Chapter 25 Tian Tian Xiangqi (Daily Xiangqi): The Chess Platform Dominator of the Mobile Internet Era

25.1 The Birth of Tian Tian Xiangqi and Tencent’s Strategy

Tian Tian Xiangqi (Daily Xiangqi, 天天象棋) was developed by Tencent’s Guanghua Studio Group (later renamed Tianjizhiqu Studio) and officially launched in 2015. It was Tencent’s first Xiangqi product on the mobile game platform and the core vehicle for migrating QQ Games’ Xiangqi business from PC to mobile.

Birth Background:

2015 was a critical turning point for China’s mobile internet. WeChat’s monthly active users had surpassed 600 million, and the mobile gaming market was experiencing explosive growth. Tencent had operated the QQ Games Xiangqi channel for over a decade, accumulating tens of millions of Xiangqi users. But these users were rapidly migrating to mobile, and Tencent had to launch a product that could meet mobile user experience expectations.

Tian Tian Xiangqi was born exactly under this background. From the start, it was mobile-native in design — comprehensively optimized for touch screen operation, small screens, and fragmented time. This stood in stark contrast to OurGame’s passive mobile transition after its PC success.

Core Advantages of the Tencent Ecosystem:

Tian Tian Xiangqi’s biggest competitive moat was not product features but deep integration with the Tencent social ecosystem:

1. WeChat/QQ One-Click Login: No account registration needed; click to start playing
2. Friend Battles: Directly invite WeChat or QQ friends to “play a game” via chat session links
3. Social Sharing: Game notations can be one-click shared to Moments and Qzone
4. Social Fission: “Your WeChat friends are also playing” prompts drive viral spread
5. WeChat Mini Program Version: No app download needed; click to play, greatly lowering the usage barrier

The combined WeChat + QQ user pool of over 2 billion monthly active users provided Tian Tian Xiangqi with a traffic foundation that no competitor could match.

25.2 Core Features and Innovations of Tian Tian Xiangqi

Tian Tian Xiangqi’s product features have undergone continuous iteration, building an ecosystem covering the full chain of “Play → Learn → Social → Tournament → Certification”:

I. Playing System:

1. Standard Games: 10-minute, 20-minute, 5-minute blitz, and other time controls
2. Variant Modes: Reveal Chess (揭棋, a flip-chess variant popular in Hong Kong/Macau/Taiwan), Flip Chess (翻翻棋, casual variant)
3. New Game Modes: Added Mini Xiangqi and Fun Doubles Xiangqi (team battle) in 2025
4. Strength Evaluation (棋力评测): Tian Tian Xiangqi’s core competitive system, establishing the most authoritative ranking system recognized across Chinese Xiangqi online platforms

Tian Tian Xiangqi’s Rank System:

Tian Tian Xiangqi’s ranks from low to high are: Amateur Level 1 → Amateur Level 2 → … → Amateur Level 9 → Divine 1 → Divine 2 → Divine 3. There are also Professional ranks (Pro 1-1 to Pro 3-3) for higher-level differentiation.

Approximate correspondence between ranks and real-world playing strength:

• Amateur 1-8: Village/County level
• Amateur 9 - Divine 1: City/Provincial level
• Divine 2-3: Master/Grandmaster level

The strength evaluation uses the ELO rating system, with a base time control of 10 minutes + 10 seconds per move. This system has become the “folk strength standard” for Chinese Xiangqi enthusiasts — when two players meet, “What’s your Tian Tian Xiangqi rank?” has replaced the traditional “What’s your level?”

II. AI and Analysis System:

Tian Tian Xiangqi comes with a built-in 9-level AI, covering the full range from beginner to top-tier, from “Newbie” to “Grandmaster.”

In 2024, Tian Tian Xiangqi underwent a major AI system upgrade, launching a search-free supervised learning AI:

• Uses hundreds of millions of human game records for supervised learning training
• No longer relies on traditional game tree search; instead learns human players’ move patterns
• Response time reduced from the traditional 1.6 seconds to 10 milliseconds
• Capable of simulating different play styles and levels of human opponents
• Training and inference costs reduced by 98%

The core value of this innovation lies in “humanization”: traditional engines have cold, mechanical, perfect play styles, while Tian Tian Xiangqi’s new AI can simulate the rhythm and preferences of human thinking, providing a more natural and educational playing experience.

Analysis features:

• Smart Review: Analyzes the quality of each move, generates game quality reports
• AI Insights: Assists with opening and midgame choice analysis
• Position Analysis: Real-time post-move analysis
• Endgame Challenges: Progressively difficult challenge mode
• 60,000+ Problem Training Library: Includes thousand-year classic endgames and real game exercises

III. Teaching and Training System:

Tian Tian Xiangqi has built the most comprehensive online education system in the Xiangqi domain:

• Online Courses: Systematic courses from beginner to basic checkmate techniques, opening setups, midgame tactics
• Master Courses: 30+ signed master live streams and video teaching (including Grandmasters Wang Tianyi, Xu Yinchuan, etc.)
• Game Notation Collection: Massive collection of ancient manuals, master games, and tournament games
• National Amateur Rank Certification: In cooperation with the Shenzhen Culture, Radio, Television, Tourism and Sports Bureau, completes online amateur rank certification in 15 seconds across 3 steps, issuing both electronic and paper certificates
• 412 Offline Teaching Locations: Tian Tian Xiangqi Education Center has established 412 offline teaching locations nationwide, with 2025 training revenue of approximately 379 million RMB

IV. Social System:

Tian Tian Xiangqi’s social features are deeply embedded in the Tencent ecosystem:

• WeChat/QQ Friend Battles: Directly invite friends, send a link to start a game
• Club System: Create or join chess clubs, organize internal tournaments and events
• 10,000-Spectator Mode: Real-time viewing of expert games
• Game Notation Wall/Sharing: One-click save game as notation, share with friends, Moments, Qzone
• Live Streaming Interaction: Watch friends’ games and comment
• Cross-Platform: WeChat Mini Program requires no download, click to play

V. Tournament System:

Tian Tian Xiangqi has established a complete tournament system from online to offline:

• “Tian Tian Xiangqi Cup” Year-End Finals: A-level official tournament certified by the Chinese Xiangqi Association, with a total prize pool of 1.2 million RMB
• National Xiangqi League (象甲联赛): Tian Tian Xiangqi has been the title sponsor for 13 consecutive years since 2013
• Huashan Lunjian (华山论剑): Weekly King selection + playoffs
• King Tournament (棋王赛): Regularly held online tournaments
• TCT Finals: Early offline finals in Sanya
• Online Tournament Tools: Video AI notation recognition, online pairing, club tournaments, notation live streaming

25.3 Why Tian Tian Xiangqi Became the Absolute Dominator

Tian Tian Xiangqi’s ability to stand out among numerous Xiangqi platforms and maintain its absolute advantage over the long term is the result of the combined effect of the following factors:

I. Tencent Ecosystem’s “Dimensionality Reduction Strike”

This is Tian Tian Xiangqi’s most core and most irreproducible competitive moat. The combined WeChat + QQ user pool of over 2 billion MAU constitutes the largest social network in human history. Tian Tian Xiangqi converts this network advantage into product advantage through the following means:

1. Zero-Friction User Acquisition: Users need not download, register, or remember account passwords — clicking a Mini Program within WeChat starts playing immediately. Compared to the cumbersome process of the OurGame era (“download client → register account → login → enter room”), this experience improvement is revolutionary.

2. Social Fission: When users share a brilliant game notation in their Moments, friends who see it can join with one click. This propagation efficiency based on real social relationships far exceeds any advertising spend.

3. Network Effects: A platform’s value is proportional to the number of people using it. Because Tian Tian Xiangqi has the most users, it has the fastest matching speed, the most opponent choices, and the most active social interactions — these advantages in turn attract more users.

II. Mobile-First Product Design

Tian Tian Xiangqi was designed for mobile from day one, standing in stark contrast to PC-era platforms like OurGame and Yitian:

1. Touch Optimization: The operation flow of touching a piece → dragging to the target position → releasing to play is an interaction paradigm specifically designed for phone screens
2. Fragmented Time Adaptation: Blitz modes (5-10 minutes) adapt to fragmented time slots in mobile scenarios
3. Clean Interface: Information density on the phone screen is carefully designed; core functions are within easy reach
4. Push Notifications: Friend invitations, tournament reminders, and other push features maintain user engagement

III. Authority of the Strength Evaluation System

Tian Tian Xiangqi’s strength evaluation system has been accepted by the entire Chinese Xiangqi industry as the de facto strength standard:

1. Broad Coverage: Tens of millions of users play here; the rank distribution provides a precise frame of reference
2. High Stability: Continuously operating for over 10 years, the rating system’s credibility is time-tested
3. Industry Recognition: Player recruitment, training institution class placement, and even some offline tournaments reference Tian Tian Xiangqi ranks

“What’s your Tian Tian Xiangqi rank?” (你天天象棋几段?) has become the standard greeting among Chinese Xiangqi enthusiasts — a clear example of how a product can set an industry standard.

IV. Full-Chain Coverage

Tian Tian Xiangqi is not just a simple playing tool but an ecosystem platform covering the full chain of “Play → Learn → Social → Tournament → Certification”:

• Want to play: Various time controls and variant modes satisfy different needs
• Want to improve: From AI analysis to expert courses, from endgame challenges to notation collection
• Want to socialize: Friend battles, club systems, 10,000-spectator mode
• Want to compete: From Huashan Lunjian to the Tian Tian Xiangqi Cup, from online tournaments to the Xiangqi League
• Want certification: Online amateur rank certification, offline teaching locations

This full-chain coverage means users do not need to switch platforms to meet all their needs, greatly improving user stickiness.

V. Continuous Iteration and Technology for Good

Over nearly a decade of operation, Tian Tian Xiangqi has continuously launched multiple industry-leading features:

• Color-Blind Mode (2021): Adjusted red-black contrast, serving over 1 million color-blind players
• Accessibility Version (2023): Integrated screen-reader voice interaction, haptic feedback, and “sighted-blind same-server” matching, allowing visually impaired players and regular players to compete fairly in the same matchmaking pool
• Health Guardian (2018): Game duration reminder mechanism based on a fatigue mathematical model, originating from a complaint letter from the daughter of a lung cancer patient in recovery
• Brain-Computer Interface (2025): In collaboration with Yansi Brain Science Research Institute, allowed a cerebral palsy player to complete formal games using only their “mind” — the world’s first use of non-invasive brain-computer interface technology in a national-level sports event
• Anthropomorphic AI (2024): Supervised learning AI trained on hundreds of millions of human game records, responding in only 10 milliseconds

These innovations not only enhanced the product’s user value but also earned Tencent a positive corporate reputation. As Tencent co-founder Zhang Zhidong said: “The ‘long’ in long-term operation must start from respect for the long-term relationship with players.”

25.4 Tian Tian Xiangqi’s Commercialization and Controversies

Business Model:

Tian Tian Xiangqi’s revenue primarily comes from the following channels:

According to market research data, the Tian Tian Xiangqi App’s comprehensive revenue for 2025 was approximately 237-259 million RMB, a year-over-year increase of 10.7%.

Major Controversies:

1. Cheating Problem: This is the biggest controversy facing Tian Tian Xiangqi. Because Tian Tian Xiangqi’s strength evaluation has industry authority, cheating using Pikafish and other engines in evaluations is rampant. Actual tests showed that cheaters could climb from Amateur Level 8 to Level 8-2 with a 92.79% win rate without being caught by the detection system throughout the process. Reporting effectiveness was inconsistent; some players reported retaliation after reporting.

2. Excessive Commercialization: Strength evaluation requires consuming “copper coins”; players must participate in blitz games or top up to obtain copper coins. This has been criticized as “finding ways to trick money,” squeezing out slow-game thinking space. VIP members enjoy privileges such as priority report handling, criticized as “resources tilted toward paying players.”

3. Professional Competition Cheating Scandals: From the Xiangqi League to individual championships, there have been multiple scandals of players using software to cheat. Liu Dahua/Dang Fei exposed suspected cheating instructions used by players in the Xiangqi League, while the Cui Ge extortion case against players revealed the software cheating interest chain within professional circles.

25.5 Tian Tian Xiangqi’s User Scale and Industry Ecosystem

User Scale:

Industry Landscape:

The Chinese Xiangqi game platform market is highly concentrated, with CR3 (top 3 concentration) reaching 96.3%:

• Tian Tian Xiangqi: Largest user base (13.2 million MAU)
• Yehu Go (野狐围棋): Highest revenue (568 million RMB/year)
• Cangjingge (藏经阁, NetEase): Revenue 215 million RMB/year

Tian Tian Xiangqi leads by a wide margin in user scale, but due to Xiangqi users’ lower willingness to pay compared to Go users, its ARPU (Average Revenue Per User) is lower than Yehu Go.

25.6 The Historical Significance of Tian Tian Xiangqi

Tian Tian Xiangqi is not just a successful mobile internet product; it is also the core vehicle for the inheritance and development of Xiangqi in the digital age:

1. The Biggest Driver of Xiangqi Popularization: Tian Tian Xiangqi allows hundreds of millions of people to play chess anytime, anywhere. For a traditional board game with a thousand-year history, this level of popularization is unprecedented.

2. Established the Modern Xiangqi Rating System: Tian Tian Xiangqi’s rank has become the “universal language” and “strength yardstick” of the Xiangqi community.

3. Key Support for Professionalization: Title sponsorship of the Xiangqi League for 13 consecutive years, providing professional players with a stable tournament platform and income source.

4. Practitioner of Technology for Good: Innovations such as the accessibility version, color-blind mode, and brain-computer interface demonstrate how technology products can serve a wider range of user groups.

5. Digitalization of Xiangqi Education: 412 offline teaching locations combined with online courses and AI analysis form a complete education system covering online and offline.

Tian Tian Xiangqi’s success is, to some extent, Tencent’s “revenge” against OurGame from nearly twenty years ago in the chess game domain — in 2003, Tencent replicated OurGame’s model with QQ Games and defeated OurGame; in 2015, Tencent used Tian Tian Xiangqi to completely replace the online Xiangqi world that OurGame had pioneered on mobile. History came full circle.

---

Chapter 26 XQF Format and Electronic Notation Standardization

26.1 History of the XQF Format

Before the widespread adoption of the UCCI protocol and FEN standard, Xiangqi electronic notation formats were in a state of “everyone fighting for themselves.” Different engines and interfaces used different notation formats, greatly hindering data exchange.

The XQF format was designed in 1999 by Dong Shiwei (董世伟, online alias “过河象” or “River-Crossing Elephant”), the developer of XQStudio (象棋演播室). It was one of the earliest widely accepted Xiangqi electronic notation formats. Its features included:

• Binary tree-based move storage structure, supporting multiple variations recorded from the first move onward
• Support for both Chinese and English move representation
• Ability to record notation comments and variations
• Extensive software support on the Windows platform

Huang Chen provided XQF tools (XQFTOOLS) on xqbase.com, which could convert XQF format to universal formats like PGN. These tools played a role in promoting the standardization of Xiangqi electronic notation. XQStudio officially published the XQF 1.0 format specification in April 2001. Subsequently, the XQF format was supported by a large number of software and websites, including Xiangqi Bridge (象棋桥), Xiangqi Wizard (象棋巫师), and Dongping Xiangqi Net (dpxq.com, 东萍象棋网).

26.2 The Standardization Process

The transition from the XQF era to the FEN standard was an important step in Xiangqi data standardization. The UCCI protocol introduced FEN (Forsyth-Edwards Notation) to Xiangqi, and FEN quickly became the de facto standard because of its simplicity and text-based nature.

26.3 Comparison of Xiangqi Engine Protocols

The communication protocols of Xiangqi engines have evolved from proprietary protocols to unified standards. Below is a comparison of the major protocols:

Note: The UCCI protocol is one of Huang Chen’s most important contributions to the Xiangqi engine ecosystem. Before UCCI, each engine had its own communication protocol, and users had to use the matching interface — which ran counter to the philosophy of “interface and engine separation.” UCCI’s standardization allowed users to use any UCCI engine on any UCCI interface, greatly promoting competition and collaboration among engines. It is worth noting that while UCCI’s Chinese coordinate format is easier for Chinese users to understand than UCI’s international coordinates, it has formed an obstacle to international promotion. In recent years, there have been discussions about unifying UCCI and UCI into a single standard.

FEN format in Xiangqi:

• Uses lowercase letters for black pieces (r=Rook, n=Knight, b=Bishop, a=Advisor, k=King, c=Cannon, p=Pawn)
• Uses uppercase letters for red pieces (R=Rook, N=Knight, B=Bishop, A=Advisor, K=King, C=Cannon, P=Pawn)
• Uses digits for consecutive empty squares
• Uses “/” to separate rows

For example, the FEN representation of the starting position is:

rnbakabnr/9/1c5c1/p1p1p1p1p/9/9/P1P1P1P1P/1C5C1/9/RNBAKABNR w - - 0 1

The standardization of FEN enabled Xiangqi game data to flow freely between different engines, interfaces, and tools, laying the foundation for later cloud-based database systems.

---

Chapter 27 Xiangqi Engines on Linux and macOS

27.1 The Xiangqi Engine Ecosystem on Unix Platforms

Compared to the rich engine ecosystem on the Windows platform, the number and variety of Xiangqi engines on Linux and macOS are relatively limited. However, there are a few notable projects and tools worth highlighting:

GMChess: GMChess is an open-source engine based on the ElephantEye code architecture, successfully entered into the Debian Linux software repository. It is one of the best Xiangqi engines available to Linux users. Users can conveniently install it via the apt-get install gmchess command. GMChess has been available on Debian/Ubuntu Linux for many years with stable maintenance.

XQBoard: A Xiangqi interface on Unix platforms, consistent with the WinBoard/XBoard style, supporting standard UCCI protocol engines. XQBoard provides familiar operating methods for Linux users accustomed to the Unix/WinBoard style.

ElephantEye for Unix: Huang Chen’s ElephantEye was inherently cross-platform; its C++ source code can be recompiled on Linux and macOS. Tools in the ElephantEye suite (such as the LEAGUE simulator, XQFTOOLS, etc.) also support Unix platforms.

Pikafish’s Multi-Platform Support: Pikafish has built-in Makefile compilation rules for Linux and macOS. On Apple Silicon (M1/M2/M3) chips, Pikafish achieves efficient NNUE inference through the ARM NEON instruction set. Pikafish’s cross-platform support has earned it a broader audience in the development community.

27.2 Compiler Optimization and SIMD Instruction Sets

The performance of modern Xiangqi engines largely depends on the utilization efficiency of CPU instruction sets. Different CPU architectures support different SIMD instruction sets:

x86/AMD64 Platforms:

• SSE/SSE2/SSE3: Basic SIMD instruction sets, supported by all x86-64 CPUs
• SSSE3: Supplemental SSE instruction set, supported by most CPUs
• AVX2: 256-bit SIMD instruction set, supported by Intel Haswell (2013+) and AMD Ryzen, critical for NNUE inference
• BMI2: Bit manipulation instruction set, significantly accelerates bitboard operations
• AVX512: 512-bit SIMD instruction set, supported only by some high-end Intel/AMD CPUs; Pikafish’s 2025 version introduced AVX512 optimizations

ARM Platforms:

• NEON: SIMD instruction set for ARM platforms, supported by all 64-bit ARM CPUs (including Apple Silicon)
• SVE: ARM’s scalable vector extension SIMD instruction set, not yet widely applied in Xiangqi engines

Impact of Instruction Set Selection on Playing Strength:

In Pikafish, the performance difference between different instruction set versions can reach 10-30%:

• SSE3 version: Minimum compatibility, performance baseline
• AVX2 version: Recommended version, significantly improved NNUE inference speed
• BMI2 version: Bitboard operation acceleration
• AVX512 version: Latest version, optimal performance on hardware supporting AVX512

Choosing the Right Compiled Version:

Users should select the most appropriate compiled version based on their hardware. Generally, Intel CPUs from 2015 onwards and AMD CPUs from 2017 onwards support AVX2. When choosing a compiled version, you can confirm supported instruction sets by checking CPU information:

• Intel: Use lscpu to view Flags; if avx2 and bmi2 are present, they are supported
• AMD: All AMD Ryzen series support AVX2
• ARM: Apple M1/M2/M3 and most ARM v8-A chips support NEON

---

Chapter 28 Performance Measurement and Benchmarking of Xiangqi Engines

28.1 NPS (Nodes Per Second) and Its Significance

NPS (Nodes Per Second) is one of the most commonly used engine performance metrics. It reflects how many game tree nodes the engine can search per unit of time.

Typical NPS ranges for different engines and hardware conditions:

It should be noted that NPS is not a direct measure of playing strength — different engines have different search algorithms, and the “quality” of each node differs. An engine with a higher pruning rate may achieve equivalent playing strength at a lower NPS. Therefore, NPS is more suitable as a comparison of optimization effectiveness between different versions of the same engine, rather than a cross-engine comparison metric.

28.2 ELO Rating: The Dilemma of Xiangqi’s Lack of a Unified System

A core fact related to the development of Xiangqi engines is: Xiangqi lacks a unified, widely accepted ELO rating system.

Unlike international chess, Xiangqi does not have a unified system like the FIDE ELO rating. This absence brings several problems:

1. Difficulty in Quantifying Engine Strength: There is no authoritative rating leaderboard for developers and users to reference.
2. Difficulty in Cross-Generational Comparison: It is impossible to accurately compare the strength differences of engines from different eras.
3. Lack of Academic Standards: It is difficult to precisely and quantitatively describe an engine’s playing strength in academic papers.

Informal Rating Systems in the Community:

Despite the lack of an official system, there are some unofficial rating references in the Xiangqi community:

1. Yitian Rating: A rating system based on game results on the Yitian platform. Although not precise, it provides a relative reference benchmark.
2. Engine vs Engine Testing: Community testers establish unofficial ELO rankings through thousands of engine vs engine games.
3. Comparison with Commercial Engines: Testing new engines against commercial engines of known strength to estimate their strength range.

28.3 Testing Methodology: From Experience to Science

The testing methodology for Xiangqi engines has evolved from “personal experience” to “statistical science”:

Personal Experience Testing Period (1990s-2000s):

• Authors ran a few dozen test games themselves
• Test results heavily influenced by random factors
• Test conditions not standardized (hardware, time, opening books all varied)

Semi-Standardized Testing Period (2000s-2010s):

• Community testers ran hundreds of test games
• Began using standardized time controls and opening settings
• But data recording and report formats were still inconsistent

Statistical Testing Period (2020s-Present):

• Pikafish introduced the Fishtest framework
• Test scale reached thousands to tens of thousands of games
• Using SPRT for statistical significance testing
• Test results are publicly transparent and reproducible

The advancement of testing methodology transformed engine development from “tuning parameters by feel” into “science driven by data” — every improvement must pass statistical significance testing before being merged.

---

Volume V Future Outlook

---

Chapter 29 Technical Frontiers of Xiangqi Engines

29.1 Transformer Architecture and Chess Evaluation

In recent years, the Transformer architecture has achieved great success in natural language processing and computer vision. This “self-attention” mechanism may bring new breakthroughs to chess evaluation:

• Global Relationship Modeling: Transformers can capture interactions between distant pieces on the board, which is especially important for Xiangqi (e.g., the long-range attack of Cannons, the long control lines of Rooks).
• Dynamic Weights: Unlike static convolutional networks, Transformers can dynamically adjust attention weights based on board content.
• Multi-Modal Fusion: Future evaluation networks may simultaneously process board position information and positional semantic information.

However, the computational cost of Transformers is far greater than NNUE’s fully connected networks. How to efficiently adapt them to real-time inference scenarios in chess search remains an open question. Some research directions include:

• Lightweight Transformer architectures
• Knowledge distillation — transferring Transformer evaluation knowledge to lighter networks
• Hybrid architectures — using different evaluation networks at different stages of the search

29.2 Reinforcement Learning and Self-Play

AlphaZero demonstrated the powerful potential of “learning from scratch” in the chess domain. In the Xiangqi domain, the application of reinforcement learning is still in its early stages:

• GGzero (Jiajia Xiangqi’s reinforcement learning branch) is currently one of the most successful Xiangqi reinforcement learning engines.
• cchess-zero and CCZero are community-driven reinforcement learning projects.
• The main bottleneck is that deep reinforcement learning requires massive GPU computing power (typically hundreds to thousands of GPU-days), and the Xiangqi community lacks computing resources similar to Google or DeepMind.

In the future, with the broader adoption of cloud computing and distributed computing, Xiangqi reinforcement learning engines may achieve breakthrough progress. Some potential breakthrough directions include:

• Using cloud GPU resources for distributed training
• Transfer learning — using network weights from international chess reinforcement learning training as a starting point for Xiangqi training
• Hybrid methods combining NNUE — using NNUE evaluation in search and reinforcement learning in training

29.3 Quantum Computing and Chess Game-Playing

The application of quantum computing in chess search is a frontier but not yet mature research direction. Grover’s search algorithm can theoretically complete unstructured search at square root complexity, which may have potential value for quantum acceleration of Minimax.

However, quantum computers are still in their early stages, with a considerable way to go before practical application. This field may achieve breakthroughs in the next 10-20 years, but will not have a substantial impact on Xiangqi engines in the short term.

29.4 Community-Driven Future: Development Directions of the Open Source Ecosystem

Looking ahead, the open source ecosystem of Xiangqi engines may develop in the following directions:

More Automated:

• Automated testing and regression frameworks will be further refined
• Continuous Integration and Continuous Deployment (CI/CD) processes will become normalized
• Automated network training pipelines
• Automated opening book updates

More Open:

• More data and models released under open licenses
• More documentation and tutorials for newcomers
• More active community discussions and collaboration
• Cross-language and cross-cultural cooperation

More Professional:

• Code quality and toolchain usage will become more professional
• Project management and decision-making processes will become more standardized
• Greater crossover with academia and other open source communities
• Introduction of software engineering best practices (code review, continuous testing, release management)

29.5 From Engine to Platform: Ecosystem Integration

Future Xiangqi engines may no longer be merely “chess-playing programs” but will evolve into integrated platforms. Pikafish has already demonstrated this trend — it is not just a Xiangqi engine but also includes a testing framework (Fishtest), data analysis tools, network training pipelines, and other components.

This evolution “from engine to platform” will make the Xiangqi computer game-playing ecosystem richer and more complete:

• Development Platform: Providing an integrated environment for engine development, testing, and deployment
• Data Platform: Providing game notation databases, cloud library queries, and data analysis services
• Education Platform: Providing engine analysis, position training, and strength testing functions
• Research Platform: Providing standardized testing tools and academic research support

---

Volume VI Conclusion

---

Chapter 30 The Cultural Legacy of Computer Xiangqi Game-Playing

30.1 The Inheritance of History

From Jiangzu/Elite’s first championship in London in 1989, to Pikafish’s continued leadership of open source engines in 2026, computer Xiangqi game-playing has traversed nearly four decades. In these forty years, we have witnessed:

• The Evolution of Technology: From dozens of lines of assembly code to millions of lines of C++ neural network systems
• The Leap in Playing Strength: From amateur-beginner level one-step-at-a-time calculation to precision computation surpassing human limits
• The Maturation of the Ecosystem: From lone-wolf solitary programmers to globally collaborative open source communities
• The Accumulation of Data: From scattered manual game notations to tens of terabytes of cloud-based endgame databases
• The Establishment of Standards: From fragmented data formats to the unified UCCI protocol and FEN standard

30.2 Unrecorded History

In the process of compiling this report, one problem we keenly feel is: much of the history of computer Xiangqi game-playing is disappearing.

• Many valuable posts from the early Qizhong Forum were lost during website redesigns
• The official websites of several early engines are no longer accessible
• Contact information for some authors is no longer valid
• The “why” behind many technical decisions can no longer be verified
• Many game notations and test data from various engines have been lost
• Sales data and user feedback for early commercial engines can no longer be traced

This report is a deliberate recording and preliminary compilation of this history. But we are well aware that this report is only a starting point. If future researchers can find more primary sources — including author interviews, forum snapshots, source code history, and tournament records — they will be able to paint an even more complete picture.

30.3 From “Chess Software” to “Civilization”: The Birth of a Knowledge Community

When we examine the development of Xiangqi engines from a broader historical perspective, we discover: this is not just a story about chess software, but a story about the birth and growth of a knowledge community.

This community is composed of the following elements:

• Knowledge: From Shannon’s Minimax to NNUE neural networks — theoretical accumulation spanning more than half a century
• Tools: From DOS to cloud computing, from floppy disks to GitHub — continuously evolving development and collaboration tools
• Rules: From the UCCI protocol to the Fishtest testing standard — technical norms collectively followed by the community
• Culture: From Qizhong Forum posts to GitHub Pull Requests — a community culture of knowledge sharing and collaboration
• Values: Openness, sharing, and the pursuit of excellence — these values propel the community ever forward

This community proves one thing: A group of people scattered across the globe, speaking different languages, from different backgrounds, can use the internet and open source tools to collectively create a knowledge product that surpasses the capability of any individual or institution.

This is the most valuable cultural heritage that computer Xiangqi game-playing leaves for the future.

---

Appendices