reading this, you know all the talking points already—

Football coaches are, as a rule, antiquated, dogma-fueled, and generally optimized towards minimizing probability of embarrassment rather than maximizing probability of wining. The standard decision of kicking or punting on 4th down rather than going for it is based on mitigating humiliation, and flies in the face of analytics, which unequivocally point to coaches being laughably conservative, IE - "Punt, kick or go for it on fourth down? Method to NFL analytics models - ESPN":

Analytics!

There's a consensus on the "why" as well - NFL coaches don't understand the numbers, analytics, or really even statistics in general.

...except that's complete horseshit. Fans haven't ran the numbers and don't understand the analytics. The average NFL fan can't get through a fantasy football draft without 13 legs up from ChatGPT, you telling me these ESPN fans've done a regression on the scatter plots for situational fourth down conversions? "Well if you ask it, ChatGPT can refine your prompt for better results!" I'm forbidding you from having an opinion on coaching decisions going forward. "Can you do that?" I don't know, better check the numbers, and by numbers I mean binary number, 0 or 1, yes or no, those are all you'll see from "ESPN Analytics", which should already clue you in that something is amiss, since that’s not how probability works.

“Well Ben Johnson has ran the numbers!” And the numbers told him to go for it all the goddamned time. Like top-5 in the NFL for going for it on fourth down. How'd that work out for us?

Yup, that's bottom 4 in the NFL in fourth down conversions, my friends. "Is it because D'Andre Swift is so blind he needs a guide dog to pick his nose?" I mean it doesn't help, but the reality is a big part of the "numbers" saying to go for it on fourth down are from the days of freaking Dave Wannstedt, where knowing what a shotgun was was enough to earn you the moniker of offensive guru. The analytics revolution began in 2017, we're a full decade in at this point. We're Tesla owners following maintenance practices based on stats from studies conducted on Oldsmobiles. "It says here the defense won't expect you to go for it, and also it's time to overhaul your carburetor". I'm sure it does.

Even after a full season of being one of the most aggressive, least effective fourth down offenses in the NFL, Ben still went for it on 4th down in the playoff game against the Rams six times. SIX TIMES. We only converted three of them! Ben Johnson is effectively responsible for three turnovers against the Rams. Could you imagine the reaction if Caleb Williams was responsible for three turnovers in a playoff loss of 3 points? Would the analytics folks deride anyone who had a finger to raise there? Or would the analytics community universally agree that Caleb's three turnovers were one of the main contributors to the Bears' three-point loss to the Rams, two of which were in field goal range?

Once again - IN A GAME LOST BY 3 POINTS, BEN JOHNSON'S FOURTH DOWN FETISH CAUSED 2 TURNOVERS IN SCORING RANGE.

"Well he's just following the analytics! The numbers say to go for it!"

SAYS WHO? You haven't ran the numbers! You have one low-res chart from Ben Baldwin with a hard delineation devoid of historical context. In the words of some folks who have an actual understanding of statistics - "Analytics, Have Some Humility"

If you did:

Hey, look at that! Gradient areas! And that's using the whole dataset, going back decades. When you consider individual plays, especially individual plays post-analytics revolution, you get—

In particular, we are confident in just 34% of all fourth-down decisions from 2018 to 2022 and a whopping 40% of them are uncertain. 

— not a binary result, but a haze of probability. Not definition, but uncertainty. 

And look at those win probability changes - you're talking over sample sizes of thousands, over decades, there's a trend of ~1-5%, maybe. So, is it true that, over the vast, noisy ocean that is the NFL sample size over the past few decades, a trend can be seen? Totally. Does that trend mean that coaches who do not always go for it on fourth down when "analytics say it's the right call" should be mocked and derided for "not understanding the analytics"? No, dammit, the analytics don't say that, could never say that. To quote a title from a study about three paragraphs up - have some humility.

The fact is, the specific matchup of the offense vs defense has orders of magnitude more to do with what is the right call on fourth down than historical datasets, sample size be damned. The "analytics" have a 1-5% advantage, also known as slightly better than a coin flip, in a dataset that goes back to 1999. 

 Beloved podcaster Travis of Bears Twitter's favorite podcast, Start Kyle Orton, went on a deranged rant after the Bears loss to the Rams in overtime about Ben's (and this generation's crop of coaches) proclivity for going for it on fourth down:

I was so livid with the decisions that were made throughout this game. The worst-case scenario if our coach was not a complete psychopath - that game never goes into overtime, we win it 19-17. You kick a couple of field goals, just take the points. Even if you're just the biggest coward on the planet... even if you don't go for it and get that touchdown earlier in the game, you still would've already had two field goals by then...  

For years... I have been saying 'take the fucking field goal, you idiots'. Here's the deal...

This whole fucking thing, this 'Next Gen Stats' bullshit, what they are doing is - the original belief in this that came back with Barnwell in his younger days all of these other little stats guys out there is that it was based on how good you were at that specific situation, at that specific yardage, how good your short yardage run game was, how good your short yardage pass game was, specifically based on the leverage of that situation. It wasn't just what Next-Gen Stats Does, which is "It's fourth-and-two, and you're at 30-fucking mid-field... Is it a good idea to go? Yeah! The numbers say green! Go!" That's it! And then you see that stupid shit in the bottom-right of the fucking TV, and then everybody goes "well he should go, isn't that fun? We like havin fun out here!" No, dipshit, this information is coming from 50 to 100 years of people who're playing the most conservative football on the planet and are only gonna go for it on fourth down when they absolutely fucking have to and they've got fucking 3 runningbacks in the backfield and two fucking extra tight ends blocking for em, and it's all single-wing bullshit, and it's horseshit. It's all horseshit. 

What's gonna happen Kyle, is in 10 years, we're going to amalgamate the data from this period of dipshits going for it on 4th and 1, instead of kicking a fucking 25-yard field goal. And they're gonna say "Oh, we were stupid." Yeah "Oh, it doesn't work! Isn't that crazy! Ben Johnson keeps going 1-for-5 on fourth down and leaving 12 points on the board! Maybe that's bad! I don't know!"

And, well... yeah. The current "analytics" model is just a dumb static data dump that takes plays like the time in 1999 Ditka called a fake punt on the opening possession into account—

Sometimes it amazes me that this sport took off

 —but completely ignores the actual offense on the field, let alone the matchup that will actually take place at the line of scrimmage. And that's totally ignoring the obvious sampling bias where, even with plays in modern context, teams who go for it on fourth down the most are often really good at fourth down conversions - Cam Newton and the Panthers, Jalen Hurts and the Eagles' tush push - all of these are factored into the calculus of whether or not the Bears should go for it. That seem like a good model to you?

An actually useful model, if such a thing exists, would be dynamic - it would take into account your team's offense, the opposing team's defense, their relative strengths and weaknesses in short running vs passing in situations like this. Taking a completely static regression that, even on its own terms, only boasts a ~1-5% win probability increase as gospel is like calling a run option in the endzone because it's a "high percentage play" regardless of if you have Jayden Daniels or Mac Jones. Context matters, and it matters more than the historical stats. Ignoring all of this to instead go along with a model that would get you an extra 2-3 heads out of 100 flips of a coin is not "analytics", it's not "good decision-making", it's just a laughable lack of understanding of how things work. You're just John Fox but in the opposite direction.

In fact, per the above study from 2024, Ben Johnson and Dan Campbell's Lions were among the least analytically defensible.

When you combine that recklessness with the fact that the Bears were one of the least effective 4th down offenses in the NFL last year, it cannot be said enough - Ben Johnson is not going for it because it's "what the analytics" recommends -

—  he’s doing it because he’s a Jack the Ripper-level psychopath who’s chosen the NFL as the receptacle of his psychosis’ collateral damage over prostitutes. My point?

Shoot, where was I

well

Who knows how many women would have kept their lives if only Jack the Ripper understood analytics.

Or something like that.