On August 21, 2026, GTO Wizard announced multiway ICM postflop solving. The update lets tournament players study three-way postflop spots with the solution accounting not just for ICM, but for bounties as well. That is a short sentence describing a genuinely awkward problem, and it is worth unpacking why this particular corner of the game has resisted study for so long — and what a player should actually do with the output once they have it.
What ICM Actually Does to Your Decisions
The Independent Chip Model converts tournament chip stacks into an estimate of real-money equity based on the payout structure. That conversion is the source of the trouble. In a cash game, a chip is a chip. In a tournament, chips have declining marginal value: the ones you win are worth less to you than the ones you risk losing, because doubling your stack never doubles your prize expectation. The payout ladder is flatter than the chip ladder.
The consequence is that chip EV and money EV come apart. A call that wins chips on average can still be a losing play in dollars. It is why a short stack on the bubble folds hands that are trivially profitable in chip terms, and why a big stack can apply pressure that would be reckless in a cash game. If you have not worked through the mechanics, our explainer on what ICM is covers the arithmetic, and the broader tournaments section puts it in context alongside structure, stack depth and payout shape.
The mental model worth carrying: ICM is a tax on variance that scales with how close you are to a pay jump, and it punishes the player who cannot afford to lose the pot more than the player who can. Two players holding identical hands, identical stacks and identical equity against a given range can have different correct actions purely because one of them has more to lose in real money.
Why Multiway Has Been the Blind Spot
Almost every solver output that players have studied over the last decade has been heads-up. There is a straightforward reason for that: computational cost.
A solve has to enumerate a game tree — every action, every runout, every subsequent action — and iterate over it until strategies converge. Adding a third player does not add a third of the work. It multiplies it. Each extra player adds another set of decision nodes at every point in the hand, another range to track, and another dimension along which strategies interact. Where a heads-up flop node has one opponent's response to model, a three-way node has two opponents whose actions are conditional on each other.
There is also a conceptual wrinkle. Nash equilibrium in two-player zero-sum games has convenient properties: the equilibrium strategy is unexploitable, and playing it you cannot do worse than break even. Those guarantees weaken with three or more players. Multiway games can have multiple equilibria, and there is no promise your equilibrium strategy pairs cleanly with the ones your opponents happen to have chosen. Multiway solutions are still enormously useful, but they are a different kind of object than a heads-up solve — a strong reference point rather than an unbeatable strategy.
The net effect is that the body of solver study available to players has been shaped by what was cheap to compute rather than by what comes up most often. That is a real distortion in how the modern player has learned the game.
Three-Way Spots Are Not Rare — They Are Concentrated Where It Matters
The blind spot matters more than it might appear, because multiway pots are not evenly distributed across a tournament. They cluster precisely where the money is decided.
Near the bubble and at final tables, stacks are shallow relative to the blinds and antes, and pots are pre-inflated by dead money. Flat-calling becomes more common as players avoid committing stacks preflop under ICM pressure. Short stacks shove, medium stacks call reluctantly, big stacks call wide because they can afford to. All of these forces push toward flops seen by three players rather than two. And the field is small enough that specific stack configurations recur — the short/medium/big triangle at a final table is a genuine repeating puzzle, not a curiosity.
So the spots that have been hardest to solve are disproportionately the spots that decide your tournament results. That is the gap this tooling addresses, and it follows a broader push at GTO Wizard: earlier in 2026 the platform rolled out an overhaul that added an instant multiway preflop solver and a feature called Single Size Solutions. Multiway postflop with ICM continues that work into the streets where the money actually moves.
Bounties Pull in the Opposite Direction
Mystery bounty and progressive knockout formats have been one of the major growth areas in tournament poker in recent years, with sizeable 2026 examples including GGPoker's $250 Mystery Millions and its $10 million guarantee on the WSOP Online schedule, and Americas Cardroom's dual Mystery Bounty Venoms carrying $15 million in combined guarantees. These are not niche side events. For a large slice of the online tournament population, bounty formats are the default.
Bounties invert part of the ICM logic. ICM rewards survival: it says risk is expensive, avoid marginal confrontations, let others bust. A bounty rewards elimination: it says there is a cash prize attached to the act of knocking a player out, and that prize is collected only by taking a risk. In a progressive knockout, part of the bounty you win is added to your own head, so eliminating a player both pays you immediately and increases your value as a target.
These two pressures do not simply cancel. They apply unevenly across players and situations. Calling a short stack's shove can be ICM-negative and bounty-positive at the same time, and which effect dominates depends on the size of the bounty relative to the remaining prize pool, how close the next pay jump is, how many players are covered, and — crucially, multiway — who else is in the pot and whether they can also collect.
That last point is why multiway bounty spots are so hard. In a three-way pot with a short stack all-in, the bounty is a prize only one of the two remaining players can win, and it is won by having the best hand at showdown. That creates an incentive to keep playing that is absent from a normal ICM calculation, and it changes how the two covering players should treat each other. Checking down to maximise the chance of elimination, betting to isolate, fold equity that is worth less than usual because your opponent is drawing to a prize you also want — none of this exists in a heads-up chip-EV solve. Combining opposing incentives across three ranges, inside a game tree that already resists computation, is a fair description of the hardest open problem in tournament solving.
The Trap: Memorising Output Instead of Extracting Principles
The predictable failure mode with any new solver capability is treating it as a lookup table. Players load a node, notice a hand doing something surprising, commit that to memory, then apply it at the table in a spot that looks superficially similar but differs in the variables that drove the result.
This is worse in ICM and bounty contexts than anywhere else, because the outputs are hypersensitive to inputs. Shift the payout structure, change one stack by a few big blinds, adjust the bounty size relative to the prize pool, and the correct action can flip. A memorised frequency from one configuration is not evidence about a different one. It is noise dressed as knowledge.
The productive use is to interrogate the shape of the solution rather than its digits. Ask which player is most constrained by ICM and why. Move a stack and watch which direction the strategy travels. Ask whether the bounty is large enough relative to the remaining equity to overturn the survival incentive, and at what point it stops being large enough. Those relationships transfer between spots. Specific numbers do not.
It also helps to keep the underlying arithmetic sharp, because ICM adjustments sit on top of ordinary hand-reading rather than replacing it. A tool like our poker odds calculator is useful for making those baselines automatic, so your attention at the table goes to the harder layer.
Is This Depth of Study Worth It for Everyone?
Honest answer: for a lot of players, no. The diminishing-returns curve on solver study is steep. A player leaking money through loose preflop opens, poor bet sizing, tilt, or playing stakes their bankroll cannot support will not recover it by studying three-way ICM nodes. The gains available from fundamentals are far larger and far cheaper. Fixing your game selection or your bankroll management will move your results more than any amount of node-crunching if that is where your leaks live.
The players for whom this genuinely matters are those who regularly reach final tables in fields large enough that pay jumps are meaningful, and those who grind bounty formats at volume. For them the multiway near-the-money spot is not a rare puzzle — it is the recurring decision that separates a good year from a flat one, and the gap between a well-reasoned call and an intuitive one compounds across hundreds of instances.
For everyone else, the concepts are worth absorbing even if the tooling is not. Understanding why ICM tightens your calls, and why a bounty pushes back against that, improves your decisions on its own. Our general poker strategy guides cover that ground without a subscription, and if bounty formats are your focus, our Americas Cardroom review covers a site running them at scale.
What This Means for Players
The practical takeaway is a hierarchy, not a to-do list.
First, know where the ICM pressure sits before the flop comes. Near the bubble or at a final table, you should be able to say without hesitation which players are most constrained by survival and which are least. That single piece of awareness — applied to your opens, your calls, and your decision to contest a multiway pot at all — captures a large share of the available edge.
Second, when there is a bounty in play, weigh it explicitly rather than treating it as a bonus. Ask whether the prize is meaningful relative to what you are risking and how far away the next pay jump is. Small bounties late, with big jumps ahead, rarely justify much extra risk. Large bounties early, with a flat ladder ahead, frequently do.
Third, if you do study multiway ICM nodes, come away with a sentence you can say at the table, not a percentage. "The covering player should widen here because the short stack's range is capped and the bounty offsets the survival cost" is portable. A memorised frequency is not.
Fourth, be honest about where you sit on the returns curve. This is a genuine advance in a part of the game that has been under-studied for structural reasons. It is also, for most recreational players, several rungs above the leaks actually costing them money. Both are true at once.