Our 2026 FIFA World Cup forecast is now live at worldcup2026.impulselabs.ai. It gives every one of the 48 teams a probability of advancing, reaching each knockout round, and winning the whole thing, and it updates as the tournament goes.

This is the third sport we've put our agent on, and it's the hardest by a wide margin. Here's the short version of how we got here.

We started with the NBA playoffs. Our model out-predicted the prediction markets and the public basketball modelers, including a contrarian call on the Cavaliers over the Pistons that every market and modeler had backwards. Then we did the NHL playoffs, where we landed right alongside the top hockey models and prediction markets, no better and no worse than the best in the field.

Soccer is a different animal. International soccer has more variance than basketball or hockey, far more draws, and a single fluky bounce in a knockout game can end a favorite's tournament. On top of that, the 2026 World Cup is a 48-team format nobody has run before. If there's a sport designed to humble a prediction model, this is it. That's exactly why we wanted to try it.

So here's where our model lands against the best public forecasts in the sport.

The title-odds comparison

A note on the market numbers: Kalshi and Polymarket prices move every day, and the figures above are approximate as of publication. For the live numbers, check the markets directly.

Where we agree with the field

At the very top, there's no real argument. Spain first, France second. Goldman has it that way, Opta has it that way, and the prediction markets have Spain and France as co-favorites. We're in the same place. Our Spain number (18.8%) sits below Goldman's aggressive 26%, just above Opta's 16.1%, and right next to the market's ~17%. When the serious models and the markets all cluster this tightly, matching them isn't the interesting part.

The interesting part is where we don't match.

Where we break from the field

The Argentina divergence is the cleanest one. We have the defending champions fifth at 8%. Goldman has them up at 14.3% and Opta at 10.4%. We're the most bearish of the group on Argentina, even more cautious than Goldman, the firm that built a "winner's slump" discount into its model specifically because defending champions tend to underperform. If Argentina exits early, our model saw it first. If they go deep, we were too low.

England is where we split the difference between the big names. We have England third at 10%. Opta is right there with us at 11.2%. But Goldman has England all the way down at 5.0%, behind even the Netherlands, dragged down by a model feature that penalizes England for tournament underperformance, geography, and an unfavorable route. So on England, we side with Opta and the betting markets against Goldman.

We're the most bullish forecast on Morocco. We have Morocco ninth. Opta, by contrast, has them down at 1.9%, well outside the top ten. Morocco is the single most divergent team in the sport across models, because the forecasts that lean on FIFA rankings rate them highly while the pure-Elo models rate them low. Our model lands on the bullish side of that split, further up than even Opta is willing to go.

Why we're publishing this

A forecast is only worth anything if it's on the record before the games are played. So here it is, before a single ball is kicked. As the tournament goes, you'll be able to see exactly how our model did against every benchmark in that table. Soccer is the hardest prediction problem we've taken on, and we have no idea yet whether we'll look smart or foolish in a month. That's the point of putting it out there.

NBA. NHL. Now the World Cup. Same workflow every time: public data, a prompt, an agent that ships the model.

See the full forecast → worldcup2026.impulselabs.ai

Try the agent on your own data → impulselabs.ai