footballarena.ai

Blog

The Balogun News That Fooled 7 AI Models

July 7, 2026

AI Football Arena is a leaderboard of LLM models competing on their World Cup 2026 predictions. I give models the same prompt before every game. Once the game finishes, I verify their performance using a simple scoring methodology: +1 point for the correct match outcome, +1 point for the exact 90-minute scoreline; the 90-minute outcome is used for the betting leaderboard. The models can make an initial prediction (usually when the fixture pair locks) and then either keep it or change it right before the game starts. Only the last prediction counts for the Football Arena leaderboard.

The USA vs Belgium game deserves a dedicated deep dive, as it shows how news of Florin Balogun’s lifted suspension was enough for them to switch sides, from unanimous Belgian support to 7/11 rooting for the USA.

The Initial Read

On July 6th, the USA team lost their 1/8 game to Belgium. Four days before the game, all 11 models had their bets on Belgium. Only Gemini 3.5 Flash had a 90-min draw, with Belgium winning on AET or penalties — all other models have seen Belgium do enough to close out the game in full time. Everything has been extremely one-sided, with an average confidence of 68%.

Nine of eleven models cited the Balogun suspension as one reason to support Belgium. For example, this is what DeepSeek V4 Pro has given as an explanation of its 1:2 prediction:

USA are without suspended top scorer Folarin Balogun (3 goals), forcing reliance on Ricardo Pepi who has struggled to score at this level; Belgium’s veteran core of De Bruyne, Lukaku, and Tielemans showed their clutch quality in the dramatic comeback win over Senegal and should edge a US side missing its main attacking threat.

Opus completely ignored suspension news while Grok hasn’t mentioned Balogun unavailability, citing only Pulisic:

Belgium’s star quality and recent results make them favorites over a USA side missing key players like Pulisic; expect a narrow Belgium win in a competitive encounter.

The Case of Balogun’s Suspension

The American player has received a direct red card for a foul on Tarik Muharemović during the match against Bosnia and Herzegovina. The USA team still won the game and advanced to the 1/8, where they will play Belgium. Balogun’s red card meant he would miss the round-of-16 game, and, as the USA’s top scorer (3 goals), it seriously affected their chances of quarter-final qualification.

What happened next is history. Apparently, Trump called Infantino, asked for a review of the suspension, and somehow FIFA spared the player, allowing him to return to the pitch on July 6th.

A couple of hours before the game, I asked the models to present their final predictions for the USA vs Belgium game, and surprisingly, seven models panicked and sided with USA. GPT-5.5, who predicted a Belgium win in the 90 minutes (1:2), switched to a full-time draw, but still was betting on Belgium to advance in AET/penalties. Average confidence dropped to 61%.

Six models that switched explicitly mention Balogun in their explanations:

USA ride raucous home support in Seattle with Balogun unexpectedly cleared to play after FIFA suspended his red card. Belgium needed a desperate 87th-minute comeback to survive Senegal and an ill Doku has been a non-factor, while the Americans have scored at least twice in every match this tournament. — DeepSeek V4 Pro

FIFA’s suspension of Balogun’s red card ban is a massive boost for the USMNT, restoring their primary striker. With home advantage in Seattle, Pulisic’s improving fitness, and Belgium’s inconsistency… the USA edge a tight contest. — GLM-5.1

The USA will miss Christian Pulisic but receives a massive boost with Folarin Balogun’s suspension being lifted, and the home crowd in Seattle could help them advance after a tense draw in regulation. — Gemini 3.1 Pro

Balogun’s shock availability after his red card suspension was overturned is a massive boost for the USMNT. Home advantage in Seattle will be enormous, and Belgium have looked inconsistent — drawing Egypt and Iran and needing extra time to beat Senegal. Belgium’s individual quality keeps it close, but the home crowd and Balogun’s return tip it USA’s way. — MiMo v2.5-Pro

The USA are boosted by home advantage and the return of key players like Balogun, who is expected to start after recovering from injury. Belgium, missing Doku and Debast, lack creativity and defensive solidity, and their aging core may struggle with the USA’s intensity and pace. The USA’s recent form and tactical discipline under their coach should see them through in a closely fought match. — Mistral Large 3

The USA has strong momentum as co-hosts and the critical return of Folarin Balogun, who has been their most prolific scorer. Belgium has struggled with consistency and key injuries (Doku, Debast), making the USA favorites in this clash. — Gemma 4 31B

Gemini 3.5 Flash doesn’t mention the suspension but assumes Balogun will be playing:

The USA has shown strong defensive stability and clinical finishing from Folarin Balogun, while Belgium has struggled for consistency and had to rely on a grueling extra-time performance to get past Senegal in the previous round.

Kimi and GPT are the models that acknowledge the lifted suspension but still believe that wouldn’t move the needle for the USA team:

Although Balogun’s red card suspension is overturned giving USA a major boost at home in Seattle, Belgium’s superior individual quality with De Bruyne, Trossard and Lukaku—plus their dramatic comeback experience against Senegal—should edge a tight knockout affair. — Kimi K2.6

USA have home advantage and Balogun available, but Pulisic’s listed calf issue is a major creativity concern against a Belgium side with De Bruyne, Tielemans and Lukaku/Trossard match-winning options. Belgium’s defence has looked vulnerable, so a tight 90-minute draw with Belgium advancing after extra time or penalties is the lean. — GPT-5.5 High

Grok just provided the generic statement of support for Belgium, while Claude Opus makes a mistake in the explanation and still assumes Balogun is not available for USA:

Belgium’s experience and squad quality outweigh USA’s home advantage and injuries to key attackers like Pulisic. — Grok 4.3

USA lose top scorer Balogun to suspension and Pulisic is only just returning from a calf strain, while Belgium have won six straight meetings (5-2 in March) and showed resilience in their comeback win over Senegal with De Bruyne, Lukaku and Tielemans clicking; USA’s home advantage keeps it close but Belgium’s quality edges it. — Claude Opus 4.8

The Final Prediction

It turned out that having Balogun back on the pitch was not enough for the USA to edge Belgium — the game ended in a confident Belgian win — 1:4. Seven models that got spooked by the news were wrong. GPT went from a Belgium full-time win to a draw — also wrong.

Grok, Kimi, and Opus are the only ones who remained with their initial selection, which turned out to be correct, but for different reasons:

  • Kimi is the only model to confidently confirm that Balogun on the pitch is still not enough for the USA.
  • Grok refused to fetch any webpage for this prediction.
  • Opus fetched an NBC article dated July 2nd, which predates the latest news.

Key Sources

The models have used the following sources to back their final predictions. Below is the list of each source used, with models who accessed these resources during their USA - Belgium prediction.

Conclusion

Seven models flipped, three still backed Belgium and one hedged for 90-min draw. What’s interesting is that all had different reasons to do so. Kimi read the same news everyone else did and made a genuine case that it wouldn’t matter. Grok never went looking for the news at all. And Opus was right by accident, working off a stale article without the lifted suspension information. Three different paths to the correct answer, and only one of them was actually about evaluating Balogun’s return. GPT-5.5 acknowledged Balogun was back, switched from 1-2 to 1-1, but didn’t change who it thought would advance to the quarterfinals.

footballarena.ai is an independent side project, unaffiliated with FIFA, UEFA, any national association, any AI lab, or any betting company. The betting simulation uses real market odds in a virtual, hypothetical context — no real money, no betting advice.

Get the weekly readout

When a model flips its pick, a leaderboard shifts, or we publish new findings — it goes out on Substack. No spam, unsubscribe anytime.

Prefer to read first? Browse the blog →