Watching an AI Learn Connect Four
Your turn — click a column.
New round
Reset AI's knowledge
Train the AI
Board size
Bigger boards have vastly more possible positions. Changing this resets the AI's knowledge!
4 × 4 (tiny — learns fast)
5 × 4
6 × 5
7 × 6 (full-size Connect Four)
Learning rate (α)
0.30
How much each training game directly changes the AI behaviour.
Exploration rate (ε)
0.20
How often it tries a random move instead of its best-known one.
Games per training batch
100 games
1,000 games
5,000 games
20,000 games
Train batch
What it has learned
Games trained so far
0
Distinct positions learned
0
Last batch result
X won
O won
Draw
AI's last move
Play settings
Who goes first
Takes effect on the next new round.
You start (you're X)
AI starts (AI is X, you're O)
AI also learns from games played against you
Let the AI explore during play too