At chess960v2.com, I'm running a full analysis of all 960 Fischer random chess starting positions using Stockfish self-play. Over 50% of the games have been completed — and while initial results looked stable, a deep review revealed a troubling anomaly: no castling occurred in any game.
The cause wasn't time pressure (0.5s per move), but a logical bug in the system: although castling was technically possible, the engine failed to recognize valid conditions during play. This distorted the data significantly, as castling is a key strategic element, especially in asymmetric setups.
The bug has now been found and fixed. A full restart with clean data is planned. The time control will be increased from 0.5s to 1 second per move — improving stability and evaluation quality, though the core issue was logical, not temporal.
The project is open-source — code is available for anyone to run longer sessions (5s, 30s, 1min+) to study position balance and fairness.
Upcoming features include online multiplayer with custom starting positions, and later, a hidden pieces mode (in development), where only pawns are shown, and players must scout to discover the true location of enemy pieces.
Brief downtime may occur in 1–2 days due to an IP address change. Feedback welcome, especially from those interested in chess asymmetry, engine-driven analysis, and novel game design.
Chess960v2 : All 960 Fischer openings analyzed, full restart planned after fix · HackerTrans
The cause wasn't time pressure (0.5s per move), but a logical bug in the system: although castling was technically possible, the engine failed to recognize valid conditions during play. This distorted the data significantly, as castling is a key strategic element, especially in asymmetric setups.
The bug has now been found and fixed. A full restart with clean data is planned. The time control will be increased from 0.5s to 1 second per move — improving stability and evaluation quality, though the core issue was logical, not temporal.
The project is open-source — code is available for anyone to run longer sessions (5s, 30s, 1min+) to study position balance and fairness.
Upcoming features include online multiplayer with custom starting positions, and later, a hidden pieces mode (in development), where only pawns are shown, and players must scout to discover the true location of enemy pieces.
Brief downtime may occur in 1–2 days due to an IP address change. Feedback welcome, especially from those interested in chess asymmetry, engine-driven analysis, and novel game design.