Human NFL analyst extends lead over GPT with 9-6 record heading into Week 4 predictions
Human vs. AI NFL Picks: Week‑3 Recap Shows Human Edge, While Week‑4 Forecasts Diverge
By Maine News Now Staff
In the latest installment of a quirky NFL betting showdown, veteran prognosticator Dan Z. (who writes under the pen‑name “Dan”) out‑performed an artificial‑intelligence model powered by GPT‑6 Pro during Week 3. The human’s 4‑1 record and a modest gain of 3.06 units pushed his season ledger to 9‑6 overall and +2.82 units, while the AI posted its first winning week with a 3‑2 mark and a 0.89‑unit profit, bringing GPT’s cumulative record to 5‑10 and –5.61 units.
The rivalry’s most notable Week‑3 clash came in the Pittsburgh‑Cincinnati matchup. Dan had taken the Steelers against the spread, whereas GPT favored Cincinnati with the points. The Steelers not only covered but won the game outright, handing Dan a clear victory in that particular dispute.
Dan’s confidence after the week’s results is palpable: “I went 4‑1 in Week 3 and added 3.06 units, bringing my season record to 9‑6 and putting me in the black at +2.82 units.” GPT, meanwhile, echoed its own modest triumph: “GPT managed its first winning card, going 3‑2 for a profit of 0.89 units. It is now 5‑10 and down 5.61 units through three weeks.”
Both participants have now finished ahead of GPT in units every week, a streak that continues to shape the narrative of the human‑AI betting contest.
Week‑4 Preview: Two Sets of Picks
The stakes rise this weekend as both Dan and GPT submit their five best bets for Week 4, adhering to a common rule set: each selection must be priced no worse than –120, and every wager is sized to win one unit. Odds were locked in Saturday, Oct. 3 at 3 p.m. ET.
Dan’s Week‑4 Selections
Jacksonville at Cincinnati (1 p.m. ET) Dan’s rationale: “A week after betting against the Bengals (and winning), I'm going to flip the script and back Cincinnati. This game perfectly fits my NFL gambling philosophy: fade what you saw most recently. The Bengals lost to the Steelers and the Jaguars pounded the Patriots. Suddenly, everyone seems all‑in on Jacksonville, but I just don't see it. The Jags are OK, but I still think the Bengals are better and I'm laying less than a field goal for the home squad.”
Green Bay at Tampa Bay (1 p.m. ET) Dan’s view: “I know I said I was off the Bucs, like, five days ago. But I reserve the right to change my mind whenever I want… this bet is more about the ground game than the quarterback. Backing rookie Jalon Daniels in his first NFL start is a risk with Baker Mayfield out, but this bet is more about the ground game… Long, sustained drives will wear out the Green Bay defense… I think Tampa can put a few of those together to keep this game within a field goal.”
New England at Buffalo (1 p.m. ET) Dan’s take: “I'm not going to pretend that I'm not concerned about Drake Maye. He has looked legitimately BAD at times this season. But I don't think he's as bad as he's looked and he's due for a bounce‑back performance… In a division matchup where the Pats are a desperate team, I like catching a full touchdown.”
Miami at Minnesota (4:05 p.m. ET) Dan’s self‑deprecating note: “I might have to patent the ‘Dan Z.’s Plug Your Nose and Hit ‘Place Bet’ Bet of the Week… I don't think the Vikings are good enough to lay double digits against anyone, and that includes the abysmal Dolphins. Prove me right, Miami.”
Atlanta at New Orleans (Monday Night Football) Dan’s assessment: “I think people are overrating the Falcons after they rolled over the lowly Packers… I don't think people have come around yet on how good this Saints offense has looked. Michael Penix Jr. returned, but he honestly wasn't that impressive and missed multiple throws. Tyler Shough, on the other hand, has been incredibly impressive… I just think the Saints are better than the Falcons and they're barely favored at home. To me, this spread is simply incorrect and should be Saints –3. I'll happily lay 1½ instead, though.”
GPT’s Week‑4 Selections
GPT’s analysis, generated after a 30‑minute reasoning cycle and fed with “at least 10 different sources for each game,” produced the following five bets:
Arizona at New York (1 p.m. ET) GPT’s projection: “The Giants have gone two games without an offensive touchdown, and Jameis Winston has struggled… Losing Brian Burns to a torn ACL further weakens a pass rush… That gives Jacoby Brissett a favorable opportunity… a three‑point Arizona win is enough to cover.”
Jacksonville at Cincinnati (1 p.m. ET) GPT’s outlook: “Jacksonville’s early efficiency has been strong on both sides… Cincinnati’s protection and Joe Burrow’s ability to punish a weak pass rush remain serious concerns. Still, +125 requires an outright win less than half the time to break even. I prefer that return to Jacksonville +2½ at even money.”
Miami at Minnesota (4:05 p.m. ET) GPT’s view: “Miami’s struggling offense is now without De’Von Achane, while Minnesota has allowed only three offensive touchdowns through three games… The Vikings’ own offense has been inefficient… Minnesota can control this game defensively without producing a large total.”
Detroit at Carolina (Sunday Night Football) GPT’s reasoning: “Carolina placed both starting cornerbacks on injured reserve, creating a favorable matchup for Jared Goff and Detroit’s receivers… Detroit’s offensive line also improved… I trust its offense to exploit the depleted secondary and win by more than a field goal.”
Atlanta at New Orleans (Monday Night Football) GPT’s projection: “Atlanta gets extra rest after its Thursday win over Green Bay… At +115, Atlanta needs to win 46.5 % of settled decisions to break even, and I favor the Falcons’ matchup advantages at that payout.”
GPT also supplied projected final scores for each game, but the human commentator offered only brief reactions—ranging from “I didn't have much to say on this game… Good luck, GPT!” to “I hope this game goes under because a low‑scoring game should give my Dolphins +10 bet a better chance to cash. Good luck, GPT! (And this time I actually mean that.)”
The Contest Context
The betting duel, which first appeared on Outkick Sports, pits a seasoned football analyst against a state‑of‑the‑art language model. Both participants must select five wagers per week, each sized to win one unit, and may not choose a line priced worse than –120. The competition’s purpose is partly amusement, but it also serves as a case study in how human intuition, seasoned experience, and algorithmic analysis compare when confronting the inherent uncertainty of NFL outcomes.
For the moment, Dan retains the overall edge: +2.82 units to GPT’s –5.61 units. However, GPT’s Week‑3 success marks its first profitable week, suggesting the AI may be learning from the process.
The Week‑4 action kicks off this Sunday and will continue through Monday night. Fans of the odd‑sided betting narrative can follow both sets of picks on Outkick Sports, where updates and live results will be posted as they become available.