What Is a Poker Solver? (And Why You Don't Need to Run One)

It’s Thursday night at your regular game. The biggest pot of the evening builds up, your buddy Dave thinks for a full minute, then shoves his whole stack in with a flush draw – four cards to a flush, hoping the fifth arrives. It doesn’t. He loses the pot, and while he’s counting out a rebuy he shrugs and says, “The solver says to jam there.” Everyone at the table nods like that settles it.

You nod along too. But on the drive home, you realize you have no idea what “the solver” actually is. A program? A website? A person? (“Jam,” at least, you know: it’s poker slang for going all-in.) And how does Dave know what this solver thing says about a random hand from his garage game?

Fair questions. Solvers are the most important thing to happen to poker strategy in decades, and also one of the most misunderstood. So let’s open the hood: what a solver actually is, how it works, what its answers really mean, and why – despite how often it gets name-dropped – you don’t need to buy one, configure one, or run one to benefit from everything it has taught us.

What a Solver Actually Is

A solver is software that computes a near-perfect strategy for one precisely specified poker situation. Not poker in general. One situation, spelled out in complete detail before the calculation starts.

To set that situation up, you feed the solver four things. First, the situation on the table itself: the community cards dealt so far and the money already in the pot. Second, both players’ ranges – a range is the complete set of hands a player could realistically hold given the action so far. If someone raised before the flop (the first three shared community cards), their range might be pairs, strong Aces, and some suited hands (two cards of the same suit); it certainly isn’t every possible holding. Third, the stack sizes – how much money each player has left to bet. Fourth, a menu of allowed bet sizes. You tell the solver that at each decision a player may check, bet one-third of the pot, bet three-quarters of the pot, or go all-in, and nothing else.

Then the machine goes to work, and what comes back is not a tip. It’s a complete strategy for both players: for every single hand in each range, at every decision point in the situation, how often to bet, check, call, raise, or fold, plus the expected value of each option. Expected value, or EV, is the average profit of a play over the long run. So the solver doesn’t just say “betting is good here.” It says betting this exact hand at this exact size is worth 4.31 big blinds on average, while checking is worth 4.28. (The big blind is the forced bet that sets the stakes; players measure stacks and profits in big blinds the way carpenters measure in inches.)

How a Solver Finds the Answer

Here’s the part most people never hear, and it’s surprisingly simple. There is no poker knowledge programmed into a solver. No hand-reading wisdom, no “bluff the scary cards,” no strategy books digested into code. A solver is just two virtual players locked in a room, endlessly adjusting to each other.

It works like this. Both virtual players start with more or less arbitrary strategies. Player one examines player two’s strategy and shifts toward the plays that make the most money against it. Then player two does the same thing back. Every adjustment by one side opens new weaknesses for the other side to attack, so they adjust again, and again – millions, then billions of tiny corrections. Over time the adjustments get smaller and smaller, until the two arrive at a pair of strategies where neither player can gain anything by changing further. That stopping point is called an equilibrium, and it’s exactly what GTO means: a strategy that can’t be exploited, because every exploit has already been tried and answered somewhere in those billions of iterations.

Notice what’s missing: understanding. The solver never grasped why any of it works. All the elegant concepts coaches now teach from solver output are human explanations, reverse-engineered after the fact. The machine just counter-adjusted, relentlessly, until there was nothing left to fix.

What the Output Looks Like (And Why It Feels Weird)

Say you raise before the flop from the button – the best position at the table, acting last on every betting round after the flop – and the big blind calls. The flop comes King-Seven-Two rainbow, meaning three different suits, so no flush draw is possible. You hold Ace-King: top pair (you’ve paired the highest card on the board) with the best possible kicker (the side card that breaks ties between matching pairs).

Ask a solver about this spot and it will not say “bet.” It will say something like: bet one-third of the pot 67% of the time, check 33% of the time. For one hand. In one situation.

If that answer makes you blink, you’re in good company: it’s the most confusing part of solver output for newcomers. The short version: when two actions are worth nearly the same EV, splitting between them keeps your overall strategy balanced and unpredictable, so observant opponents can’t profit from your patterns. We wrote a full guide to mixed strategies that unpacks this, including the reassuring practical takeaway: when the solver mixes, both actions are fine, and simply taking the more frequent one is never a serious mistake.

Now multiply that kind of answer across every hand in your range, and across every possible turn and river card (the fourth and fifth community cards), and you can see what solver output really is. Not advice. A giant map: thousands of frequencies and EVs describing an entire strategy, where every number is correct and none of them explain themselves.

The Fine Print: Solvers Solve a Simplified Game

Now for the part that gets left out at the table. A solver never solves real no-limit hold’em. It solves a simplified version of it, and the simplification matters.

In a real hand you can bet any amount, from the minimum up to your whole stack. In a $200 pot that’s hundreds of legal bet sizes, at every decision point, for every combination of future cards. The full game tree – every possible sequence of actions and cards – is so astronomically large that no computer can hold it, let alone solve it. So the solver only considers the handful of bet sizes you gave it. Allow only a one-third pot bet and it produces the perfect strategy for a game where that’s the only bet available. Hand it a three-quarter pot menu instead and it returns a visibly different strategy for the same King-Seven-Two flop: fewer bets, more checks, different hands doing the betting.

The ranges you enter matter just as much. Feed it a preflop range that’s slightly too tight or too loose, and every frequency downstream shifts. And nearly all solver work is heads-up – exactly two players. Your Thursday night pot with five players in it is far beyond what the standard tools will map for you.

Put all that together and “the solver says” turns out to be shorthand for “a solver, given this exact spot, these assumed ranges, and this bet-size menu, says.” Two players can study the same spot with slightly different assumptions and walk away quoting different answers, and both are telling the truth. The output is only ever as good as the inputs.

How Solvers Changed the Way Poker Is Played

A little history explains why everyone suddenly started talking like Dave.

For most of poker’s existence, strategy came from books and arguments. Strong players wrote down principles, other strong players disagreed, and there was no referee. When online poker exploded, tracking databases arrived: software that recorded every hand you played online and let you study millions of them. Now you could see what winning players actually did, which was real progress. But “what winners do” is not the same as “what’s optimal.” The winners were guessing too, just more successfully.

Then, around 2015, solvers like PioSOLVER reached the public, and for the first time strategy questions had answers instead of opinions. The effects were fast and visible. Pros discovered that the solver bet the flop small – often one-third of the pot – where the standard of the era was much bigger bets. Check-raising (checking, then raising after your opponent bets) turned out to be dramatically underused by even elite players. And the solver mixed almost everything, splitting its play between actions where humans had always hunted for the one right answer. Watch any high-stakes game today and you’re watching strategies that barely existed before machines mapped the territory.

Why You Don’t Need to Run One Yourself

Here’s the honest version of what doing your own solver work involves, because nobody who says “the solver says” ever mentions it.

First you buy the software. PioSOLVER and GTO+ are the standard do-it-yourself routes, and to be fair, both are genuinely powerful and reasonably priced for what they do. Then the real work starts. You construct realistic ranges for both players, which requires already knowing a lot about good poker – a genuine chicken-and-egg problem. You choose which bet sizes to include, knowing the answer is only as good as that menu. You wait while your computer churns: minutes for a rough answer, hours for a precise one. And when it finishes, you’re staring at that giant map – a grid of hands and frequencies with no explanations attached. Turning “check 33%” into a transferable lesson about why is a skill in its own right, and it takes hundreds of hours to develop.

Professional players take this seriously enough that many of them literally pay other people to do it. There are pros whose entire job is running and interpreting solver simulations for other pros. That’s the level of effort real solver work demands.

Now be honest about your own situation. You play a home game every week or two. You want to stop making the mistakes that cost you buy-ins and start making decisions you can actually defend. Do you need to personally operate industrial strategy-calculation software? Or do you need the lessons buried in its output, pulled out and explained in plain English?

Getting the Lessons Without the License

The good news is that solver knowledge doesn’t expire. The equilibrium for a given situation is the same today as it was last year, which means someone can compute it once, carefully, and thousands of players can learn from it. So the sensible path for a home game player is a study tool built on precomputed solutions: you practice realistic decisions, compare your choice against the solver’s, and spend your energy on understanding instead of computation.

The missing piece, historically, was the explanation – the “why” that a wall of frequencies never provides. That’s where AI poker training has genuinely changed things, and it’s the approach we took with Poker Sense: you practice hands with instant GTO feedback computed from real solver output, and when a recommendation surprises you, the “Ask Why” coaching explains the reasoning behind the recommended play. The free tier includes twenty practice hands and three AI coaching conversations a day, enough to find out whether this way of learning suits you.

However you study, the principle is the same. The slow, expensive, technical part of solver work has already been done, by people with better hardware and more patience than any of us. Your job is the part that was never automatable anyway: the learning.

The Bottom Line

A solver is a counter-adjustment machine: two virtual players sparring for billions of rounds until neither can improve, inside a simplified game defined entirely by what a human fed it – the board and pot, the stacks, the ranges, and the menu of bet sizes. Its discoveries rewired modern poker strategy, and its fine print – the answer is only as good as the inputs – is routinely ignored by the people who quote it.

You don’t need to run one. You need what it found, translated into lessons you can use on Thursday night. And the next time Dave announces that the solver says to jam there, you can smile and ask him which ranges he gave it. Watch what happens.