reddit · SideProject · 10-03 18:50 55 分 ← 返回列表

A memory researcher's side project: finding the recurring stories and identity in your chess (no LLMs)

When you remember a loss, do you remember one bad move or a slow slide, and which one actually happened? Are your losing streaks one story told over and over, or a different story every time? When does a story you keep telling about your games stop being a story and start being your identity, like "I'm an attacker" or "I always find a way back"? These are questions about autobiographical memory, which is what I studied for years while getting a PhD in cognitive psychology. A

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：棋手复盘时只看到引擎评分和"失误标注"，无法看清自己反复犯的是同一类叙事性错误（如总是慢速崩盘、总是同一种输法），也难以把单局教训和长期棋风身份联系起来。

目标用户：有自我复盘习惯的成人进阶棋手（chess.com / lichess 付费用户）、教练与棋类内容创作者。

证据："In everyday life you can rarely check someone's story against what actually happened. In chess, every move is recorded, so you can." / "Single-game reviews show where a game actually turned, whether it was one bad move or a slow slide, and connect your past games to the one in front of you."（原帖未出现任何付费用户、收入或他人求购的原话）

变现：订阅制：免费单局叙事复盘，付费解锁多局长期模式报告（月付 5–10 美元）；或按份出售"棋风身份报告"，面向教练做批量授权。

MVP：用 python-chess 解析用户上传的 PGN，调用本地 Stockfish 计算每步评估差，做两件事：(1) 单局转折点检测，区分"一步崩"和"慢速滑坡"；(2) 多局聚类，找出重复出现的失误模式（开局阶段、时间压力、残局转换），输出一页可视化叙事报告，2 周内可上线。

竞品：Chess.com Game Review、Lichess Insights、Aimchess、DecodeChess、Chessable / ChessMood

摘要：一位认知心理学博士（研究自传体记忆）用 Stockfish 搭了不带 LLM 的 StoryArc 引擎，从叙事角度复盘国际象棋：判断一局是"一步崩"还是"慢速滑坡"，并把多局串联成反复出现的棋风故事。这是个人 side project 展示，帖中没有任何收入、付费用户或他人求购的证据，但"棋手为深度复盘付费"的相邻市场已被 chess.com、Aimchess 等验证，差异化在叙事/心理视角而非引擎算力。