AI's Probability of Winning
- The AI estimates its probability of winning based on its playing history.
- Board configurations leading to wins will have their probability increased, while those leading to losses will have their probability decreased.
- Probabilities are updated in real-time. If multiple people are visiting this site at the same time they will update each other's probabilities.
- If the AI's memory is wiped, all of its probabilities will go back to neutral (50%).
- The AI can be quickly retrained by running simulations in batch mode.
Temporal Difference Learning
- Initially, the AI moves randomly unless its next move results in a win.
- Moves leading to a win will become preferred in the future. Similarly, moves leading to a loss will be avoided.
- The AI will sometimes make an exploratory move. These moves will ignore what has been learned in the previous games in order to explore alternate strategies.
- Click here for more information on temporal difference learning.
- AI Logic: Python
- Backend: Python Flask Framework