國立中正大學課程大綱
115 學年第 1 學期　教育學院運動與休閒教育研究所－碩士班
課程名稱(中文) 運動情報蒐集與大數據分析實務研究
課程名稱(英文) Practical Intelligence Sport & Big Data Analytics Research
授課教師 林晉榮 博士/教授
開課時段 星期一下午 14:10-16:00（2 學分／每週 2 小時）
先修科目或先備能力 無（不需程式設計基礎）
課程概述 本課程以體育運動為主體，訓練學生運用AI代理（Claude Code）完成運動情報的蒐集、清理、分析與呈現。著重在於把研究問題轉換為可執行的任務、指揮 AI 完成資料工作，並查證其產出是否可信。全學期以一套運動情報系統為貫穿專案，逐週累積為可重現、可交付的研究成果，培養學生在 AI 時代進行運動科學研究與產業應用的實作能力。
學習目標 1. 將運動場域的研究問題轉換為可執行的分析任務。
2. 指揮 AI 代理完成資料蒐集、清理、入庫與分析。
3. 查證產出品質，辨識幻覺、資料洩漏與過度樂觀的驗證方式。
4. 判斷資料授權、個人資料與研究倫理風險。
5. 交付可重現的情報報告，並誠實揭露 AI 參與程度。
教科書 主教科書：Claude Code Vibe Coding 開發手冊 第二版（施威銘研究室／旗標）
參 考 書：Vibe Coding CLI 頂級開發－Claude Code 前瞻菁英育成手冊（胡嘉璽／深智）
　　　　　Python for Data Analysis（Wes McKinney）
　　　　　The Signal and the Noise（Nate Silver）
自編講義：運動情報蒐集與大數據分析（16 週）
影音教材：Claude Code 完整教學－新手必學的 15 個核心觀念
　　　　　https://youtu.be/f10jf_PQU2E
（請尊重智慧財產權，不得非法影印教師指定之教科書籍）

教學要點概述
教材編選 ■自編教材　■教科書作者提供
教學方法 ■投影片講述　■上機實作　■專題討論
評量方法 ■上課點名 15%　■作　業 15%　程式實作 %
■專　　案 30%　其它（學習日誌）0%
□小考 0%　■期中考 10%　■期末考 30%　□實習報告 0%
教學資源 ■課程網站　■教材電子檔供下載　■線上影音教材
教學相關配合事項 學生自備筆記型電腦；開學前辦理 2 小時環境安裝日；上機採 2 人一組。
課程進度
260914　W1 課程導論：運動情報與 AI 代理（CH1-1、1-3）｜實作：認識參考專案與資料流盤點
260921　W2 安裝、啟動與執行環境（CH1-2、1-4、2-2）｜實作：跑通管線、專案初始化與存檔點
260928　W3 與檔案協作（CH1-3、2-3）｜實作：運動資料生態盤點與合規
261005　W4 AI 代理使用心態（CH8-1、4-3）｜實作：抽樣查證與責任歸屬
261012　W5 Prompting：PRD 與 plan 模式（CH2-3、9-1）｜實作：資料擷取與合規檢核
261019　W6 Token、成本與模型選擇（CH3-1、3-2）｜實作：擷取成本估算與大量取得
261026　W7 Context 管理與 Subagent 分工（CH3-1、4-1~4-2）｜實作：資料清洗與實體解析
261102　W8 CLAUDE.md 專案規範與 /init（CH2-1、2-2）｜實作：資料入庫與資料字典（期中提案）
261109　W9 設定檔範圍與分層規範架構（CH1-3、2-4）｜實作：運動指標設計
261116　W10 權限模式與 Hook 防線（CH2-4、3-2、6-1~6-2）｜實作：統計陷阱辨識
261123　W11 隱私與資料流向（CH6-3；主教材無專章，見講義）｜實作：預測模型建立與驗證
261130　W12 API Key 與 .env 安全、Hook 進階（CH6-4~6-5）｜實作：資料洩漏獵捕
261207　W13 MCP 與遠端協作（CH5-1~5-3、7-1~7-4）｜實作：視覺化與情報報告
261214　W14 Skills 封裝與任務排程（CH5-4、8-5、3-3）｜實作：可重現性與一鍵重跑
261221　W15 實戰範例研讀與專題工作坊（CH8、CH9）｜交叉重現審查
261228　W16 期末專題發表
講授內容
本課程採雙軸並行：每週上半場依主教材章節講授、同學專章報告，下半場以同一貫穿專案進行運動情報實作，逐週累積為完整的分析系統。
主教材軸 PART I　入門與基礎（CH1-CH3，W1-W6）：Claude Code 簡介與安裝、內建工具、對話紀錄與自動記憶、CLAUDE.md 專案規範、權限模式與 /checkpoint、工作樹與 plan 模式、常用斜線命令、Token 與成本管理。
主教材軸 PART II　進階能力（CH4-CH6，W7-W12）：Subagent 分工合作與 context 管理、MCP 與 Skills、Claude in Chrome 與 Playwright 自動化測試、Hook 防線（阻擋修改、通知、JSON 介面與精準條件）。
主教材軸 PART III　協作與實戰（CH7-CH9，W13-W15）：Claude Code Desktop、Remote Control、雲端 Agent 與 Telegram 協作；數獨遊戲與 YouTube 轉投影片兩個實戰範例，作為期末專題的方法示範。
實作軸（運動情報系統）：資料生態盤點與合規判斷、API 與網頁擷取、資料清洗與實體解析、資料庫與資料字典、運動指標設計與效度論證、統計陷阱辨識、預測模型的時間切分驗證與基準比較、資料洩漏獵捕、視覺化與情報報告、可重現性與交付。
影音教材之 15 個核心觀念與各週主題對應，時間碼載於各週講義，供課前自習使用。
核心能力
1.情報問題界定能力：能將教練、球團、媒體或產業的提問，轉為有明確對象、指標與時效的可分析問題。
2.資料蒐集與整備能力：能指揮 AI 代理完成多來源運動資料的擷取、清理、實體解析與入庫，並產出資料品質報告與資料字典。
3.分析與驗證能力：能設計運動指標並論證其效度，建立預測模型並以時間切分、基準比較與校準檢查驗證其可信度。
4.品質把關能力：能查證 AI 產出的正確性，辨識幻覺、資料洩漏、過度宣稱與統計誤用，不以「跑得出來」作為正確的依據。
5.研究倫理與交付能力：能判斷資料授權與個人資料風險，產出可被他人重現的分析流程，並誠實揭露 AI 參與程度與研究限制。

National Chung Cheng University Course Syllabus
Academic Year 2026, Semester 1
College of Education, Graduate Institute of Sports and Leisure Education — Master's Program
Course Title (English) Practical Intelligence Sport & Big Data Analytics Research
Instructor Lin Chin-Jung, Ph.D. / Professor
Class Time Monday 14:10–16:00 (2 credits / 2 hours per week)
Prerequisites None (no programming background required)
Course Description Centered on sport and physical activity, this course trains students to use an AI agent (Claude Code) to collect, clean, analyze, and present sports intelligence. The emphasis is on translating research questions into executable tasks, directing the AI to carry out the data work, and verifying whether its output can be trusted. Throughout the semester, students build a single sports-intelligence system as a course-long project, accumulating week by week into reproducible, deliverable research results, and developing the hands-on capability to conduct sports-science research and industry applications in the AI era.
Learning Objectives 1. Translate research questions from sport settings into executable analytical tasks.
2. Direct an AI agent to complete data collection, cleaning, database loading, and analysis.
3. Verify output quality: recognize hallucinations, data leakage, and overly optimistic validation.
4. Assess risks concerning data licensing, personal data, and research ethics.
5. Deliver reproducible intelligence reports and honestly disclose the extent of AI involvement.
Textbooks Main textbook: Claude Code Vibe Coding Development Handbook, 2nd ed. (Flag's R&D Lab / Flag Technology, 旗標)
References: Vibe Coding CLI Top-Tier Development: Claude Code Elite Training Handbook (Hu Chia-Hsi / 深智)
Python for Data Analysis (Wes McKinney)
The Signal and the Noise (Nate Silver)
Course reader: Sports Intelligence Collection and Big Data Analytics (16 weekly handouts, instructor-compiled)
Video material: Claude Code Complete Tutorial — 15 Core Concepts for Beginners
https://youtu.be/f10jf_PQU2E
(Please respect intellectual property rights. Unauthorized photocopying of the designated textbooks is prohibited.)

Teaching Overview
Teaching Materials ■ Instructor-compiled materials ■ Publisher-provided materials
Teaching Methods ■ Slide lectures ■ Hands-on lab work ■ Project-based discussion
Evaluation ■ Attendance 15% ■ Assignments 15% Programming practice ___%
■ Project 30% Other (learning journal) 0%
□ Quizzes 0% ■ Midterm exam 10% ■ Final exam 30% □ Internship report 0%
Teaching Resources ■ Course website ■ Downloadable e-materials ■ Online video materials
Course Logistics Students bring their own laptops; a 2-hour environment-setup session is held before the semester begins; lab work is done in pairs.
Course Schedule
260914 W1 Course introduction: sports intelligence and AI agents (CH1-1, 1-3) | Lab: exploring the reference project and mapping the data flow
260921 W2 Installation, startup, and execution environments (CH1-2, 1-4, 2-2) | Lab: running the reference pipeline; project initialization and checkpoints
260928 W3 Working with files (CH1-3, 2-3) | Lab: surveying the sports-data ecosystem and compliance
261005 W4 The right mindset for working with AI agents (CH8-1, 4-3) | Lab: spot-check verification and accountability
261012 W5 Prompting: PRD and plan mode (CH2-3, 9-1) | Lab: data extraction and compliance checks
261019 W6 Tokens, cost, and model selection (CH3-1, 3-2) | Lab: extraction cost estimation and bulk retrieval
261026 W7 Context management and subagent teamwork (CH3-1, 4-1–4-2) | Lab: data cleaning and entity resolution
261102 W8 CLAUDE.md project rules and /init (CH2-1, 2-2) | Lab: database loading and the data dictionary (midterm proposal)
261109 W9 Settings scopes and layered configuration (CH1-3, 2-4) | Lab: sports metric design
261116 W10 Permission modes and Hook safeguards (CH2-4, 3-2, 6-1–6-2) | Lab: recognizing statistical pitfalls
261123 W11 Privacy and data flows (CH6-3; no dedicated chapter — see handout) | Lab: building and validating prediction models
261130 W12 API-key and .env security; advanced Hooks (CH6-4–6-5) | Lab: hunting data leakage
261207 W13 MCP and remote collaboration (CH5-1–5-3, 7-1–7-4) | Lab: visualization and intelligence reporting
261214 W14 Packaging Skills and task scheduling (CH5-4, 8-5, 3-3) | Lab: reproducibility and one-command reruns
261221 W15 Case-study chapters and project workshop (CH8, CH9) | Cross-team reproduction review
261228 W16 Final project presentations
Course Content
The course runs on two parallel tracks. In the first half of each week's session, lectures follow the main textbook chapter by chapter, supplemented by student chapter presentations; in the second half, students advance a single course-long sports-intelligence project, accumulating week by week into a complete analytical system.
Textbook track, PART I — Fundamentals (CH1–CH3, W1–W6): introduction to and installation of Claude Code; built-in tools; conversation history and auto memory; CLAUDE.md project rules; permission modes and /checkpoint; worktrees and plan mode; common slash commands; token and cost management.
Textbook track, PART II — Advanced Capabilities (CH4–CH6, W7–W12): subagent teamwork and context management; MCP and Skills; Claude in Chrome and automated testing with Playwright; Hook safeguards (blocking edits, notifications, the JSON interface, and precise conditions).
Textbook track, PART III — Collaboration and Case Studies (CH7–CH9, W13–W15): Claude Code Desktop, Remote Control, cloud agents, and Telegram collaboration; two end-to-end case studies (a Sudoku game and a YouTube-video-to-slides converter) as methodological models for the final project.
Project track (the sports-intelligence system): surveying data sources and compliance assessment; API and web extraction; data cleaning and entity resolution; database design and the data dictionary; sports metric design and validity argumentation; recognizing statistical pitfalls; prediction models validated with time-based splits, baseline comparison, and calibration checks; hunting data leakage; visualization and intelligence reporting; reproducibility and delivery.
The 15 core concepts in the video material map onto the weekly topics; timestamps are listed in each weekly handout for pre-class self-study.
Core Competencies
1. Intelligence problem framing: turn questions from coaches, clubs, media, or industry into analyzable problems with a clear audience, metrics, and timeliness.
2. Data collection and preparation: direct an AI agent to extract, clean, entity-resolve, and load multi-source sports data, producing data-quality reports and a data dictionary.
3. Analysis and validation: design sports metrics and argue their validity; build prediction models and establish their credibility through time-based splits, baseline comparison, and calibration checks.
4. Quality assurance: verify the correctness of AI output; recognize hallucinations, data leakage, overclaiming, and statistical misuse; never treat “it runs” as proof of correctness.
5. Research ethics and delivery: assess data-licensing and personal-data risks, produce analytical pipelines that others can reproduce, and honestly disclose the extent of AI involvement and the limitations of the research.