Human vs. ChatGPT NFL Week 2 betting picks: AI looks to rebound after disastrous 0-5 debut
AI vs. Human in NFL Betting: One Week of Results
By Maine News Now
In a week‑long experiment that has attracted both curiosity and skepticism, independent sports bettor Dan — who has been following the National Football League closely for years — challenged the popular language model ChatGPT to a head‑to‑head betting contest. The rules were simple: each bettor would place five wagers on Week 1 games, using identical odds and a unit‑sized betting system. No pick could be priced worse than –120, ensuring that each wager carried a clear risk‑to‑reward profile. The experiment began on September 5, when the NFL opened its new season, and concluded on September 12, after the final Week 1 match.
The Experiment’s Structure
Dan and the AI each selected their five “best” bets from the same pool of spreads, totals and moneylines. The unit system meant that a –110 favorite required a stake of 1.10 units to win a single unit, while a +110 underdog required a stake of roughly 0.91 units to win the same. The objective was not only to win more money but to assess whether an artificial intelligence could consistently outperform a human bettor over a significant sample.
Week 1 Results
Dan’s record in the first week was 3‑2, yielding a net gain of 0.99 units. His wins came on a spread bet on the Chicago Bears, a total bet on the Raiders‑Dolphins matchup, and a spread bet on the Green Bay Packers. He also lost two of his five picks. In contrast, ChatGPT finished with a 0‑5 record, incurring a loss of 5.23 units. The AI’s five losses were spread across a range of games, and none of its picks hit the target.
“One week is obviously far too small a sample to draw any sweeping conclusions,” Dan wrote. “So, naturally, I’m going to draw one anyway: humanity is safe. At least for now.” While Dan’s win‑rate and unit gain were modest, the contrast with the AI’s performance was stark enough to spark discussion among bettors and AI enthusiasts alike.
Analysis of the AI’s Performance
ChatGPT reportedly spent nearly twenty minutes analyzing every eligible spread, total and moneyline before producing its card. The AI’s selections included bets on the Minnesota Vikings versus the Chicago Bears, the Pittsburgh Steelers versus the New England Patriots, the Jacksonville Jaguars versus the Denver Broncos, the Indianapolis Colts versus the Kansas City Chiefs, and the New York Giants versus the Los Angeles Rams. None of these picks matched Dan’s winning bets.
The experiment also highlighted a few inaccuracies in the AI’s output. For instance, ChatGPT incorrectly stated that the Denver Broncos had played an overtime game in Week 1, when in reality they lost to the Kansas City Chiefs 31‑10 in regulation. It also misidentified the injured player on the Pittsburgh Steelers, naming Marquise Brown instead of New England’s A.J. Brown. These errors illustrate that, while the AI can process large volumes of data quickly, it is not immune to factual mistakes.
The Broader Context
The experiment is part of a growing conversation about the role of artificial intelligence in sports betting. Some bettors fear that algorithms will eventually dominate the market, while others see AI as a tool to enhance human decision‑making. Dan’s approach—combining human intuition with a structured, data‑driven betting system—serves as a useful case study for both sides.
In Dan’s own words, “ChatGPT felt sorry for me because it proceeded to go 0‑5.” Although the statement was informal, it underscores the frustration that can arise when a machine’s output falls short of expectations. The experiment, however, did not produce evidence of an AI advantage, at least in a single week of NFL play.
Looking Ahead
Dan plans to continue the experiment for the remainder of the season, adding additional weeks to the data set. “We’ll keep going to see if any patterns emerge,” he said. The next round will involve Week 2, during which both Dan and ChatGPT will again select five bets. Dan’s next picks include a spread bet on the Washington Commanders, a total bet on the Los Angeles Chargers, and a spread on the Las Vegas Raiders. ChatGPT’s projected outcomes for Week 2 are still in development, but its previous record suggests it may need to adjust its strategy.
For now, the experiment remains an anecdotal but intriguing snapshot of the intersection between human expertise and machine intelligence in the high‑stakes world of NFL betting. While one week of data is insufficient to claim definitive superiority, the stark contrast between Dan’s modest gains and ChatGPT’s losses provides a concrete starting point for further inquiry. Whether the next season will tilt the balance is still unknown, but the conversation it has sparked will likely continue long after the final bell rings.