Indianapolis Colts vs Washington Commanders prediction: 5 of 5 AI models back Indianapolis Colts

A one-score London game that the market — and one outlier model — misread

All five experts land on the same side: Indianapolis to win, but narrowly. The disagreement worth resolving is not who wins, but by how much, and here the panel splits into a 4-1 shape. Four models (Claude Opus, GPT-5.6, Kimi K3, Grok 4.6) converge on a field-goal Colts win in a game Washington stays live in until the end — projected scores of 23-27, 21-24, 24-27 and 20-24. Only GLM-5.3 projects a blowout (17-30), and its reasoning is notably weaker: it explicitly admits it cannot verify why the posted spread is ~15 and simply defers to the market rather than analyzing the game. The spread on this board is disconnected from the rest of the market and from every model's read.

How the game actually unfolds

The quarterback situation is the headline but not the story. Jayden Daniels is confirmed out (dislocated left elbow), and Marcus Mariota starts again — but the backup-cliff narrative is contradicted by evidence: Mariota has thrown seven touchdowns on 47 attempts this season and just engineered a 33-31 win over Seattle. As Kimi K3 notes, Mariota is stylistically similar to Daniels (mobile, RPO-capable), so Washington runs its full menu rather than a stripped-down playbook. The real concerns around Mariota are quieter: right guard Sam Cosmi is out, Rachaad White is out, and Terry McLaurin popped up Friday with a hamstring and is questionable — a late-week soft-tissue addition that historically misses at a high rate. That thins the margin for error around a veteran backup.

For Indianapolis, Daniel Jones remains a genuine limiting factor: roughly an 82.5 passer rating, three touchdowns against three picks, and responsibility for five of the Colts' six turnovers. The Colts' receiving corps is thin (Alec Pierce on IR, Keenan Allen downgraded late in the week), which pushes the plan toward Jonathan Taylor and tight end Tyler Warren. Here the experts split on the run game — Claude Opus sees Washington's 27th-ranked total defense giving Taylor a soft landing zone, while GPT-5.6 and Grok correctly note that Washington's run defense specifically is elite (3rd in rushing yards allowed at 82 per game, zero rushing touchdowns surrendered). The resolution: Taylor won't run wild, which forces Jones into the matchup that actually decides the game — a thin passing attack against Washington's 31st-ranked pass defense (280 yards per game) and 31st-ranked scoring defense (30.7 per game). Jones doesn't need to be good; he needs to be adequate against a secondary that has stopped nobody. That's why Indianapolis wins the majority of simulations without ever pulling away.