Sorry, this interactive is only available on screens wider than 490px, and optimised for those wider than 610px.
trAIning ground
In this game, take turns with the AI (Nathaniel) picking 1, 2, or 3 sticks.
Whoever picks the last stick wins!
Watch how Nathaniel's picks change as it plays more games.
Sticks remaining:
AI wins:
Games played:
Game parameters:
How likely is the AI to start each game?
How many sticks in the pile?
How much to change after each game.
Nathaniel can practice before we start.
What is Nathaniel's opponent during simulations?
The smartbot has a probability table that favours "correct" choices.
The fewer sticks that remain, the more likely it is to choose a number that prevents its opponent from winning.
The randobot has an equal probability of choosing any valid number of sticks.
It cannot choose more sticks than there are remaining.
Both AI will use the same probability table to face eachother.
Whether they win or lose, both will contribute to Nathaniel's neural network.
Simulations will be hidden but much faster.
This definition is not available in English, sorry!