Project showcase

AI Poker Tells Detection

An AI system that watches poker players the way a great live pro does — blink rate, gaze, posture, stillness, smile symmetry — and flags the moments they are likely bluffing or likely holding it. Built by Luke Geel and featured on ESPN's 2026 World Series of Poker Main Event broadcast, in a five-minute NBC News interview segment, and in WIRED.

By Luke Geel - Boston, MA - Contact: geelluke@gmail.com

ESPN 2026 WSOP Main Event broadcast NBC News five-minute interview segment WIRED Sportico X & Instagram viral poker coverage
AI poker tells detection Poker bluff detection AI Poker tells Computer vision poker Poker AI WSOP Main Event ESPN poker broadcast

See It In Action

These clips show the system's actual output: annotated broadcast footage where the model surfaces a read on a player and the hand reveals whether it was right. No hidden information is used — the model sees only what the camera sees.

Demo: the model flags a likely bluff from behavioral signals before the hand is revealed.
Demo: annotated output overlay showing the signals the system is tracking during a hand.

What It Does

Every poker player leaks information. Most of it is too subtle, too fast, or too scattered across hours of footage for a human to catch. The AI poker tells detection system scans video of a player across many hands, learns which repeated physical patterns correlate with that player's hand strength, and then flags live moments where those patterns appear.

On ESPN's Main Event coverage the feature was used responsibly: analysis was shown only for players after they had been eliminated, so nothing on the broadcast could influence live play.

Blink rate

Changes in blink frequency under pressure, compared against the player's own baseline.

Eye gaze

Where a player looks — at the board, the chips, the opponent — and how that shifts by hand strength.

Posture & stillness

Shifts in posture and sudden changes in stillness after betting.

Smile symmetry

Asymmetric smiles read differently than genuine ones — one of many micro-expression cues.

How the model combines these signals — the architecture, the training data, and the code — is private. This page is a showcase of what the system produces, not a technical write-up. For the high-level ideas, read Can Poker Tells Be Measured With Computer Vision?

Honest limits: the system needs footage of a player to find their patterns, and some players simply don't have detectable ones. As Luke put it: "If I deployed it on a top professional like Daniel Negreanu, the system would probably find no signals. I assume he's worked hard to eliminate them."

As Featured On

Common Questions

What is AI poker tells detection?

A computer vision system that analyzes poker players on video — blink frequency, gaze, posture, stillness, and micro-expressions — and compares those patterns against previous hands to flag likely bluffs and likely strong hands.

Was it really on ESPN?

Yes. Omaha Productions featured the system on ESPN's 2026 WSOP Main Event coverage, applied only to players who had already been eliminated.

Can AI really detect poker bluffs?

Sometimes. It finds player-specific patterns when there is enough footage, and it is honest when there is no signal — some players, especially elite pros, don't have detectable tells.

Can I see the code or use the tool?

The code and methodology are private. For media, licensing, or collaboration inquiries, contact geelluke@gmail.com.

About The Creator

Luke Geel is an AI engineer for the US Air Force with degrees in AI, math, and economics. He built the AI poker tells detection system independently over six months. As an avid poker player he can often be found at the live tables in the Boston area, as well as occasional appearances on the Garage Game live stream. He also created Poker Stream Clipper and writes the AI Insider column for PokerOrg.