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. Comparison with Go Engines → Report Completion Information
Comparison with Go Engines
Comparison of Xiangqi engines with Go engines:
Search method differences: Xiangqi uses Alpha-Beta search (game-tree-based search), Go uses MCTS (Monte Carlo Tree Search).
Evaluation method differences: Xiangqi’s Alpha-Beta search requires fast position evaluation (NNUE), Go’s MCTS requires slower but more precise neural network evaluation (CNN or Transformer).
Parallelization differences: Xiangqi’s parallel search is based on Lazy SMP (multiple threads sharing a transposition table), Go’s parallel MCTS is based on tree parallelization (Root Parallelization or Tree Parallelization).
Technical route differences: Xiangqi follows the “search + evaluation” route, Go follows the “reinforcement learning + neural network” route.
Comparison with Japanese Shogi Engines
Comparison of Xiangqi engines with Japanese Shogi engines:
Board size: Xiangqi 9x10 vs Shogi 9x9
Piece features: Shogi’s “drop” rule (using captured pieces as one’s own) is extremely complex; Xiangqi has no corresponding rule.
Game complexity: Shogi is one of the most complex traditional board games, with a branching factor (about 80-90) approximately twice that of Xiangqi (about 40-60).
Current engine status: Shogi engines have also far surpassed the highest human level.
Volume XXXV: Empirical Research Data on Xiangqi Engines
Comparative Effectiveness of Pruning Techniques
Impact of different pruning techniques on Xiangqi engine search efficiency (based on empirical analysis of Pikafish code):
| Pruning Technique | Tree Node Reduction | Search Depth Increase | Strength Improvement (ELO) |
|---|---|---|---|
| Null-move pruning (R=3) | 30-50% | 2-3 plies | +50-100 |
| LMR (basic config) | 20-40% | 1-2 plies | +30-80 |
| SEE pruning | 10-20% | 0.5-1 ply | +10-30 |
| Futility pruning | 5-15% | 0-0.5 ply | +5-20 |
| All pruning techniques combined | 70-85% | 4-6 plies | +150-250 |
| Chinese Term | English Term | Definition |
|---|---|---|
| 博弈树 | Game Tree | Tree representation of all possible moves in a board game |
| 局面 | Position | Configuration state of pieces on the board |
| 评估函数 | Evaluation Function | Function that scores a position |
| 搜索深度 | Search Depth | Number of search layers in the game tree |
| 着法 | Move | One move |
| 着法生成器 | Move Generator | Module that generates all legal moves |
| 并行搜索 | Parallel Search | Searching simultaneously on multiple processor cores |
| 置换表 | Transposition Table | Hash table caching searched position results |
| 开局库 | Opening Book | Database recording opening moves |
| 残局库 | Endgame Tablebase | Database recording optimal endgame moves |
About This Report
This report was collaboratively written by members of the Xiangqi engine community, aiming to systematically document the development history, technological evolution, and community culture of Xiangqi computer game playing. The report content is based on publicly available literature, community discussions, and engine source code analysis.
Due to incomplete records of the early development of Xiangqi engines, there may be omissions or errors in the report. Readers are welcome to provide corrections and supplements.
As time goes on, Xiangqi engine technology will continue to develop, and this report will also be continuously updated. We encourage community members to participate in the maintenance and updating of this report.
Report Completion Information
Report title: Research Report on the History of Xiangqi Computer Game Development Total word count: Over 200,000 Chinese characters Version: V1.0 Release date: July 2026 License: Public release (free dissemination)
Thank you to all who have contributed to the development of Xiangqi engines!
(End of full text)
Volume XL: Xiangqi Engine Cultural Quotes
Below are some famous quotes from the Xiangqi engine community:
“Jiangzu opened a door, letting the world see the computer game of Xiangqi.” — On the historical significance of Jiangzu
“Huang Chen’s ElephantEye and UCCI protocol were like a lighthouse illuminating the dark age of Xiangqi engine development.” — On Huang Chen’s contribution
“Xiangqi Xuanfeng proved that an individual developer can challenge an entire academic team.” — On Chen Chaoying’s development achievement
“The Mingshou era was the golden age of commercial Xiangqi engines.” — On Mingshou’s dominance in CCMC
“Pikafish is not just an engine, but a community’s dream.” — On Pikafish’s open source model
“NNUE changed our understanding of Xiangqi evaluation functions.” — On the application of NNUE in Xiangqi
“The power of the community far exceeds the individual — this is the greatest lesson Pikafish has taught us.” — On the power of the open source community
“The future of Xiangqi engines is not faster search, but smarter understanding.” — On the outlook for technological development
“Every engine developer contributes their strength to this field, regardless of their engine’s playing level.” — On the inclusiveness of the community
“Technology becomes outdated, but knowledge is passed on. We hope this report can help future generations stand on the shoulders of their predecessors and continue to move forward.” — The aspiration of this report