AI Betting Experiment Sinks to 2-8 as Human Handicapper Stays Steady in NFL Picks Contest
AI Betting Experiment Sinks to 2-8 as Human Handicapper Stays Steady in NFL Picks Contest

Sports

AI Betting Experiment Sinks to 2-8 as Human Handicapper Stays Steady in NFL Picks Contest

An AI model tasked with picking NFL bets each week has fallen to a 2-8 record through two weeks, while a human competitor in the same experiment stands at 5-5.

An ongoing experiment pitting a human sports bettor against ChatGPT has produced a lopsided early result: the AI model is 2-8 on its picks through the first two weeks of the NFL season, while its human counterpart sits at a more respectable 5-5.

The project, run by an Outkick writer identified as Dan, has both competitors selecting five “best bets” each week using identical odds, with no wager priced worse than -120. Each bet is sized to win one unit. After going 0-5 in Week 1, ChatGPT improved slightly in Week 2, going 2-3, but both of its winning picks were under bets rather than spread or moneyline calls. On point spreads and moneylines specifically, the AI is now 0-7 through two weeks.

The human competitor’s Week 2 struggles were less severe. He correctly backed Cleveland to beat Tampa Bay outright and correctly predicted an under in the Panthers-Falcons game, which finished with 37 combined points. He acknowledged missing badly on the Packers and Chargers, however, and noted that Washington quarterback Jayden Daniels’ injury hurt a bet on the Commanders.

Through two weeks, the season report card shows Dan at 5-5 with a loss of 0.24 units, compared to GPT’s 2-8 record and a loss of 6.50 units.

The Week 3 Picks

For Week 3, Dan’s five selections included backing the Pittsburgh Steelers getting 3½ points at home against the Cincinnati Bengals, taking the New York Giants below a field-goal spread against Tennessee despite New York starting backup quarterback Jameis Winston after Jaxson Dart’s season-ending injury, and picking Tampa Bay as a home underdog against Minnesota. He also backed Las Vegas plus 3 points against New Orleans, citing the Raiders’ 53 points scored through two games despite missing tight end Brock Bowers, and leaned toward the under in Monday’s Eagles-Bears game, given Chicago’s likely absence of starting quarterback Caleb Williams.

ChatGPT’s five picks for Week 3 were Buffalo minus points against the Los Angeles Chargers, Jacksonville against New England, Cincinnati against Pittsburgh, New Orleans against Las Vegas, and the under in the Rams-Broncos Monday night game, projecting a final score of Rams 23, Broncos 20.

Notably, the two competitors landed directly opposite each other in the Bengals-Steelers game — Dan took Pittsburgh with the points, while ChatGPT backed Cincinnati to cover, projecting a 24-17 Bengals win.

The article’s author disclosed that the AI picks were generated using a system described as “GPT-6 Pro, powered by GPT-6 Astra,” with the reasoning effort set to “Max.” The model was given the odds board as it appeared Saturday, Sept. 26, at 6 p.m. ET, and was instructed to select five bets, explain its reasoning for each selection, justify why other eligible bets were passed over, and consult at least ten sources per game before fact-checking its own results. The process reportedly took nearly 30 minutes.

The weekly series is set to continue, with the next update expected to show whether the AI can record its first winning week of picks.

Click to comment

You must be logged in to post a comment Login

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

You May Also Like

Government Corruption

Updated 5/17/19 9:52am Jack Crane | Opinion  James Baker, Former-FBI General Counsel has joined Russian hoax media collaborator Michael Isikoff on his podcast, yesterday....

US Politics

I do not even know where to begin with this one.  Just when you think you have seen the worst that humanity has to...

US News

Education is considered to be one of the pillars of a successful life. Without a college degree, many believe these students will earn lower...

US News

ICYMI| If it were not for Tom Fitton and Judicial Watch, it is more than likely that the world would never know the extent...