What is a poker solver, and what can it tell you?
Updated November 2026
A poker solver is a program that plays a simplified version of a hand against itself, over and over, until neither side can do better by changing what it does. What it gives you back is a balanced strategy for that one spot.
The tool below is a real solver, cut down to the smallest game that still has bluffing in it. Move the bet size and it works out the answer again.
What you give it
- Both players' ranges
- The board
- The pot and stacks
- A short list of bet sizes
What you get back
What it can't do
- Know how your opponent really plays
- Use bet sizes it wasn't given
- Guarantee anything with 3 players
How does a solver find the answer?
It starts with a poor strategy and plays itself. After every round it checks each option and asks how much better that option would have done, which is called regret. Options it regrets not playing get played more next time.
Repeat that enough and the two sides settle into a point where neither can gain by changing. Game theory calls that an equilibrium, and our guide to GTO poker explains why it's useful.
The loop
- Play every option
- Score what each would have earned
- Play the regretted ones more
- Repeat many thousands of times
What is the small solver above doing?
It's a one-street game. The first player either has the best hand or nothing, half the time each. He always bets the good hands and has to decide how often to bet the bad ones. The second player can only beat a bluff and has to decide how often to call.
The dashed lines are the answer from the formula. The solid lines start in the wrong place and find their way there by themselves.
Half-pot bet, the exact answer
What does regret look like with real numbers?
Take rock, paper, scissors. You play rock and your opponent plays paper, so you lose 1. Paper would have tied, which is 1 better than what happened. Scissors would have won, which is 2 better.
Those regrets add up to 3. So next round the solver plays paper 1 time in 3 and scissors 2 times in 3, and it never plays rock. Repeat that for millions of rounds and the mix settles at one third each.
You played rock, he played paper
How do you know when a solver has finished?
It measures how much a perfect opponent could win from the strategy so far. That number shrinks as it runs. People usually stop when it falls below a small share of the pot.
A solver result is close, never perfect. The differences left are far too small to matter at a real table.
Over time
What does real solver output look like?
Numbers for every hand and every action. This summary comes from our own solver on a simplified betting tree: the button raised, the big blind called, and the big blind has checked.
How the button plays after a check
Our own solver, simplified tree, by type of flop
On high flops the button bets nearly all the time, and mostly small. On low flops it checks far more often. That's a pattern you can take to the table, and the exact percentages aren't.
Why do solvers need a simplified game?
Because the full game is far too big. A player could bet any amount, so the number of ways a hand can go is enormous. You give the solver a short list of sizes and it works only with those.
Change the list and the answer changes. That's worth remembering any time someone quotes a solver at you.
A typical list of sizes
Where do solvers fall short?
- They assume perfect opponentsMost players are far from balanced. A plan built for their mistakes wins more.
- They depend on what you feed themGive it the wrong ranges and you get a precise answer to the wrong question.
- Three players breaks the guaranteeMulti-way pots have no balanced strategy that can't lose.
- Nobody can memorize the outputLearn the patterns and the reasons, and leave the exact numbers.
How strong have poker programs become?
- 2015CepheusHeads-up limit Hold'em was described as essentially solved, with play that couldn't be told apart from perfect.
- 2017LibratusBeat four top professionals over 120,000 hands of heads-up no-limit.
- 2019PluribusBeat top players at six-handed no-limit, the first program to do that with more than two players.
What's the difference between a solver and other study tools?
- SolverWorks out a balanced strategy for one situation you set up.
- Hand trackerRecords your own hands and shows your results and habits.
- TrainerDeals you spots and tells you what you got wrong.
- BotPlays for you at a real table. That's banned, and it's cheating.
How do you read a strategy that mixes?
When a hand shows 70% bet and 30% check, the solver is close to indifferent between the two. It isn't telling you the hand is special. If you always bet that hand you'd give up very little.
Worry about the big frequencies and the hands with a clear answer. Those are the ones that cost real money when you get them wrong.
What to look at
How should you study with a solver?
Take a spot you meet every session, like a flop bet from the button. Look at what happens on a few different flops, and write down the pattern in plain words. Then test it in a drill.
Start after you know ranges and pot odds. Everything in a solver is built on those two.
One spot at a time
- Pick a common spot
- Compare a few flops
- Write the pattern down
- Drill it
Can you use a solver while you play?
No. Real-time help is banned on the major online sites and live cardrooms won't allow it either. Solvers are for study between sessions.
Allowed?
Do you know what a solver does?
Twelve spots in three rounds.
Questions players ask about solvers
Do I need a solver as a beginner?
No. You'll win more from learning what your opponents do wrong.
Is a solver the same as GTO?
A solver is the tool. GTO is the kind of strategy it looks for.
Has poker been solved?
Heads-up limit hold'em essentially has. No-limit is far bigger, so solvers work on simplified versions.
Does a solver always give the same answer?
For the same inputs, very nearly. Change the ranges or the bet sizes and it can change a lot.
Are poker bots allowed?
No. Using a program to play for you at a real table is cheating on every serious site.
How long does a solver take?
From seconds for a small spot to many hours for a full game tree, depending on the computer and the options.
How the numbers on this page were worked out
The tool uses regret matching, running live in your browser and starting from a deliberately lopsided position.
For a bet of b times the pot, the exact answer is to bet the bad hands b / (1 + b) of the time and to call 1 / (1 + b) of the time. The flop chart comes from our own solver on a simplified betting tree.
Know someone who quotes solvers at the table?
Send them this page.