Human vs. AI NFL betting results show human still in profit

The current standings

The ongoing competition between a human bettor and ChatGPT has reached a point of high tension. As reported by Fox News, the human bettor is currently clinging to a slim profit of 0.09 units for the season. This represents a significant lead over the AI, which is currently down 4.03 units.

However, the momentum has shifted recently in favor of the machine. ChatGPT has outperformed the human for three straight weeks, including a Week 6 performance where it finished 2-2-1. While the AI lost 0.16 units that week, the human lost 1.17 units, making the robot the more efficient bettor in the short term.

The human's overall season record stands at 15-14-1, while the AI sits at 13-16-1. Despite the recent losses, the human remains the only one of the two with a positive balance. The human notes that the robot doesn't even have to make a profit to beat them anymore, as it simply needs to lose less money than the human does.

How the betting works

The competition follows a strict set of rules to ensure a fair comparison between human intuition and machine logic. Both participants receive five selections from the same odds board, with a maximum price of -120. This price cap ensures that neither side is taking on disproportionately high risks to skew the results.

To standardize the results, the stakes are adjusted so that every winning bet produces exactly one unit of profit. For example, a bet at -110 requires risking 1.10 units, while a bet at +110 requires risking 0.91 units. This system allows the audience to see the relative success of the picks without being distracted by varying dollar amounts.

The human bettor also utilizes a specific 'plug your nose and hit place bet' for a special weekly selection. This specific bet has a record of 4-1, yielding a profit of 2.90 units. This secondary metric helps track the human's ability to identify high-value opportunities that may differ from their standard five picks.

Week 6 analysis

The results from Week 6 highlighted different strengths and weaknesses for both the human and the AI. The human struggled with specific team performances, notably losing on the Los Angeles Chargers and the Dallas Cowboys. The human attributed the Chargers loss to the team's inability to perform, while the Bears-Falcons under failed due to late-game defensive rotations.

ChatGPT, on the other hand, focused heavily on totals during Week 6. The AI secured a win on the Buffalo Bills-Las Vegas Raiders over and a win on the Los Angeles Chargers-Kansas City Chiefs under. It also recorded a push on the Carolina Panthers-Philadelphia Eagles game, which resulted in zero units of profit.

The human's season report card shows they are more successful with totals, winning 5-4 for +0.66 units. In contrast, the AI has been more successful with totals as well, going 6-5-1 for +0.67 units. This suggests that while the AI is currently struggling with spreads and moneyline picks, it is performing competitively in predicting total scores.

Week 7 outlook

Looking ahead to Week 7, both participants have identified key matchups with varying levels of confidence. The human is looking to back the Tennessee Titans moneyline at +112, despite acknowledging it as a risky move following a previous loss on the team. The human argues that the Titans have shown more fight in recent weeks than the Cleveland Browns' record suggests.

The human and the AI actually find common ground on the Cincinnati Bengals -2½ at Baltimore. Both participants agree that the Bengals have a significant advantage coming off a bye with a healthier roster. They both believe Joe Burrow will be able to produce the necessary scoring to cover a three-point margin of victory.

The AI's other picks include the Pittsburgh Steelers -3 against the New Orleans Saints in Paris. The AI's logic is based on rushing DVOA, noting that Pittsburgh ranks second in rushing DVOA while New Orleans ranks last. The human, however, remains skeptical of the Saints' offense, though they are not ready to bet against them yet.

The role of AI in sports

This competition serves as a practical test of how well large language models can handle real-world, high-variance data. To generate its picks, the human used GPT-6.1 Sol with the thinking effort set to Ultra. The human provided the AI with the OddsChecker board and instructed it to use at least 10 different sources for each game to ensure a broad understanding of the facts.

The AI's ability to synthesize information from multiple sources allows it to identify trends that might be missed by a human. For instance, the AI correctly identified the quarterback advantage for Cincinnati and the rushing advantages for Pittsburgh. These data-driven insights are the core of the AI's strategy in this competition.

Despite these capabilities, the human's ability to account for