ESPN Reads Poker Faces With AI. The Math Outperforms the Stare. Microexpressions Were Always Overrated.
ESPN deployed an AI tells detection tool during its 2026 World Series of Poker Main Event broadcast in early July. The system displayed live metrics on player movements and a hand strength model charting probabilities of what cards a player might hold. Poker pro Gagliano reviewed every second of the live streams afterward, suggesting the tool's impact on competitive strategy is already being taken seriously.
This demonstrates the principle of observational inference at scale. The mechanism is probabilistic tell detection: combining movement tracking with hand strength modeling to surface patterns invisible to unaided observation. The lesson is that human intuition about deception is increasingly a subset of what statistical models can quantify. The house doesn't always win. The model does.
ESPN built and deployed the AI tells detection tool for its World Series of Poker broadcast. Poker player Gagliano, who started the final table in eighth chip position, reviewed the streams frame by frame.
Step 1: Record a short video of yourself describing something true and something false on your phone. Expected outcome: two clips where your physical mannerisms may differ between truthful and deceptive statements. Step 2: Watch both clips back at quarter speed and note any differences in eye movement, fidgeting, or pause patterns. Expected outcome: a rudimentary sense of your own unconscious tells. Step 3: Ask a friend to guess which clip was the lie, then compare their accuracy to your own slow-motion analysis. Expected outcome: a small demonstration of how systematic observation, even without AI, can outperform gut instinct.