← EVolveLab · R&D

Watch a poker solver compute itself

The "GTO" everyone talks about comes from an algorithm: CFR (Counterfactual Regret Minimization). It's the same one behind PioSolver and the bots that beat the pros (Libratus, Pluribus). It works by playing the game millions of times and, at each decision, measuring how much it "regrets" not doing something else — the average strategy converges to the Nash equilibrium.

Below it runs for real in your browser, on two toy poker games where it converges in seconds. Hit solve and watch it discover the bluff, the slowplay and when to fold — with no one teaching it.

Kuhn Poker

3 cartas (J<Q<K), una ronda. El juego mínimo donde ya existe el farol. El solver converge en una fracción de segundo.

Leduc Hold'em

6 cartas, dos rondas y una carta comunitaria. Ya tiene la maquinaria del poker real: calles, board, semi-faroles. Tarda un par de segundos (corre de verdad en tu navegador).

GTO preflop range (solved by us)

This isn't a copied range: our own ES-MCCFR solver computed it over the abstracted NLHE game (heads-up, 100bb) with 600k iterations. A single solve yields all three decisions — the open, the BB defense and the response to a 3-bet. Each cell blends the action colors by their GTO frequency.

AA
AKs
AQs
AJs
ATs
A9s
A8s
A7s
A6s
A5s
A4s
A3s
A2s
AKo
KK
KQs
KJs
KTs
K9s
K8s
K7s
K6s
K5s
K4s
K3s
K2s
AQo
KQo
QQ
QJs
QTs
Q9s
Q8s
Q7s
Q6s
Q5s
Q4s
Q3s
Q2s
AJo
KJo
QJo
JJ
JTs
J9s
J8s
J7s
J6s
J5s
J4s
J3s
J2s
ATo
KTo
QTo
JTo
TT
T9s
T8s
T7s
T6s
T5s
T4s
T3s
T2s
A9o
K9o
Q9o
J9o
T9o
99
98s
97s
96s
95s
94s
93s
92s
A8o
K8o
Q8o
J8o
T8o
98o
88
87s
86s
85s
84s
83s
82s
A7o
K7o
Q7o
J7o
T7o
97o
87o
77
76s
75s
74s
73s
72s
A6o
K6o
Q6o
J6o
T6o
96o
86o
76o
66
65s
64s
63s
62s
A5o
K5o
Q5o
J5o
T5o
95o
85o
75o
65o
55
54s
53s
52s
A4o
K4o
Q4o
J4o
T4o
94o
84o
74o
64o
54o
44
43s
42s
A3o
K3o
Q3o
J3o
T3o
93o
83o
73o
63o
53o
43o
33
32s
A2o
K2o
Q2o
J2o
T2o
92o
82o
72o
62o
52o
42o
32o
22
FoldLimpSubirAll-in
Pasa el cursor sobre una mano para ver su mezcla GTO.

With this bet-sizing abstraction (limp / 3× / all-in) the equilibrium mixes a lot of limp — expected in heads-up with a coarse betting menu. Refining the menu is the next step.

Live river solver

This is what makes the bots superhuman (Libratus, Pluribus): solving the subgame in real time. Nothing here is precomputed — when you hit it, our solver solves this exact river (board + ranges) in your browser and shows you the GTO strategy for every hand.

Board:
A
K
7
2
9
bote 20 · stack 40

From Kuhn to a real-time river solver — EVolve's GTO engine, built in-house.