Reversi AI Research Project — Meeting Slides (Bilingual)¶

Note: English first, followed by Chinese translation for each bullet. Use as slide notes directly.

1) Project Title & Ethical Compliance｜项目标题与伦理合规¶

Reversi AI Algorithm Benchmarking & Optimization Research — 黑白棋AI算法基准与优化研究
Student: , , BSc Computer Science — 学生：<姓名>，<学号>，计算机科学本科
Supervisor meeting: Introductory overview (30 mins) — 与导师初次会面：项目概览（30分钟）
Statement of Ethical Compliance: Data Category B; Participant Category 2; will follow COMP390 ethical guidance — 伦理合规声明：数据类别B；参与者类别2；遵循COMP390伦理指南

Visual suggestion: Minimal title slide; small footer with ethics line. — 视觉建议：极简标题页，页脚写伦理声明。

2) Why Reversi? Motivation & Fit｜为何选择黑白棋：动机与契合¶

Simple rules and no hidden info; a solid AI test — 规则简单、无隐藏信息；适合做AI测试
Right size: interesting to study, small enough to finish — 规模合适：有研究性，也能按时做完
Fits C++ well for fast bitboards and clear structure — 很适合用C++做位板加速和清晰架构
Fits COMP390: software + analysis + good process — 符合COMP390：软件+分析+良好过程

Visual suggestion: One mini-board sketch + 3 bullet icons. — 视觉建议：简化棋盘草图+3个要点图标。

3) Project Overview & Objectives｜项目概览与目标¶

Distributed‑first: online multiplayer that just works (rooms, matchmaking, fair play) — 分布式优先：可用的在线对战（房间、匹配、公平对局）
Works on real networks: smooth latency, reconnect, save results — 面向真实网络：延迟平滑、可重连、保存结果
Then improve AI step by step; keep one clean interface for later methods — 然后逐步增强AI；保留统一接口便于以后接入方法
Clear modules: networking and AI are separate; upgrade one without breaking the other — 模块清晰：联网与AI分离；升级一边不影响另一边
Later show progress with small tests and simple charts — 之后用小测试和简明图表展示进展

Visual suggestion: Networking/Multiplayer block on top, AI block beneath. — 视觉建议：上方“联网/多人”，下方“AI”模块。

4) Tech Stack & Tools (Why Pure C++ vs Engine)｜技术栈与工具（为何纯C++而非引擎）¶

Core approach: Modern C++ + CMake; lightweight UI — 核心做法：现代C++ + CMake；轻量UI
Why pure C++: performance & control; simple packaging; easy to reproduce — 选择纯C++：性能与可控性；打包简单；结果易复现
Trade‑off: engines have stronger UI/UX; we prioritise netcode + AI now — 取舍：引擎UI/UX更强；当前优先网络与AI
Networking plan: add asio/boost::asio when needed, no engine lock‑in — 联网方案：需要时加入asio/boost::asio，避免引擎绑定

Visual suggestion: One‑line contrast (Pure C++ now → engine optional later). — 视觉建议：一行对比（现用纯C++ → 将来可选引擎）。

5) High-level System Overview｜系统高层概览¶

Layers: Algorithms / Optimizations / Evaluation / UI / Visualization — 分层：算法/优化/评测/UI/可视化
Clear interfaces: one AI interface; easy to swap implementations — 清晰接口：一个AI接口，便于替换实现
Simple events: UI listens to changes; background search does not block — 简单事件：UI监听变化；后台搜索不阻塞
Core bitboard Board with clean separation of concerns — 位板Board为核心，关注点分离清晰

Visual suggestion: Layered architecture diagram. — 视觉建议：分层架构图。

6) Direction Options: Distributed vs AI vs Both｜方向选择：分布式 / AI / 双线¶

Option A: Distributed search (different computers) — 方案A：分布式搜索（多机）
Pros: scalability research, networking/OS skills — 优点：可扩展性研究、网络/系统技能
Cons: higher complexity, debugging risk — 缺点：复杂度高、调试风险大
Option B: AI-first depth (algorithms + optimizations) — 方案B：AI优先（算法+优化）
Pros: stronger AI quality, clearer scope — 优点：AI质量更强、范围更清晰
Cons: fewer distributed systems elements — 缺点：分布式要素较少
Preferred: Distributed-first; add AI as staged enhancement — 倾向：分布式优先，AI分阶段增强

Visual suggestion: 3-column pros/cons matrix; highlight preferred. — 视觉建议：三列利弊矩阵，高亮首选。

7) AI Basics & Plan｜AI基础与计划¶

Start simple: basic rules to choose moves — 先从简单规则入手
Cut bad branches early to speed up search — 提前剪掉不佳分支，加快搜索
Try random trials when helpful to explore — 需要时用随机试走，探索更多可能
Use bitboards to make move checks fast — 用位板让走法判定更快
Keep one AI interface so we can swap methods later — 保持统一AI接口，便于后续替换方法

Visual suggestion: Simple flow from input position → choose move (with quick checks). — 视觉建议：从局面到选步的简易流程图（含快速检查）。

8) Speed-ups & Quality｜加速与质量¶

Go deeper step by step with time limit — 按时间限额逐步加深
Remember past positions to avoid repeat work — 记住已见局面，减少重复计算
Check good moves first to save time — 先检查更可能的好棋，节省时间
Add simple multi-thread later if needed — 需要时再加简单多线程
Measure basics: speed, depth, memory; keep small tests — 量基本指标：速度、深度、内存；保持小测试

Visual suggestion: Simple pipeline with 3 speed-up notes. — 视觉建议：简易流程图，标注3处加速点。

9) Compressed Roadmap & Milestones｜压缩路线图与里程碑¶

M1 (Week 1–2): Setup & plan — 建仓与计划
Repo, build, simple local play — 仓库/构建/本地对战
Goal: build passes; local play works — 目标：构建通过；本地对弈可用
Check: start, move, no crash; README + task list — 验收：能启动、能落子、不崩溃；README与任务清单
M2 (Week 3–5): Online basics — 在线基础
Rooms & matchmaking; send/receive moves — 房间与匹配；收发走子
Goal: online play loop works — 目标：联网对弈闭环打通
Check: two PCs in one room finish a game — 验收：两台电脑同房完成一局
M3 (Week 6–8): Network polish — 联网打磨
Smooth latency, reconnect, save results — 延迟平滑、重连、保存结果
Goal: playable on real networks — 目标：真实网络下可玩
Check: 100–200ms ok; resume same game; view records — 验收：100–200ms可玩；可回到同局；可查看记录
M4 (Week 9–11): AI v0 — AI初版
Simple rules; one clean AI interface — 简单规则；统一AI接口
Goal: basic AI via one interface — 目标：通过统一接口接入基础AI
Check: human vs AI works; netcode unchanged — 验收：人机可用；联网代码无需改动
M5 (Week 12–15): Tests & small charts — 测试与小图表
Speed/depth charts; multiplayer stats — 速度/深度图；多人统计
Goal: make small, repeatable charts — 目标：产出可重复的小图表
Check: fixed script outputs speed/depth; basic multiplayer counts — 验收：固定脚本输出速度/深度；多人对局数量
M6 (Week 16–19): AI v1 — AI进阶
Faster search; remember positions — 更快搜索；记住局面
Goal: faster/steadier AI — 目标：AI更快/更稳
Check: deeper or faster at same time budget — 验收：同等时限下更深或更快
M7 (Week 20–23, optional): Distributed add‑ons — 可选分布式增强
Spectators or simple ranking — 旁观或简单排行
Goal: one nice add‑on — 目标：完成一个增强点
Check: spectators or ranking works — 验收：旁观或排行可用
M8 (Week 24–30): Video & dissertation — 视频与论文
Goal: video + paper draft — 目标：演示视频与论文稿
Check: ≤10 min video; clear charts; paper structure ready — 验收：视频≤10分钟；图表清晰；论文结构就绪

Visual suggestion: Single line timeline with 6–8 markers. — 视觉建议：单线时间轴6–8节点。

10) Discussion & Next Steps｜讨论与后续¶

Questions: OK to go distributed‑first? how simple can AI v0 be? first demo content — 待讨论：是否认可分布式优先？AI初版多简单？首个演示内容
Risks & plan B: time, network bugs; fallback to offline play; AI can stay simple — 风险与备选：时间与网络问题；可退回本地对战；AI先保持简单
Next steps: lock MVP (online rooms + basic AI), set demo date, weekly goals — 下一步：锁定MVP（在线房间+基础AI）、定演示日期、订周目标
Delivery: small demos often; keep repeatable tests and charts — 交付：小步快演示；保留可重复的小测试与图表

Visual suggestion: Simple checklist; space for notes. — 视觉建议：简易清单，留白做笔记。

Appendix (optional for printout)｜附录（打印可选） - Glossary for non-experts (Reversi basics, bitboards) — 非专家术语表（黑白棋基础、位板） - Reference to COMP390 criteria & ethics link — COMP390评分标准与伦理链接提示