AlphaLudo V17 · Two and four players
AlphaLudo is a 587,691-parameter graph transformer that plays both two- and four-player Ludo with the exact same network. Trained from random weights without shaped rewards, it runs entirely in your browser: no server, no account, no tracking.One 587,691-parameter network for two- and four-player Ludo, running right here in your browser. No server, no account.
01 Chance
Every turn starts with a roll you don’t control. What you do with it is the whole game, and it’s the only part a network can learn.
02 One move
It doesn’t look ahead. In a single pass it scores every square you could move from, all 225 of them, and plays the best. About 13ms, on your device.
03 Two games, one brain
Earlier models were locked to two players. V17 treats every square as a graph node carrying 5 numbers: piece counts for every seat relative to the mover, and safety. A player who isn’t at the table is simply absent, which is how one single network plays both games.
Inside the model04 Your turn
It beats V15.2, the model it replaced, 56.1% of the time, and at a table of four it holds its own against three copies of the search player it learned from. Against you, the dice still have a say.
It knows its odds
Besides choosing moves, the network estimates who will win. This is one real game of AlphaLudo (V15.2) against itself: it rated green at 78% at best and 9% at worst, and the dice swung it more than once. You see the same estimate live while you play.
Your turn
Challenge AlphaLudo head-to-head or face three neural seats at a four-player table. Pick your colour, pick up the dice, and see real-time win estimates move by move.
Take your seatThe board is empty.
Loading the model…
AlphaLudo plays gold, the corner opposite yours.
You always move first. Your colour changes nothing but where you sit.