Tiny GTO

A vectorized CFR+ solver & trainer for Kuhn poker, Leduc hold’em and Rhode Island hold’em — the same algorithm family that essentially solved heads-up limit hold’em (Cepheus, 2015). Kuhn and Leduc are solved live in your browser; Rhode Island hold’em (~871k information sets) is solved offline and shipped as a precomputed strategy, GTO-Wizard-style.

iterations0 infosets0 exploitability game value (P1)
Bot
?
pot: —
You
?
session (saved in this browser) — reset
hands0 decisions0 chips won0 EV lost vs GTO0.000 avg loss / decision drill
Press Deal to start. You alternate seats each hand; the bot plays the solved (near-Nash) strategy. After each of your decisions you’ll see the GTO mix and how much EV your choice gave up. Costly mistakes go to the Drill queue — retire each one by getting it right twice in a row.
EV lost by street
worst open leaks (drill to retire them)
hand history (latest first, last 100 kept)
Rows: hand | board | betting (c check, b bet, k call, r raise, / new street). Bars show GTO action frequencies.