martha889 /
catan-advisor
I wanted to know what actually wins at Catan, so I wrote a rule-based bot, had it play itself 20,000 times (1v1 to 15 points), and tested strategy ideas head to head: each change played thousands of games against the ver…
About
I wanted to know what actually wins at Catan, so I wrote a rule-based bot, had it play itself 20,000 times (1v1 to 15 points), and tested strategy ideas head to head: each change played thousands of games against the version before it. Some results: - Holding Longest Road and Largest Army at the end wins 99% of games; holding neither wins 2%. - Each pip of opening production over your opponent is worth about 4 points of win rate. - The biggest improvement was patience: saving cards for the next city instead of spending them on roads and dev cards (62% vs. the original strategy). - Several "obvious" ideas did nothing or hurt: valuing ports highly (44.8%), an endgame mode, a deeper search. The README lists every experiment, including the rejected ones, with game counts and 95% intervals. Live demo (runs in your browser, nothing to install): https://martha889.github.io/catan-advisor/ – click "Try an example board". During setup it plays 360 whole games to rank your starting spots.
Discovery history
- Show HN
Discuss this product