Hook
The 2018 World Series of Poker Main Event didn't end with a hero call. It ended with a fold John Cynn didn't make.
Ten hours earlier, he made one he shouldn't have. On a paired board against Tony Miles' river all-in, he folded a full house — though nobody in the room, Cynn included, seemed to know it was a full house. The commentary called it two pair. Miles flashed 7-high. The room gasped.
QuintAce's solver went back and looked at three of the turning-point hands of that final table. The first — Cynn's pocket Kings against Nicolas Manion's A-10 shove — is the snap-call the solver would make a hundred times in a hundred games. The third — Cynn's trip Kings against Miles' Queen-high on the final hand — is the same story: pure, mandatory, trivially correct.
The middle hand is where the solver and the champion disagree.
decisions
(full house, folded)
after the disagreement
What makes a champion isn't that the disagreement never happens. It's what happens after.
Hand 1 — Nicolas Manion, Day 8, 9-handed final table
The final table started nine-handed at the Rio on July 12, 2018. Nicolas Manion had arrived as the overall chip leader. By midway through Day 8, he was short-stacked. John Cynn had built a mid-stack into the top three. The hand that closed the first night was the one that announced where the stacks were actually going.
Solver verdict: mandatory call. This is the cleanest peer-validation in the slate. The solver gives K♠K♣ facing a 30-big-blind all-in a pure call — no fold frequency, no mixed strategy, no argument. Kings are the second-best starting hand in No Limit Hold'em. Against an opponent pushing over a button open, the typical shoving range is smaller pairs, high broadway hands, and strong Aces — all of which pocket Kings crushes.
Cynn's snap-call was the shape of a player who had internalized that math years earlier and didn't have to think about it. The cards cooperated: Manion's A-10 never found a ten or an ace. But even if they had, the solver would tell you the same thing every time. This wasn't a read. This was a preflop commit.
Source: QuintAce solver decision_analysis_tool on the reconstructed preflop node. hero_reach = 100% (pure preflop node — maximum trust). See Methodology & caveats.
Hand 2 — Tony Miles, Day 10 heads-up, early in the marathon
Then the 199-hand heads-up match began, and somewhere inside it, the most-replayed hand of the tournament happened.
Miles had built a reputation across nine days of Main Event play as the player who wasn't afraid to put his tournament life on the line. Across ten hours of heads-up, the lead changed eleven times. They traded three-bets and barrels and stares. And then, on a strange paired board, Miles fired a river shove that was 3.5 times the size of the pot.
The commentary, watching live, called John Cynn's hand "two pair." So did most of the poker media in the weeks after. It was wrong.
Solver verdict: pure call. Every time. The solver runs K♣J♣'s — no, 4♦3♠'s — three of a kind of 3s with a pair of 4s behind it, which is a full house — through this exact river node and goes all-in 100% of the time. Every combo. No mixing. No equity-vs-bet-size tradeoff. Folding isn't in the strategy grid. The expected value of the call is +78.4 big blinds — a figure dwarfed only by the +94 that the final hand will carry ten hours later.
Cynn folded. He tanked for several minutes. He settled on a fold. Miles flashed the bluff. The room gasped. The ESPN production crew had footage of a hand that would re-circulate on poker social media for the next decade.
Why did he fold? The solver doesn't know; the solver has full range information. In the moment, under fluorescent lights at the end of a ten-hour heads-up session on the third day of a three-week tournament, Cynn likely saw a paired board and a massive overbet and thought about the hands that beat him. A 3 in Miles' hand (for quads). A J (for Jacks-full). A King (for Kings-full on the river card). A straight — the wheel, or a 6-5 rivering a gutshot earlier.
He may not have registered — in the moment — that his own hand was a full house. The board pair does that. The commentary team didn't see it. The poker media that reported the hand didn't see it. A player trying to fold quickly to protect what he has left doesn't always do the full rank-calculation on a paired board at midnight.
This is the point at which our article could become "solver grades Cynn." It doesn't. The hand the solver calls is the hand Cynn folded. That's the data. But what makes a champion is not that this never happens. What makes a champion is that 199 heads-up hands later, he's still in his seat. The fold doesn't break him. Most players who fold a winner in a moment that big at a stakes that high don't recover in the same session. Cynn did.
Source: QuintAce solver decision_analysis_tool + solver_tool on the reconstructed river node. hero_reach = 16.7% (solid — above the 5% trust floor). See Methodology & caveats.
Hand 3 — Miles eliminates Dyer, and the heads-up begins
The hand that created the 199-hand heads-up wasn't Cynn's. Three-handed play had opened Day 10 with Michael Dyer, the long-time chip leader, and short-stacked Miles and Cynn. On about the 18th hand of 3-handed, Miles opened for 4 million with A♠J♥ from the button, and Dyer jammed 22.2 million over the top from the big blind with A♥10♦.
Miles called. Dyer had a kicker problem. The board ran out Q-5-3-J-Q. Dyer finished 3rd for $3.75 million. Miles got the chips and a ten-hour heads-up match with Cynn.
We don't run the solver on this one. John Cynn wasn't in the pot. But it's the hand that set up everything that followed — the A-J vs A-T cooler between the two players who'd both outlasted the field. One of them was going home with $5 million, the other with $8.8. It took 199 hands to figure out which was which.
Hand 4 — The final hand
Ten hours later, they were still playing.
Cynn was the chip leader by the time this hand dealt. He raised to 9 million on the button with K♣J♣. Miles three-bet to 34 million with Q♣8♥. Cynn called. On the flop of K♥-K♦-5♥, Miles fired 32 million and Cynn called with trip Kings and a Jack kicker. On the 8♦ turn, Miles had picked up a pair of 8s — apparently thought he was ahead or near it — and shoved his last 114 million.
Solver verdict: call. Massive value. At a stack-to-pot ratio of 0.6 on the turn, the solver enters pure commit territory. Trip Kings with a Jack kicker is the premier value combo in Cynn's checking range — the solver's expected value for this class of hand at this node is +94.2 big blinds. In plainer terms: this is one of the largest +EV calls the solver ever recommends. The board is paired and two-toned, which slightly raises the threat of a full house, but it substantially reduces the real equity of the flush and straight draws Miles can have. Cynn's own trip Kings block most of Miles' big-King-x combos.
Cynn's earlier check on the turn — which gave Miles the chance to shove into him — was itself an acceptable variation in the solver's strategy. The solver prefers betting 8.2 big blinds on the turn with K♣J♣ at 73.7% frequency; checking is a 7.4% mix. Cynn took the low-frequency line and walked into exactly the result the line is designed to produce: a massive value transfer when villain commits.
Miles was drawing dead. The 4♠ on the river was a formality. John Cynn won the 2018 World Series of Poker Main Event.
Source: QuintAce solver decision_analysis_tool + solver_tool on the reconstructed turn node. hero_reach = 0.17% (low reach — combo-level figure is directional rather than precision-calibrated; the range-level claim "trip Kings at SPR 0.6 is a mandatory commit against any turn shove" is robust). See Methodology & caveats.
The pattern
Three hands. Two solver-approved. One solver-disagreement. Then a bracelet.
The middle hand is the story. The outside two are what happens when the middle hand doesn't break you.
Cynn's 2018 Main Event is not the story of a player who ran GTO lines before GTO existed. He didn't. The tank-fold against Miles on a paired-board river is a full house the solver calls 100% of the time, and Cynn folded. That happened. The historical record doesn't need to be adjusted for it.
What it is, instead, is the story of a player whose discipline survived one concrete disagreement with optimal play. Ten minutes after the fold, he was still in his seat. An hour after the fold, the chip lead had moved back to him. Four hours after the fold, he had Q♣8♥ committed on K♥-K♦-5♥-8♦ and he was making a call the solver agrees with at +94 big blinds of expected value. And on the first hand of the final table, two days earlier, he had Kings against Manion's A-10 and snap-called the only play that made sense.
The solver says Cynn got two of these three right. The solver also says the one he got wrong was worth 78 big blinds of folded expected value — not a rounding error. A lot of tournaments end when that fold happens. This one didn't, because Cynn kept his seat, stayed composed, and played 199 more hands.
That's the champion's version of correct play. Not "never be wrong." Not "solve every spot pre-session." Take the one you got wrong and keep playing — because the field is still in your seat, and the next 199 hands still have to happen, and you still know how to play them.
The 2018 Main Event didn't reward the player who ran the cleanest strategy. It rewarded the one who recovered fastest from the moment he didn't.
Methodology & caveats
Hands graded: Three hands were constructed as game-state JSONs from publicly reported final-table coverage (PokerNews hand-by-hand, WSOP.com, 888poker, ESPN, PokerGO, CardPlayer). Hole cards, betting lines, and outcomes were cross-checked across at least two independent sources per hand. Game states were run through QuintAce's decision_analysis_tool and solver_tool at each hand's key decision node. Outputs cited include solver action frequencies, expected value in big blinds, range composition, and hero_reach (the probability that the hero's exact combo reaches the exact node under equilibrium play).
The hand the commentary got wrong. Widely reported during and after the event as "John Cynn folded two pair (4s and 3s) to Tony Miles' 7-high bluff." This is inaccurate. Cynn's hole cards were 4-3 on a board of J♥-4♣-3♥ / 3♦ / K♣. The turn paired the board 3, giving the board two 3s. Cynn's 3 combines with the board's two 3s for three-of-a-kind 3s, and his 4 combines with the board 4 for a pair of 4s — a full house, 3s full of 4s. This is what the solver evaluates, and it's what was actually in Cynn's hand when he folded. The media shorthand ("two pair") collapsed the hand into a simpler category. We report the accurate hand rank throughout.
Solver model: QuintAce's cash NLHE solver. The final-table hands in this article are either heads-up with top-two payouts locked in (Hands 2 and 4) or a 3-max preflop spot where ICM still applied at the margin (Hand 1). For Hands 2 and 4, the cash approximation is tight because ICM has collapsed once HU starts at locked top-two. For Hand 1, ICM adds defensive weight to Cynn's call with KK (a short stack shoving into the button's open range at a pay-jump bubble is even more skewed toward playable hands) — which only strengthens the "call" verdict. The cash-model +EV figures are chip-EV lower bounds; tournament-EV at the pay-jumps would equal or exceed.
Reconstructive sizing: Blinds, ante, and stack depths were reconstructed to match the publicly reported pot progression. Where source coverage didn't preserve an exact pre-decision-node bet size, the reconstruction targets the documented pot at the decision node. Minor preflop or early-street sizing variance doesn't change any of the decision-node solver verdicts at SPR below 1 (Hands 1 and 4) or above medium-depth SPR (Hand 2 river at SPR 6.8 with a 3.5x pot overbet).
Reach verification (per METHODOLOGY §6c): every combo-level claim in the article carries a hero_reach value — the probability this exact hand reaches this exact node at equilibrium. Target is 40%+; hard floor is 5%; below 5% we shift prose claims to range level.
| # | Hand | Decision node | hero_reach |
Trust level |
|---|---|---|---|---|
| 1 | KK vs Manion A-10 shove | Preflop call-off | 100% | Maximum — pure preflop node |
| 2 | 4-3 facing 3.5× river overbet | River fold (solver says call) | 16.7% | Solid — above the 5% floor, combo verdict calibrated |
| 4 | K♣J♣ trip Kings facing turn shove | Turn call | 0.17% | Low — claims shifted to range level |
Why Hand 4 is low-reach: iconic televised hands frequently involve non-equilibrium actions from at least one player. Miles' 3-bet preflop with Q-8-offsuit and his turn-shove line are both polarization deviations from equilibrium. Cynn's specific K♣J♣ at the specific turn node with these specific stack depths rarely arises in equilibrium simulations. The combo-level figure (+94.2 BB EV for "the class of trip Kings at this node") is directionally correct but not precision-calibrated. The range-level claim — at SPR 0.6 on a paired two-toned board, trip Kings with a kicker must commit against any turn shove range — is robust and independent of reach.
What's not in this article. We do not solve the 199-hand heads-up match end-to-end; the article picks three turning-point hands. We do not analyze Miles' A-J vs Dyer's A-10 (Hand 3 in narrative order) with the solver because Cynn was not the hero. We do not build a claim about Cynn's decision quality over the full 442 final-table hands — we don't have the hand-by-hand data to support that.
What this article is. A retrospective on three specific decisions that defined the outcome. Two the solver agrees with. One it doesn't. And a champion's response to the one it didn't.
Solver analysis by QuintAce. Hand provenance and per-hand solver runs documented in the hands/ and phase1-solver-runs.md folders of this article's workspace. John Cynn's post-publish response, if any, will be captured as a separate closing note.