The final score of the Super Bowl will be 27-24. I’m not sure who’s going to win, but I am telling you, with absolute certainty, that the final score will be 27-24.
Lock of the century.
Gold star play.
Lead pipe lock.
How sure am I? So sure I’m willing to dub it my “platinum-plated penguin double rainbow grandma’s secret recipe intergalactic play of the year,” that’s how sure I am.
Why? Because Hayden Winks of Underdog Fantasy says so. Kind of.
See, each year he runs one of the more eye-watering columns of the entire football season — 99 Stats for the Super Bowl. I say “eye-watering” because it goes deep, and trying to make sense of it all is … difficult.
An example:
Chiefs LG Mike Caliendo has allowed a 7% pressure rate this year and will be partially tasked with stopping All Pro DT Jalen Carter (11% pressure rate) and star DT Milton Williams (14%).
And ..
The Chiefs’ offense has a +0.14 EPA per dropback with exactly 4 pass rushers. That drops to -0.06 versus a drop 8 (that’s 3 pass rushers) and +0.12 against 5+ pass rushers (that’s usually a blitz). The change up to use, if there is when, is the super soft pass rush plan on critical plays. The Eagles were 5th-lowest in blitz rate (25%) already.
I mean, what am I supposed to do with this? It’s good information, might be actionable, but … [insert shrugging shoulders emoji].
99 Stats For The Chiefs Eagles Super Bowl.
And what ChatGPT thinks my final score prediction will be after reading the 99 stats. https://t.co/LX1NFrTYoQ pic.twitter.com/o1J3ubJo7v
— Hayden Winks (@HaydenWinks) January 30, 2025
I’m guessing Winks had the same thought, as he ended his column this year a little bit differently: He plugged it into ChatGPT and asked it to spit out an answer. He tasked it with coming up with a final score based on all the 99 stats he fed it.
Eagles 27, Chiefs 24. That’s what it came up with.
I took it a step further, and plugged it all into both Google’s Gemini and Anthropic’s Claude.
Chiefs 27, Eagles 24. Both of them.
Spooky.
I then asked for boxscores for all three games, and let me just say this: Hollywood Brown is scoring a touchdown.
Now, obviously, there’s about 23,292 caveats here, but come on: Three different LLMs, and they all come up with the same score. And it just so happened that this score — either side — has the lowest odds at DraftKings. (Chiefs to win 27-24, +5000, Eagles to win 27-24, +5500).
Which means that somehow, some way, all three of these came up with what oddsmakers have as the most likely final score. And it’s all based on Winks’ exhaustive research, and nothing more.
If you’re not a little spooked out by this, you’re not paying attention.
To be clear: Winks took 99 stats, it was run through three large language models, and all three came back with the same score Vegas sees as the most likely outcome.
Gee, ya think this AI crap is getting pretty good?
For the record, I went back into all three and asked for some other outcomes, like low-scoring, high-scoring, turnover-laden, and blowouts. I then looked at those boxscores and threw together some Showdown lineups on DraftKings.
Listen: Our computer overlords are not here just yet, but I’m really starting to think these LLMs can be used to help make lineup decisions in DFS and betting decisions. Take as much information as you can gather, and let the AI do the heavy lifting. At worst, it’s another tool. At best, it’s … well, it’s a secret sauce not many others are taking advantage of, at least not yet.
And really — really — if the final score comes in 27-24, I’m going to find a way to upload my consciousness onto my ThinkPad. I mean, mind blown.
