WSOP Just Ran the Numbers 1 Million Times, You Won’t Believe Who The AI Says Wins Poker’s $10M Prize

Last Updated on July 31, 2026 by Bala Kumar

Every summer, poker fans play the same guessing game once the WSOP Main Event final table is set: who actually wins this thing? Usually that debate stays in the realm of gut feeling and Twitter hot takes. This year, someone decided to answer it with math — a lot of math. One million simulated final tables, to be exact.

The project comes from Advanced Poker Training, a poker software company run by Florida-based developer and longtime poker author Steve Blay. And if his name rings a bell, there’s a reason for that: this isn’t his first rodeo.

The Guy Who Called Qui Nguyen a Decade Ago

Back in 2016, the WSOP Main Event final table also had a break between the last nine being set and the champion being crowned, the same format we’re seeing again this year. Blay ran a model back then too, though a far smaller one, using just 100 simulations. The output pointed to Qui Nguyen as the most likely winner, which raised some eyebrows at the time since Nguyen wasn’t the chip leader heading into the final table. Nguyen went on to win it all.

Ten years and several generations of computing power later, Blay’s system is a different beast entirely. This year’s model doesn’t just crunch numbers off a spreadsheet — it builds a distinct computerized “personality” for each of the nine finalists based on their tournament history, playing style, risk tolerance, and known tendencies at the table, then plays out the final table under realistic conditions: actual seat draw, actual starting stacks, actual blind structure. The system can reportedly simulate a full nine-handed final table roughly three times per second, which is how it managed to grind through 1,000,000 complete tournaments between July 24 and July 26.

That’s 10,000 times the sample size of the 2016 prediction. So, what did all that computing power spit out?

The Full Win Probability Breakdown

Chip leader Lucas Jumalon, who takes 194,000,000 chips and roughly 35% of the total chips in play into the final table, was the runaway favorite — winning the tournament outright in 40.4% of all simulated runs. Nobody else came close to that number, though the gap between second and eighth place was tighter than you might expect.

Here’s how all nine finalists finished across the full run of simulations:

Even with more than a third of the chips in play, Jumalon still busted in ninth place in roughly 1% of the simulated tournaments — a good reminder that tournament poker punishes overconfidence no matter how big your stack is.

What ICM Reveals That Chip Counts Don’t

Raw win probability only tells part of the story. Advanced Poker Training also compared each player’s simulated average winnings against their Independent Chip Model (ICM) value — essentially, the theoretical dollar value of their stack based on the remaining payout structure. Beating your ICM number means you’re outperforming what your chip stack alone would predict; falling short means the opposite.

Two names stood out immediately: Jamie Shaevel and Greg Mueller both finished well ahead of their ICM expectation, with Mueller’s average simulated winnings landing 12.33% above what his stack alone would suggest — the single biggest positive gap at the table. Mueller enters the final table with three WSOP bracelets and decades of high-level tournament reps, and the model clearly weighted that experience heavily. He’s also a former professional hockey player, which — for whatever it’s worth in a simulation — apparently correlates with performing well under pressure.

On the flip side, Rami Hammoud posted the model’s biggest underperformance, coming in 4.73% below his ICM value despite sitting in second place on the actual chip counts. Part of that comes down to something no player can control: seat draw. Hammoud is seated just two spots to Jumalon’s left, meaning every time he’s in position to attack the blinds from the cutoff or button, the chip leader is sitting there with maximum leverage to punish him. It’s a small structural disadvantage, but across a million simulations, it clearly added up.

Unsurprisingly, the five shortest stacks at the table — Gagliano, Boos, Saaskilahti, Feng, and Evagorou — all underperformed their ICM value to some degree, a pattern that lines up with what most tournament regulars already know intuitively: short stacks at a final table this deep are stuck walking a tightrope between preserving their tournament life and accumulating enough chips to survive rising blinds.

So Who Does the Model Actually Pick?

Here’s where it gets interesting. By the raw numbers, Lucas Jumalon is the obvious pick — more chips, higher win rate, no real argument against him statistically. But Blay didn’t stop there. When it came time to put his name on an official prediction, he went against the favorite and backed Greg Mueller instead, pointing to Mueller’s consistent ICM outperformance and championship-level composure as reasons to believe the veteran finds a way to outplay his stack size when it matters most.

It’s worth being upfront about what this tool actually is: it’s not a large language model or generative AI in the ChatGPT sense — it’s a purpose-built poker simulation engine that models player behavior and runs a massive number of tournament outcomes. But the underlying idea — using data and behavioral modeling to move beyond gut-feel predictions — is exactly the kind of thing that’s increasingly showing up across sports and esports forecasting, and poker is a natural fit given how well-defined the mathematics of tournament equity already are.

Should You Trust a Simulation Over Your Gut?

Probably somewhere in between. A million simulations is a genuinely large sample size, and the model correctly flagged an underdog winner back in 2016 with a fraction of the computing power it has now. But no simulation knows what cards will actually come out, how a player will feel with $10,000,000 on the line under television lights, or whether someone changes their entire strategy heading into the biggest week of their career. What these numbers do offer is a much more precise picture of where the real edges sit — which stacks are dangerously exposed, which players are quietly better positioned than their chip count suggests, and just how much of tournament poker’s outcome really comes down to variance even when the math is stacked in your favor.

The final nine return to the felt August 3-5 at Horseshoe Las Vegas to find out whether the numbers hold up.

FAQs

Bala Kumar

Founder of PokerClubGames.com and a Poker Researcher with 10+ years of experience in SEO, WordPress development, and gaming content strategy. Specializes in researching online poker sites, poker apps, tournaments, bonuses, and poker strategies. Experienced in poker platform reviews, affiliate marketing, and creating SEO-focused poker content for global audiences.

For collaborations, media inquiries, or poker-related partnerships:
Contact: Info@hugecount.com