What is GTO in poker?

What is GTO?

GTO stands for Game Theory Optimal. It is a way of playing poker, derived from the mathematics of game theory, that is perfectly balanced. A player using a GTO strategy is unexploitable: whatever their opponents do, those opponents cannot gain an edge on them in the long run.

In technical terms, a GTO strategy is a Nash equilibrium. That is a set of strategies where no player can improve their own result by changing their own strategy while everybody else keeps theirs. At a poker table it means mixing your actions, raising and calling and folding the right hands at the right frequencies, so that no counter-strategy your opponent can find shows a profit against you.

Where the idea came from

The mathematics is much older than the poker. John von Neumann proved the first version of it in the 1920s, and the example he worked through was a stripped-down one-card betting game he could solve with a pencil. John Nash generalised the result in 1950, which is where the equilibrium takes its name. Neither of them could have applied it to a real table, because knowing that an equilibrium exists is a different problem from computing one for a game with more decision points than there are atoms in the observable universe.

What closed that gap was hardware. In 2015 a program called Cepheus played heads-up limit Hold’em closely enough to perfectly that the game is considered solved, and the same family of algorithms became the solvers no-limit players train against today. A solver does not look an answer up. It plays the spot against itself millions of times and lets the frequencies settle, which is how a solver actually works and why two different solvers handed the same spot come back with nearly the same strategy.

This is why GTO became the reference point rather than one school of thought among several. It is the output of an algorithm and it is reproducible. Two players can disagree about whether a river bet was good, and now both of them can go and check.

Poker terms, explained

The rest of this page uses the working vocabulary of poker: ranges, equity, c-bets, outs, board texture. If any of that is unfamiliar, the poker glossary defines every term in the game in plain language, and does the arithmetic on the ones that turn out to be questions about maths. The hand rankings are in it, along with a full outs table and a pot odds calculator that says exactly how much equity a call needs to break even.

Why GTO is unbeatable

Every poker strategy has a best response, which is the way an opponent would play in order to punish it most. A strategy has a leak whenever some adjustment beats it. GTO is unusual because it is its own best response. There is no adjustment an opponent can make that turns a profit against it.

It is worth being precise about what that guarantee covers, because it is narrower than it sounds. A GTO strategy does not promise to win. Against another perfect player it breaks even, minus the rake. What it promises is that your result stops depending on your opponent’s competence. You cannot be studied, you cannot be countered, and nothing anybody does at the table makes your strategy worse than it already was.

The money comes from the other side of that. GTO does not punish mistakes, it declines to be punished. Mistakes leak value to it anyway. When an opponent folds too often, calls too wide, or bets a size their range cannot support, the equilibrium collects the difference without adjusting at all and without taking on any risk of its own.

What unexploitable costs you

The guarantee is not free. Playing the equilibrium means passing up money a read would have earned. If somebody at your table folds to three quarters of all river bets, the exploitative play is to bluff far more often than the solver does, and the solver will not do it. It keeps bluffing at its equilibrium frequency and leaves the rest on the table, because taking it would unbalance the range and a competent opponent would then attack the imbalance.

That trade is the entire argument between the two styles, and it is the subject of the next section.

GTO vs exploitative play

There are two broad ways to play poker.

• GTO, or unexploitable play. You play the balanced equilibrium strategy and accept that you cannot be beaten, even though you also cannot maximally punish a specific weak opponent.
• Exploitative play. You deviate from the equilibrium on purpose, to attack a tendency you have observed. This wins more against players who are wrong, and it opens leaks of your own.

The insight that matters is about order rather than preference: you can only deviate safely from a baseline you know. Take an opponent who almost never bluffs the river. The exploitative response is to fold more. How much more? Without the equilibrium calling frequency for that spot, "fold more" is not a strategy, it is a mood, and it will follow you to the next table against somebody it does not fit.

Deviations are also not symmetric in cost, which is genuinely useful to know. Expected value near an equilibrium is flat. Move a few percentage points off the solved frequency and you lose almost nothing, so a soft read is cheap to act on. Move thirty points and the loss accelerates, and your whole range is unbalanced downstream of the decision. Small adjustments on weak evidence, large adjustments only on evidence you will rarely have.

When to leave the equilibrium behind works through the deviations that are actually worth making, and what each one costs when the read is wrong.

Four things GTO is not

Most of the resistance to GTO comes down to four beliefs about it. All four are worth taking seriously, because each one is half right.

That it means playing like a robot

Mixed strategies look mechanical written down and are anything but at the table. A solver playing a hand 70 percent raise and 30 percent call is not hedging. It has found that both actions earn the same amount against a correct opponent, and that varying between them is what stops the opponent ever knowing which is coming. The unpredictability is the point. What actually looks robotic is the opposite: one hand, one action, every single time.

That it does not work against bad players

It works, it just does not maximise. Against a table of weak players the equilibrium still wins, because their mistakes pay it automatically. What it will not do is extract everything available. That is a real argument for learning to deviate. It is not an argument against knowing the baseline, since the size of any deviation is measured from it.

That you have to memorise everything

Nobody memorises a solver. The full solution to no-limit Hold’em does not fit in a human head and never will. What strong players carry is much smaller: the preflop ranges for the seats they actually play, a feel for which board textures favour which range, and enough repetition that the common spots come back without calculation. The rest is recognising which known shape a new spot resembles.

That it is only for high stakes

The opposite is closer to true. High stakes is where opponents are competent enough that deviating from the equilibrium gets punished quickly. At low stakes the equilibrium is mostly a way of not donating: it removes the leaks that cost the most, which are the ones you have without knowing you have them.

GTO preflop fundamentals

Preflop, GTO strategy is expressed as ranges, meaning the full set of hands you play a given way from a given seat, each with a frequency attached. Four ideas carry most of the weight.

Ranges, not hands

You do not ask what to do with ace-king in isolation. You ask what your whole range does from that seat, and where ace-king sits inside it. The standard 13 by 13 grid shows every starting hand at once, coloured by action and by how often each action is taken. Reading one properly takes about ten minutes and is the highest-return thing a new player can learn.

Position decides how wide you play

Acting last is a structural advantage that lasts the whole hand, so equilibrium ranges widen steadily as the button approaches. The effect is large enough to be worth carrying as a shape rather than as a set of numbers.

Facing a raise

Opening ranges are called RFI, for raise-first-in. Once somebody has opened, the decision splits into 3-betting, meaning re-raising, and flat calling. A GTO 3-betting range is built from two kinds of hand at once: the ones strong enough to want a bigger pot, and a measured number of hands too weak to call, chosen as bluffs because of the cards they remove from the opponent. Holding an ace makes it less likely anybody else holds one, which is why so many equilibrium 3-bet bluffs contain one.

Defending the blinds

The blinds are where most preflop money is lost, and the reason is that folding there is not free. Money is already in the pot and the price to continue is discounted, so the equilibrium defends the big blind far wider than most players find comfortable. Wide defence is not the same as wide calling, though. A large part of that range is played as a 3-bet rather than a call, because calling out of position with a weak hand is how the discount gets handed straight back.

Exact frequencies depend on format, stack depth and seat count, and guessing at them is the part worth skipping. The free chart library has the solved ranges for cash, tournaments and Sit and Gos, with Omaha charts alongside them, and what the numbers on a range chart mean covers how to read the frequencies rather than just the colours.

GTO postflop fundamentals

After the flop the same range logic continues, now shaped by the board. Postflop decisions come down to expected value, which in practice means the average number of big blinds a choice wins or loses against a correct opponent.

Who the board belongs to

Before anything else, a flop divides the two ranges unevenly. An ace-high board favours whoever raised preflop, because their range holds far more aces. A board of low connected cards favours the caller, whose range contains the suited connectors and small pairs the raiser folded. This is range advantage, and it decides how often somebody can bet more than the two cards in their hand do.

Continuation betting

The preflop raiser often c-bets the flop, but not always and not with everything. The equilibrium picks a frequency and a size that keep the betting range balanced between value hands and bluffs, so the defender cannot profit by folding everything or by calling everything. On boards where the raiser holds a large range advantage, the solver tends to bet small and very often. Where the advantage is thin, it bets rarely and large.

Sizing is part of the strategy, not a matter of taste

A large bet needs a range polarised into strong hands and bluffs, because middling hands cannot profitably put that much in. A small bet suits a wide range of medium-strength holdings that want to deny equity cheaply. Choosing a size you cannot support with a plausible range is one of the most reliably punished mistakes in poker, and it is visible to anybody paying attention after about three hands.

Board texture

Dry boards with few draws and wet boards with many call for different strategies, and the gap is wider than most players expect. On a dry board equity is largely settled, so betting is about protection and thin value. On a wet board equity is still moving, both ranges contain far more semi-bluffs, and the correct frequencies shift with every card that lands. This is the part that only clicks through repetition on real boards.

The concepts that do the work

Underneath the frequencies are half a dozen ideas that recur in every spot. Once these are familiar, most solver output stops looking arbitrary.

Indifference

Much of the strangeness in solver output comes from one principle. At equilibrium, a well-constructed betting range makes the opponent indifferent between calling and folding with their marginal hands, meaning both choices earn them exactly the same. That is not a coincidence, it is the mechanism. If calling were better the opponent would simply always call and the bluffs would stop working. Frequencies are chosen precisely to sit on that line.

Minimum defence frequency

The other side of the same coin. Facing a bet, there is a share of your range you have to continue with in order to stop a bluff being automatically profitable. Against a pot-sized bet that share is 50 percent; against a half-pot bet it is 67 percent. MDF is a useful check against overfolding, but it applies cleanly only against an opponent whose bluffing is already balanced. Against somebody who does not bluff enough, folding more than it suggests is correct.

Polarisation

A polarised range is split into strong hands and bluffs with little in between, and it is what justifies a big bet. A linear or condensed range is weighted towards medium strength and wants small bets. Asking which of the two a range resembles answers the sizing question most of the time.

Blockers

The cards in your own hand change what your opponent can hold. Holding the ace of spades on a three-spade board means the opponent cannot have the nut flush, which makes your hand a better bluff even though it is worth nothing at showdown. Blocker effects are why solvers bluff with specific junk rather than with any junk, and why two hands holding identical equity can have opposite correct actions.

Equity realisation

Raw equity is not what you collect. A hand realises less than its share when it is out of position, dominated, or hard to play on later streets, and more when position lets it see cheap cards and fold cheaply. This is why suited connectors outperform their equity and offsuit broadways underperform theirs, and it is most of the reason position widens ranges at all.

One hand, start to finish

All of that is easier to follow attached to a hand. Here is one, six-handed, at 100 big blinds.

It folds to you on the button with the ace and five of spades. The equilibrium opens this hand every time, and not because it is strong, because it is not. It plays well. It is suited, it makes the best possible flush rather than a second-best one, and position means it can fold cheaply on the streets where it misses. You raise to two and a half big blinds and the big blind calls.

The flop is the king of spades, the nine of spades and the three of diamonds. Ask who the board belongs to before asking what you hold. You raised from the button, so your range holds more kings and more strong aces than a big blind defending range does. The equilibrium therefore bets here often and small, around a third of the pot, with almost its entire range.

Your own hand is a good illustration of why. It has no showdown value whatsoever right now. It also has a draw to the best flush available, and an ace that removes some of the opponent’s strongest continues. It is a hand that wants to bet, is happy to be called, and can keep going profitably if raised. You bet a third of the pot and get called.

The turn is the seven of hearts. Nothing has changed except that one card remains. Now sizing starts to matter: with a draw this strong the equilibrium bets large some of the time and checks some of the time, and that split is not indecision. Betting every time would leave the range too bluff-heavy for the opponent to go on calling. Checking every time would announce that a check means weakness. The mix is what keeps both options credible.

Say you bet, the opponent calls, and the river is the two of clubs. The flush missed. Your hand is now ace-high, which beats nothing the opponent will call with and only their own busted draws, which they will fold. That is exactly what qualifies it to bluff: a hand that can never win by checking gives up nothing by betting, while a weak pair that wins at showdown sometimes would far rather check and take its chances. This single rule decides most river bluffs, and it is why the missed draw fires and the small pair does not.

Nothing in that sequence needed arithmetic at the table. It needed knowing who the board favoured, what a size implies about a range, and why a hand with no value is the right one to bet. That is what repetition builds, and it is what a chart on its own cannot.

Cash games, tournaments and short stacks

The principles hold across formats. The inputs change, and two of them change enough to produce genuinely different strategies rather than adjusted ones.

Stack depth

Everything above assumed 100 big blinds. Shorten the stacks and the game simplifies in a particular direction: fewer streets of betting remain, implied odds fall, and hand values shift towards raw equity and away from playability. At 20 big blinds a suited connector has lost most of what made it valuable, while a small pair keeps nearly all of its. Below about 15 big blinds the game collapses into a single decision and has been solved exactly, which is why push-fold charts are not an approximation of anything. They are the answer.

Chips that are not money

In a tournament, chips won are worth less than chips lost, because a stack converts to prize money on a curve rather than a straight line. This is ICM, and it makes correct tournament play measurably tighter than the chip equilibrium in exactly the spots where busting hurts most: near the bubble, and at a final table with steep pay jumps. A call that is clearly correct in a cash game can be a clear fold in a tournament with the same cards and the same stack.

More players, more folding

Seat count moves ranges the same way position does. A nine-handed table has four more players to get through before the button, so opening ranges tighten across the board. Everything else, the range logic and the sizing logic and the board reading, carries across unchanged.

The chart library is split along exactly these lines, with separate solved ranges for cash, tournaments and Sit and Gos at each stack depth, because they are different solutions rather than one chart with adjustments applied.

Key GTO terms

• Nash equilibrium. The balanced set of strategies where no player can profit by unilaterally changing their play. GTO poker is a Nash equilibrium.
• Range. All the hands you would play a certain way in a spot, with frequencies, rather than a single hand.
• Solver. Software that computes GTO strategy for a spot by iterating towards equilibrium.
• Mixed strategy. Playing the same hand different ways a set percentage of the time, for example raise 70 percent and call 30 percent, so the action carries no information.
• EV, or expected value. The average long-run result of a decision, measured in big blinds.
• Balance. Mixing value hands and bluffs so opponents cannot exploit you by always calling or always folding.
• MDF, or minimum defence frequency. The share of your range you must continue with against a bet to stop bluffs being free.
• Blocker. A card in your hand that reduces the number of strong holdings your opponent can have.

Those eight are the ones this page rests on. For the other two hundred, from the turn to the nuts to equity realisation, see the full poker glossary.

How to actually learn GTO

Reading about GTO builds understanding. Playing it builds instinct, and the two are not close substitutes. Anybody who has studied a chart, felt confident, and then misplayed the same spot an hour later at the table has already met the gap.

The reason is that poker decisions arrive under time pressure and with imperfect recall. What closes the gap is repetition with immediate feedback: get dealt a real spot, decide, and find out at once what the solver does and what the difference cost.

1. Learn to read one chart before memorising any. A 13 by 13 grid is a compact object and takes about ten minutes to understand properly. Until it is readable, every chart is a wall of colour.
2. Take one seat at a time. Start with the button and the big blind. They are the two seats you play most often and where the most money is decided. Adding a seat once the previous one is automatic is far faster than working at all six together.
3. Train the spots rather than reading them. Deal yourself the seat you are learning, act, and read the correction. A hundred graded repetitions of one spot beats a week of studying charts you are never asked to recall under pressure.
4. Move to postflop only once preflop is boring. Postflop decisions inherit their ranges from preflop ones. Studying c-betting while still unsure which hands you opened with is building on sand.
5. Then bring your own hands back. Once the equilibrium is familiar, the most valuable study material is the hands you played for money, priced against the solution, because those mistakes are actually yours.

That sequence is what the guided path is: seventeen chapters in order, dealing real spots at a real table, with every decision priced in big blinds against a solved reference. Chapters one and two are free and need no card. The chart library and the solver are open with no account at all, and choosing a poker solver covers what to look for if you would rather run your own.

Start learning

Understanding the equilibrium and being able to play it are different skills, and only one of them is built by reading. The trainer deals real spots and prices every decision against a solved reference, which is the shortest route from knowing what GTO is to playing it without thinking.

Frequently asked questions