Research Report on the History of Computer Chinese Chess (Xiangqi) Game-Playing
A structured history of Chinese chess engine development from the 1980s to 2026, covering major engines, protocols, and community tooling. Engine Performance Metrics → Visualization of Test Results
☰ Contents
Engine Performance Metrics
Core performance metrics of a Xiangqi engine:
- NPS (Nodes Per Second): Indicates the engine’s search speed
- Average search depth: Indicates the engine’s search depth per unit time
- Cache hit rate: Indicates transposition table utilization efficiency
- Pruning rate: Indicates the effectiveness of various pruning techniques
- ELO rating: Indicates the engine’s strength level
Volume XXXII: Best Practices for Xiangqi Engine Testing
Test Environment Setup
Environmental setup specifications for Xiangqi engine testing:
- Use the same hardware and operating system
- Use the same compiler and compilation options
- Use the same opening book and time control
- Control processor core isolation (avoid interference from other processes)
- Record environmental information for result reproduction
Test Game Arrangement
Specifications for arranging Xiangqi engine test games:
- Opening selection: Use diverse openings (or standard opening test suites)
- Color rotation: Alternate Red and Black between the test engine and reference engine
- Time control: Use standard time control (e.g., 15s + 0.5s)
- Number of games: At least 1000 games to obtain statistically significant results
- Result recording: Record the result of each game (win/loss/draw)
Result Analysis Methods
Methods for analyzing Xiangqi engine test results:
- Win rate calculation: Win rate = Number of wins / Total number of games
- ELO calculation: ELOd = -400 * log10(1/win_rate - 1)
- Confidence interval calculation: Using normal distribution or Bootstrap methods
- SPRT decision: Using SPRT to determine whether a modification is effective
Visualization of Test Results
Methods for visualizing Xiangqi engine test results:
- ELO trend chart: Display ELO changes of the engine over time
- Win rate comparison chart: Bar chart showing win rate comparisons under different settings
- Search depth distribution chart: Display the engine’s search depth distribution
- Time usage distribution chart: Display the engine’s thinking time distribution