Title: 从个人训练到竞技产业，AI 把"运动过程"变成可计算资产——哪里被吃透，哪里是硬骨头 URL Source: https://sport.rushos.com/ Markdown Content: 市场报告口径差异很大：Grand View Research 估 $10.6B(2025)→$49.9B(2033)，WSC Sports 引用口径为 $8.9B(2024)→$27.6B(2030)。应看结构和方向，不宜机械相信单一数字。**「用户觉得有帮助」不是训练效果**：某运动社交平台的 AI 助手已从测试版转正、覆盖跑步与骑行并纳入虚拟活动与功率分析，官方称**约 80% 用户评估其「非常有帮助」**——这是**满意度自评，测的是体验不是适应**；同侧的数字力量训练机**每秒采集约 500 个数据点**，由其智能层分析表现与疲劳并生成每日计划，厂商称方法源自内部教练学、宣称「最个性化最有效」，**但缺第三方验证**。**传感器密度与满意度都在涨，而「执行层」——你到底做没做完那组动作——仍然没有被自动化。**Estimates diverge sharply: Grand View Research puts it at $10.6B (2025) → $49.9B (2033); WSC Sports cites $8.9B (2024) → $27.6B (2030). Read the structure and direction — don't take any single number literally. **«Users find it helpful» is not a training effect**: one activity platform has moved its AI assistant from beta to general availability across running and cycling, adding virtual activities and power analysis, and reports that **about 80% of users rate it «very helpful»** — a **satisfaction self-rating that measures experience, not adaptation**. On the same side, a digital strength-training machine captures **around 500 data points per second**, with its intelligence layer analysing performance and fatigue to generate a daily plan, credited to in-house coaching methodology and claimed to be «the most personalised, most effective», **without third-party validation**. **Sensor density and satisfaction both keep rising, while the execution layer — whether you actually finished the set — remains un-automated.** 版图 A Map A 个人训练闭环 · 从发现到分享 The personal training loop · from discovery to sharing 面向普通运动用户、家长、私教、运动社区。钱在降低找教练成本、提高反馈频率、让进步可视化、让运动可分享。九个阶段按训练干预时间顺序排列。For everyday exercisers, parents, personal trainers and communities. The money is in cutting coaching cost, raising feedback frequency, making progress visible and sport shareable. Nine stages, ordered along the training-intervention timeline. 01 Before · Discover & Book Before · Discover & Book 发现与预订 · 找场地 / 找教练 / 找搭子 Discovery & booking · courts, coaches, partners 运动的真正起点，也是原版漏掉的漏斗顶层——先要能约到场、约到人。这层是需求侧聚合器，独立于线下场馆 SaaS（中游）。The real start of the funnel — and the top layer the old map missed. Before anything else you have to book a court and find people. This is the demand-side aggregator, distinct from the venue SaaS in the midstream. 传统 Traditional 电话订场 · 熟人拉群约球 · 大众点评找馆 · 教练靠口碑介绍 Phone bookings · WeChat groups to organize games · reviews to find venues · coaches by word of mouth 2C 面向用户 · 约场/约人 Consumer · book & match 01 Playtomic★ 全球最大网球/匹克球/桨板约场平台，2024 GMV €2.4 亿+、约 6000 家俱乐部、150 万月活；2025 融 $70M。AI 做水平匹配与开局撮合 World's largest racquet-booking platform (tennis/pickleball/padel): €240M+ GMV 2024, ~6,000 clubs, 1.5M MAU, $70M raised 2025. AI matches skill levels and fills open games 02 ClassPass / Mindbody 课程/场馆聚合，2026 年 3 月与 EGYM 合并为 Playlist、估值 $75 亿；ClassPass 累计 3.43 亿次预约 Class & studio aggregator; merged with EGYM into 'Playlist' at a $7.5B valuation (Mar 2026); 343M+ ClassPass reservations 03 Keep 约练 / 咕咚 / 小红书搭子 Keep booking / Codoon / RED partner-finding 国内约课、约跑、找搭子的分散入口，尚无统一的拍类约场聚合器——空白 Fragmented Chinese entries for booking classes, runs and finding partners; no unified racquet-court aggregator yet — a whitespace 02 Before · Planning Before · Planning 训练前 · 计划编排 Before · training plans 周期怎么排——强度起伏、练什么、何时减量。这层 AI 渗透相对成熟，因为排课表是纯文本推理，不涉及看动作。How the cycle is laid out — intensity waves, what to train, when to deload. AI penetration here is relatively mature: planning is text reasoning, no need to watch movement. 传统 Traditional 私人教练一对一排课 · 健身杂志/书籍模板 · Excel 手动记录周期 1-on-1 coach programming · magazine / book templates · Excel-tracked cycles by hand 2C 面向用户 Consumer 01 Freeletics / Runna★ Freeletics 5600 万+ 用户 AI 计划；Runna 跑步专项计划，2025 年 4 月被 Strava 收购 Freeletics: 56M+ users on AI plans; Runna: running-specific plans, acquired by Strava (Apr 2025) 02 Pelaris 5 层 LLM 流水线生成周期化方案，覆盖 10 大类、40+ 方法论 A 5-layer LLM pipeline that builds periodized plans across 10 categories and 40+ methodologies 03 WOD App 号称首个自适应 CrossTraining 程序，10 年教练经验 + 自研 AI Claims the first adaptive cross-training program — 10 years of coaching plus in-house AI 2B 教练/场馆 Coaches / gyms 01 Wodify★ 5000+ 健身房管理系统，含 AI 流失预测与编程分发市场 5,000+ gyms on its management system, with AI churn prediction and a programming marketplace 02 TrainHeroic / TrueCoach 教练端远程编程与交付平台 Coach-side remote programming and delivery platforms 03 During · Real-time During · Real-time 训练中 · 实时监测与指导 During · real-time monitoring & coaching 动作发生的当下就给反馈。依赖两条硬件路径：摄像头实时看，或传感器实时测——2C 智能设备与 2B SDK 供应商分野最清晰的一层。Feedback the instant a movement happens. Two hardware paths: a camera that watches, or a sensor that measures — the clearest split between 2C smart hardware and 2B SDK vendors. 传统 Traditional 心率带/秒表人工计时 · 教练现场喊口令纠正 · 健身镜自我观察 HR straps / stopwatches by hand · a coach shouting cues on the floor · self-checking in a fitness mirror 2C 智能硬件 Smart hardware 01 Tonal 2 / Speediance★ 内置摄像头墙装力量设备，实时纠正 + 练后回看 Camera-equipped wall-mounted strength machines, real-time correction plus post-set replay 02 Tempo / Magic AI 3D 传感器/摄像头捕捉动作，实时 form 反馈 + 计数 3D sensors / cameras capture movement for live form feedback and rep counting 03 WurQ 哈佛 Wyss 孵化，腕带+胸带传感器，CrossFit 官方 + HWPO 合作，100+ 动作。走传感器绕开视频 Harvard Wyss-incubated wrist+chest sensors; official CrossFit + HWPO partner, 100+ movements. Chose sensors to sidestep video 2B 动作 SDK 供应商 Motion-SDK vendors 01 Kemtai★ 计算机视觉 B2B，111 数据点、2000+ 动作库；FDA 登记/CE 医疗器械。**可自定义评分**：可上传自有动作、按 API 取 rep/ROM 分数——买视觉留评估。融资约 $5.8M，团队小 CV B2B, 111 data points, 2,000+ exercises; FDA-listed / CE medical device. **Fully customizable scoring**: upload your own exercises, pull rep/ROM scores via API — 'buy the vision, keep the eval.' ~$5.8M raised, small team 02 KinesteX★ 白标 SDK，7 语言框架，疲劳/跌倒检测、ROM，**端侧**（只传坐标不传视频），500+ 动作库、可全 UI 白标 White-label SDK across 7 frameworks; fatigue/fall detection, ROM, **on-device** (only coordinates leave, not video); 500+ exercises, full UI white-labeling 03 Sency 嵌入式 Motion SDK，几分钟接入；计价最清晰——前 100 活跃用户小额月费，之后**按月度独立活跃用户**付费；已覆盖数十家、150 万+ 会员 Embedded Motion SDK, integrates in minutes; clearest pricing — small monthly fee up to 100 MAU, then **pay per unique monthly active user**; tens of companies, 1.5M+ members 04 ASENSEI 白标动作识别 + 端侧 LLM 语音教练，深度整合 Tonal 生态；卖的是教练智能层而非纯姿态 White-label movement recognition + on-device LLM voice coach; deeply tied into the Tonal ecosystem — sells coaching intelligence, not just pose 05 QuickPose 姿态估计 SDK，唯一全透明定价：≤10 设备免费，往上按月活设备 $25/100→$1000/10万。廉价原语，评估逻辑要自己搭 Pose-estimation SDK, the only fully transparent pricing: free ≤10 devices, then per active device $25/100 → $1,000/100k. A cheap primitive — you build the eval logic yourself 04 After · Movement Analysis · Core Layer After · Movement Analysis · Core Layer 训练后 · 动作与技术分析 After · movement & technique analysis 上传一段视频问我哪里没做对——AI 渗透最密集、最能看出项目冷热的一层。红边＝技术上有明确硬骨头，虚线卡片＝还没专门玩家。同类按 AI 热度排序，可用上方筛选。Upload a clip and ask 'what did I do wrong' — the densest layer of AI penetration and the clearest read on which sports are hot or cold. Red border = a clear technical hard bone; dashed card = no dedicated player yet. Sorted by AI heat; filter above. 泛类 App（SportsReflector、FormCheck AI）几乎每项都捆绑覆盖，不单列。此层 2B 供应商见阶段 03——同一批 SDK 支撑着这些 2C 产品。**注意：Kemtai/KinesteX/Sency 这些 SDK 的动作库都是通用健身+康复动作，不含网球/高尔夫/棒球/舞蹈/CrossFit 专项模型；专项技术分析必须自建评估逻辑。**Generic apps (SportsReflector, FormCheck AI) bundle nearly every sport and aren't listed separately. The 2B vendors for this layer are in stage 03 — the same SDKs power these 2C products. **Note: the exercise libraries of Kemtai/KinesteX/Sency are all generic fitness + rehab movements, with no tennis/golf/baseball/dance/CrossFit models; specialized technique analysis must build its own evaluation logic.** 红海/在位 Red ocean / incumbent 成熟/技术壁垒 Mature / barrier 空白/机会 Whitespace / opening 撸铁 · 力量 Lifting · strength 红海 Red ocean **传统：Trad:**私教纠正、镜子自查 PT correction, mirror self-check **AI：AI:**CueForm、FormCheck、Gymscore、AiKYNETIX、ChAIron、Lift it、Metric VBT — $10–13/月价格战 CueForm, FormCheck, Gymscore, AiKYNETIX, ChAIron, Lift it, Metric VBT — a $10–13/mo price war 高尔夫 Golf 强势在位 Strong incumbent **传统：Trad:**PGA 教练现场、录像回放 PGA coach on-site, video replay **AI：AI:**Sportsbox AI（LPGA 背书，单摄 3D 重建，2026 年被 DeChambeau 团队收购）、GEARS、TrackMan Sportsbox AI (LPGA-backed, single-camera 3D, acquired by DeChambeau's group in 2026), GEARS, TrackMan 硬骨头：Hard bone: 球杆轨迹＝器械识别，Sportsbox 靠自研 Kinematic AI 把球杆当关键点一起追——高尔夫专用，非通用方案 Club path = equipment tracking; Sportsbox's Kinematic AI tracks the club as keypoints alongside the body — golf-specific, not general 棒球/垒球 Baseball / softball 最拥挤 Most crowded **传统：Trad:**球队教练组拆解 Team coaching staff breakdown **AI：AI:**Onform、Mustard（投球）、Blast Motion、Rapsodo — 12+ 竞品测评 Onform, Mustard (pitching), Blast Motion, Rapsodo — 12+ competitors reviewed 硬骨头：Hard bone: 球棒轨迹：职业靠视觉（MLB Statcast 用 Hawk-Eye 12 机位 300fps 追球棒），消费级仍靠传感器（Blast 手柄 IMU）绕开视觉 Bat path: the pros use vision (MLB Statcast tracks bats with 12 Hawk-Eye cameras at 300fps); consumers still use a sensor (Blast's knob IMU) to sidestep vision 篮球 Basketball 强势在位 Strong incumbent **传统：Trad:**教练录像逐帧讲 Coach walks the tape frame by frame **AI：AI:**HomeCourt（NBA 球星投资，融资 $4M）、Zepp、Huupe 智能篮球架 HomeCourt (NBA-star backed, $4M raised), Zepp, Huupe smart hoop 瑜伽 Yoga 饱和 Saturated **传统：Trad:**工作室现场指导 In-studio guidance **AI：AI:**Skill Yoga、Zenith Yoga、FitYoga、Prayoga — QuickPose 当免费 demo SDK Skill Yoga, Zenith Yoga, FitYoga, Prayoga — QuickPose as the free demo SDK 网球 Tennis 成熟 Mature **传统：Trad:**教练场边指导 Coach on the sideline **AI：AI:**SwingVision（端侧 Core ML，球+人+击球点，97% 电子司线，已扩匹克球）、PlaySight SwingVision (on-device Core ML: ball + player + contact, 97% line-calling, now covers pickleball), PlaySight 硬骨头：Hard bone: 球拍从击球事件反推而非当刚体 6DoF 追踪；快速小球 + 运动模糊是手机端天花板 The racket is inferred from contact, not tracked as a rigid 6DoF object; a fast small ball + motion blur is the phone-side ceiling 跑步 Running 成熟 + 渠道创新 Mature + channel play **传统：Trad:**跑步店步态分析、理疗师评估 In-store gait analysis, physio assessment **AI：AI:**Ochy（Runners Need 零售店合作）、Movaia、GaitLab、OnForm/Dartfish Ochy (Runners Need retail tie-in), Movaia, GaitLab, OnForm/Dartfish 游泳 Swimming 技术壁垒 Technical barrier **传统：Trad:**泳池边教练目测 Coach eyeballing poolside **AI：AI:**SportsReflector、iSWIM、Swimlytics（均为水上部分）SportsReflector, iSWIM, Swimlytics (all above-water only) 硬骨头：Hard bone: 水下环境让手机视频直接失效，全行业只能做水上，越难越值钱 Underwater conditions break phone video outright; the whole industry stays above water — the harder it is, the more it's worth CrossFit / 功能性 CrossFit / functional 部分已破 Partly cracked **传统：Trad:**Box 教练现场带练 Box coach leading in person **AI：AI:**WodVision（专项分析，33 关键点 + 杠铃路径 + 深度）、Lift it、Gymscore 已覆盖杠铃部分 WodVision (dedicated analysis: 33 keypoints + barbell path + depth), Lift it, Gymscore cover the barbell portion 硬骨头：Hard bone: 体操动作 + wall-ball 是否中靶 + double-under 是否过绳＝目标物识别 + rep 分割 + 规则引擎。最可攻——身体 + 已知简单器械（杠铃）Gymnastics + did the wall-ball hit + did the double-under clear = object recognition + rep segmentation + a rules engine. Most tractable — body plus simple known objects (a barbell) 滑雪 / 单板 Ski / snowboard 生态已成型 Ecosystem formed **传统：Trad:**雪场教练现场带滑 Instructor riding alongside **AI：AI:**视频派 Google×美国滑雪队（隔冬装提骨架 + Gemini 复盘，2026 实验版）、PisteAI、Ride；传感器派 Carv、SkiCoach Video camp: Google × US Ski Team (skeleton through winter gear + Gemini review, 2026 prototype), PisteAI, Ride; sensor camp: Carv, SkiCoach 硬骨头：Hard bone: 厚重冬装 + 装备遮挡——Google 用 DeepMind 单目 3D 姿态在肢体被遮时仍估骨架；但仍不追雪板本身。硬骨头被啃下的样本 Bulky gear + occlusion — Google's DeepMind monocular 3D infers the skeleton even when limbs are hidden; but it still doesn't track the board. A sample of a hard bone being gnawed down 舞蹈 Dance 空白 Whitespace **传统：Trad:**镜面自查、老师现场纠正 Mirror self-check, teacher correction **AI：AI:**未找到专门玩家 — 唯一确认的蓝海 No dedicated player found — the one confirmed blue ocean 硬骨头：Hard bone: 音乐性 + 双人遮挡 + 什么算好无法量化，标准问题非算力问题 Musicality + partner occlusion + 'what counts as good' can't be quantified — a definition problem, not a compute problem 拳击/格斗/MMA Boxing / combat / MMA 空白 Whitespace **传统：Trad:**教练打靶手把手 Coach on the pads, hands-on **AI：AI:**仅泛类 App 捆绑 — 健身拳击品类近年增长快，无专门玩家 Only bundled into generic apps — fitness boxing is growing fast, yet no dedicated player 羽毛球/乒乓/匹克球 Badminton / table tennis / pickleball 亚洲视角空白 Asian-blindspot gap **传统：Trad:**陪练/教练对拆 Sparring partner / coach drills **AI：AI:**欧美 SwingVision/PB Vision 已成型；中国仅小球圈（2025 新创，四拍通吃）在补——详见版图 C 拍类空白 SwingVision/PB Vision are established in the West; in China only Xiaoquanquan (a 2025 startup covering all four) is filling it — see the racket-gap in Map C 攀岩 Climbing 空白 Whitespace **传统：Trad:**路线设定员/同伴建议 Route-setter / partner advice **AI：AI:**未找到专门玩家 No dedicated player found 硬骨头：Hard bone: 正确动作高度依赖具体路线，比舞蹈更难定义标准 The 'right' move depends heavily on the specific route — an even harder standard to define than dance 05 Long-term · Performance Data Long-term · Performance Data 长期 · 表现与负荷数据 Long-term · performance & load data 单次训练之外把周/月/季串起来看趋势——训练量有没有堆多、PR 是否稳步涨。头部方向是从 dashboard 转向自然语言问数据。Beyond a single session, stringing weeks/months/seasons into trends — is volume piling up, are PRs climbing. The leading edge is shifting from dashboards to asking your data in natural language. 传统 Traditional 手写训练日志 · 教练手动统计历史成绩 · 纸质 PR 记录表 Hand-written logs · coaches tallying history by hand · paper PR sheets 2C 面向用户 Consumer 01 Strava Athlete Intelligence★ 全球最大社交训练平台，从数据展示转向训练解释 The largest social training platform, shifting from displaying data to explaining training 02 Garmin Training Status 六因子复合算法给训练负荷与恢复趋势，随设备免费 A six-factor model for training load and recovery trends, free with the device 2B 教练/场馆 Coaches / gyms 01 TrainingPeaks★ 耐力项目训练负荷管理的标准工具 The standard tool for managing training load in endurance sports 02 Wodify Performance Box 端 PR 追踪 + 排行榜，2012 年 CrossFit 成绩追踪起家 Box-side PR tracking + leaderboards, started as a CrossFit results tracker in 2012 06 Injury Prevention & Rehab Injury Prevention & Rehab 伤病预防与康复 Injury prevention & rehab 个人版图里资本规模最大的一层——2B（保险/雇主付费）远超所有 2C 消费级 App 融资总和。付费能力最强，但需专业验证与合规。The most capitalized layer of the personal map — 2B (paid by insurers/employers) dwarfs the combined funding of every 2C consumer app. The strongest willingness to pay, but it demands clinical validation and compliance. 传统 Traditional 线下物理治疗诊所人工评估 · 功能性动作筛查（FMS）人工打分 · 徒手关节角度测量 In-person physio assessment · manual FMS scoring · goniometer joint-angle measurement by hand 2C 个人可用 Consumer-usable 01 Kaia Health★ 摄像头 Motion Coach，2026 年 1 月被 Sword 以 $2.85 亿收购 Camera-based Motion Coach, acquired by Sword Health for $285M (Jan 2026) 02 Exer Health / Kemtai 面向个人的计算机视觉康复 App，摄像头追踪反馈 Consumer computer-vision rehab apps, camera-tracked feedback 2B 雇主/保险/球队 Employers / insurers / teams 01 Hinge Health★ 数字 MSK 龙头，收购计算机视觉公司 wrnch，2025 年 5 月 IPO（约 $26–30 亿市值）Digital MSK leader; acquired CV firm wrnch; IPO'd May 2025 at a ~$2.6–3B market cap 02 Sword Health 融资超 $453M、2025 年 6 月估值升至约 $40 亿，整合传感器 + 计算机视觉双路线$453M+ raised, valuation up to ~$4B (Jun 2025); combines sensors + computer vision 03 Zone7 / Kitman Labs / VALD 球队级伤病风险预测、运动表现平台、肌骨力学测量数据解释 Team-level injury-risk prediction, performance platforms, and musculoskeletal force-measurement interpretation 07 Recovery & Readiness Recovery & Readiness 恢复与身体状态管理 Recovery & readiness 今天该不该练、能练多狠——过去靠主观感觉，现在靠 HRV 和睡眠数据算每日分数。头部可穿戴正把生成式 AI 做成随问随答的解释器。Should you train today, and how hard — once a gut feel, now a daily score from HRV and sleep. The leading wearables are turning generative AI into an ask-anything interpreter. 传统 Traditional 主观疲劳感受判断 · 按摩师/理疗师人工评估 · 纸质睡眠日记 Judging fatigue by feel · masseur / physio assessment · paper sleep diaries 2C 可穿戴 Wearables 01 WHOOP Coach★ 生成式 AI 结合生物数据随问随答，0–100% 恢复分 + Strain。公司估值约 $101 亿 Generative AI over biometric data, ask-anything; 0–100% recovery + Strain. Company valued ~$10.1B 02 Oura Advisor 戒指形态，把睡眠/活动/恢复长期趋势转成行动建议。估值约 $110 亿 Ring form factor turning long-term sleep/activity/recovery trends into actionable advice. Valued ~$11B 03 Garmin Body Battery 全天电量式恢复展示，随设备免费 An all-day 'battery' view of recovery, free with the device 2B 数据整合层 Data-aggregation layer 01 ROOK★ 健康数据聚合 API，供开发者接入多品牌可穿戴数据 A health-data aggregation API letting developers plug into many wearable brands 02 SensAI / Vora / Athlytic 跨设备整合，把 WHOOP/Oura/Garmin 多份数据汇成一份训练建议 Cross-device integration, merging WHOOP/Oura/Garmin data into one training recommendation 08 Nutrition Nutrition 营养管理 Nutrition 训练闭环最后一环——吃够没、吃对没，AI 把手动搜数据库变成拍张照。The final link in the loop — did you eat enough, eat right; AI turns searching a database into snapping a photo. 传统 Traditional 营养师人工制定方案 · MyFitnessPal/Cronometer 手动搜索 · 目测估算分量 Dietitian-built plans · manual search in MyFitnessPal/Cronometer · eyeballing portions 2C 面向用户 Consumer 01 Cal AI★ 拍照识别，2025 年底被 MyFitnessPal 收购、2026 年 3 月公布，整合进 MFP Premium Photo recognition; acquired by MyFitnessPal (closed late 2025, announced Mar 2026), folded into MFP Premium 02 Keep 卡卡 / PlateLens Keep 'Kaka' / PlateLens Keep 的 AI 教练卡卡含饮食识别；PlateLens 独立拍照识别头部 Keep's AI coach 'Kaka' includes food recognition; PlateLens is a standalone photo-recognition leader 2B 数据/识别 API Data / recognition API 01 Nutritionix / Passio★ 食物识别与营养数据库 API，供第三方 App 接入 Food-recognition and nutrition-database APIs for third-party apps 09 Community & Content Community & Content 社区 · 内容 · 分享 Community · content · sharing 运动天然可传播——训练报告、进步图、个人集锦、挑战赛。这层是 2C 留存的关键，也是把运动数据变成社交货币的地方。Sport is inherently shareable — training reports, progress charts, personal highlights, challenges. This is the key to 2C retention and where sports data becomes social currency. 传统 Traditional Strava/Keep 手动发帖 · 小红书/微信群打卡 · 俱乐部线下活动 · 人工剪训练视频 Manual posting on Strava/Keep · check-ins in RED / WeChat groups · offline club events · hand-edited training clips 2C 面向用户 Consumer 01 Strava / Keep 社区 Strava / Keep community★ AI 自动生成训练报告、进步曲线、个人集锦 AI auto-generates training reports, progress curves and personal highlights 02 挑战赛 / 排行榜 Challenges / leaderboards 游戏化激励 + 排行榜反作弊（识别异常数据）Gamified motivation plus leaderboard anti-cheat (flagging anomalous data) 2B 俱乐部/赛事 Clubs / events 01 WODProof★ 25.7 万用户录像/比赛证明，防作弊链接、AI 精彩片段 257k users for workout/competition proof, anti-cheat links and AI highlights ◆ 横向层 · 传感与生物数据基座◆ Cross-cutting · the sensing & biometric base 可穿戴/传感生态——贯穿个人与竞技的采集底座 The wearable & sensing ecosystem — the capture base under both personal and pro sport 传感器不属于任何单一阶段，却给上面每一层供数据。这一层正从卖硬件转向卖 AI 解释，并同时服务 2C（恢复/训练）与 2B（球队监测）。Sensors belong to no single stage yet feed data to every layer above. This tier is shifting from selling hardware to selling AI interpretation, and it serves both 2C (recovery/training) and 2B (team monitoring). WHOOP 订阅制手环，估值约 $101 亿、2025 融 $575M；Coach 把生物数据做成随问随答 Subscription band, ~$10.1B valuation, $575M raised 2025; Coach turns biometrics into ask-anything Oura 智能戒指，估值约 $110 亿、2025 年 10 月融 $907M；Advisor 做长期趋势顾问 Smart ring, ~$11B valuation, $907M raised Oct 2025; Advisor is a long-term trend advisor Garmin 运动手表在位者，Training Status/Body Battery 随设备免费，装机量巨大 The sports-watch incumbent; Training Status/Body Battery free with the device, huge installed base Catapult 2B 精英端，5500+ 职业球队，可穿戴运动员监测 + 视频分析 The 2B elite side, 5,500+ pro teams; wearable athlete monitoring + video analysis 版图 B Map B 竞技体育 · 赛事产业 Pro sport · the event industry 面向球队、联盟、学校、场馆、媒体、赞助商。钱在提高胜率、降低伤病、降低内容成本、增加版权/赞助/粉丝收入。这是 AI 在体育里资本规模最大、也是消费视角容易完全漏掉的一整层。For teams, leagues, schools, venues, media and sponsors. The money is in higher win rates, fewer injuries, lower content cost and more rights/sponsorship/fan revenue. This is AI's most capitalized layer in sport — and the one a consumer lens misses entirely. Value Chain · 产业链十节点 Value Chain · ten nodes 从数据采集到商业变现 From data capture to commercial monetization 越靠左越接近采集，越靠右越接近变现。真正的壁垒（绿色高亮）出现在采集＋标注/规则＋闭环反馈同时成立的地方。The further left, the closer to capture; the further right, the closer to monetization. The real moats (highlighted green) appear where capture + labeling/rules + a feedback loop all hold at once. 01 参与/发现 Participate / discover 赛事、课程、场地、社区触发 Events, classes, courts, community triggers 02 · 壁垒 moat 数据采集 Data capture 视频、GPS、心率、IMU、球轨迹 Video, GPS, HR, IMU, ball trajectory 03 · 壁垒 moat 结构化 Structuring 人/球追踪、姿态、事件切分 Player/ball tracking, pose, event splitting 04 技术动作 Technique 挥拍、投篮、发球、跑姿 Swing, shot, serve, gait 05 战术/比赛 Tactics / match 阵型、跑位、xG/xT、回合价值 Formation, runs, xG/xT, possession value 06 训练计划 Training plan 周期化、强度管理、动态调整 Periodization, load management, dynamic tuning 07 · 壁垒 moat 恢复/伤病 Recovery / injury 疲劳、HRV、复健、回归比赛 Fatigue, HRV, rehab, return to play 08 裁判/公平 Officiating / fairness 鹰眼、VAR、越位、电子司线、诚信 Hawk-Eye, VAR, offside, e-line, integrity 09 内容/直播 Content / broadcast 自动导播、集锦、多语言、粉丝 Auto-directing, highlights, multilingual, fans 10 商业闭环 Commercial loop 订阅、SaaS、数据 API、版权、投注 Subscription, SaaS, data API, rights, betting B1 Team & League Infrastructure Team & League Infrastructure 球队/联盟栈 · 采集 + 数据一体化 Team / league stack · capture + data, unified 头部玩家把视频、可穿戴、招募、高级数据整合进组织操作系统。纯 2B、销售周期长、但数据授权本身就是护城河。The leaders fold video, wearables, recruiting and advanced data into an organizational OS. Pure 2B, long sales cycles — but the data licensing itself is the moat. 传统 Traditional Sportscode 人工打标签 · 教练会议 · Excel 报告 · 训练日志 + RPE Manual tagging in Sportscode · coaches' meetings · Excel reports · training logs + RPE 2B 球队/联盟操作系统 Team / league OS 01 Hudl★ 23 万+ 球队用，从视频工具扩展到 AI 摄像头 + 数据；2024 收购 StatsBomb（3400+ 事件/场、190+ 联赛），整合 Wyscout 招募 230k+ teams; grew from a video tool into AI cameras + data; acquired StatsBomb in 2024 (3,400+ events/match, 190+ leagues), integrated Wyscout recruiting 02 Catapult 5500+ 精英球队，可穿戴运动员监测 + 视频分析 + 招募，OpenField 分析平台 5,500+ elite teams; wearable athlete monitoring + video analysis + recruiting, the OpenField platform 03 Teamworks / Kitman Labs 把训练、医疗、负荷、伤病风险整合进组织操作系统；Teamworks 2026 年 1 月收购 Sportlogiq Fold training, medical, load and injury risk into an org OS; Teamworks acquired Sportlogiq (Jan 2026) B2 Match Data & Tracking · Rights = Moat Match Data & Tracking · Rights = Moat 比赛数据层 · 追踪与版权 Match-data layer · tracking & rights 把 tracking 与 event 数据结合，向分析、媒体、博彩、粉丝输出。关键洞察：真正的护城河是独家官方数据版权，不是追踪技术——挑战者用单摄追踪把技术追平了，却拿不到必须消费官方源的博彩/媒体收入。Combine tracking with event data, then feed analysis, media, betting and fans. Key insight: the moat is exclusive official-data rights, not the tracking tech — challengers matched the tech with single-camera tracking, yet can't touch the betting/media revenue that is contractually required to consume the official feed. 传统 Traditional 人工数据供应 · 内部数据仓库 · 比分 API · 转播画面人工标注 Manual data provision · in-house warehouses · score APIs · hand-tagged broadcast footage 2B 数据/追踪供应商 · 按版权规模排 Data/tracking vendors · by rights scale 01 Genius Sports★ 2025 营收约 $6.55 亿（+28~38%）；**独家官方数据**：NFL、英超/EFL、意甲；拥有 Second Spectrum（NBA/英超/MLS 追踪）。版权即护城河~$655M revenue 2025 (+28–38%); **exclusive official data**: NFL, Premier League/EFL, Serie A; owns Second Spectrum (NBA/EPL/MLS tracking). Rights are the moat 02 Sportradar 2025 营收约 €12.9 亿；博彩为中心，NBA/WNBA 独家博彩数据（8 年、含 3% 股权），2025 收购 IMG ARENA 版权组合~€1.29B revenue 2025; betting-centric; exclusive NBA/WNBA betting data (8 yrs, incl. 3% equity); acquired IMG ARENA's rights portfolio in 2025 03 Stats Perform / Opta Vista 旗下，Opta Vision 把 tracking 与 event 结合、7.3PB 数据 + 140+ AI 模型；企业级授权，无自助 Vista-owned; Opta Vision fuses tracking with events, 7.3PB of data + 140+ AI models; enterprise licensing, no self-serve 04 SkillCorner / Sportlogiq 用单摄/转播源做追踪，成本更低；SkillCorner 融资约 $60M 按结果计价，Sportlogiq 已被 Teamworks 收购 Tracking from single-camera / broadcast feeds at lower cost; SkillCorner (~$60M, outcome-based pricing), Sportlogiq (acquired by Teamworks) B3 Sports Science & Biomechanics Labs · New Sports Science & Biomechanics Labs · New 运动科学 · 生物力学实验室 Sports science · biomechanics labs 补图：精英测试与无标记动捕这一层，区别于消费级动作分析——是喂给上面战术/伤病模块的专业底座。A fill-in: the elite testing and markerless-mocap tier, distinct from consumer movement analysis — the pro-grade backbone feeding the tactical and injury modules above. 传统 Traditional 力板 + 测功计人工测 · Vicon 贴点动捕 · 实验室专家判读 Force plates + dynamometers by hand · marker-based Vicon mocap · lab-expert interpretation 2B 测试 / 无标记动捕 Testing / markerless mocap 01 VALD★ ForceDecks 力板等肌骨力学测量，从英超到 NFL 广泛使用，数据解释层最强 ForceDecks and MSK force measurement, used from the Premier League to the NFL; the strongest data-interpretation layer 02 Theia Markerless Theia3D 无标记三维步态/动作，2025 已对标 Vicon 完成验证，省去贴点 Theia3D markerless 3D gait/motion, validated against Vicon in 2025 — no markers needed 03 Kitman Labs / Vicon Kitman 运动员数据平台整合 VALD 等；Vicon 仍是金标准贴点动捕 Kitman's athlete-data platform integrates VALD and others; Vicon remains the gold-standard marker mocap B4 Tactical AI Copilot · Live Now Tactical AI Copilot · Live Now 战术 Copilot · 从 dashboard 到自然语言问数据 Tactical copilot · from dashboards to asking your data 头部方向不是再多一个 dashboard，而是自然语言问数据、自动找片段、自动生成分析。这个趋势正在 2026 世界杯上真实运行。The leading edge isn't one more dashboard — it's asking your data in natural language, auto-finding clips, auto-generating analysis. This is running for real at the 2026 World Cup. 传统 Traditional 分析师团队人工看录像 · 制作战报 PPT · 依赖财力和技术资源的深度分析 Analyst teams watching tape by hand · building report decks · depth gated by budget and tech resources 2B 赛事级 AI 助手 Event-grade AI assistant 01 FIFA Football AI Pro FIFA Football AI Pro★ FIFA×Lenovo 为 2026 世界杯全 48 队开发的生成式 AI 助手，基于 FIFA 足球语言模型，分析 2000+ 指标、15 种语言自然语言问答 + 3D 可视化（不用于实时比赛）。48 队全部使用 FIFA×Lenovo's generative assistant for all 48 teams at the 2026 World Cup; built on FIFA's Football Language model, 2,000+ metrics, 15-language natural-language Q&A + 3D viz (not for live matches). All 48 teams use it 02 Stats Perform / VALD AI 让专业人员更快理解可信数据的同类趋势，AI 解释层 The same trend of helping pros understand trusted data faster — an AI explanation layer ⬢硬骨头 Hard bone 运动专项壁垒 + 数据权。通用视觉模型只懂人在动，不一定懂动作为什么错；真正壁垒是专项规则 + 专家标注 + 训练闭环。且比赛视频、球员数据、健康数据都涉及权利和隐私——B2B 里数据授权本身就是护城河。Sport-specific expertise + data rights. A general vision model knows 'a person is moving,' not why the move was wrong; the real barrier is sport-specific rules + expert labeling + a training loop. And match video, player data and health data all carry rights and privacy — in B2B, data licensing is the moat itself. B5 Recruitment & Scouting Recruitment & Scouting 招募 · 选材 · 青训 Recruitment · scouting · youth 找下一个球星既是数据挑战也是艺术。AI 让俱乐部远程试训、在低级别联赛里找符合顶级球员统计画像的隐藏璞玉；青训/校园招募是同一条链的下沉。Finding the next star is both a data problem and an art. AI lets clubs run remote trials and find hidden gems in lower leagues that match a top player's statistical profile; youth and school recruiting is the same chain pushed downmarket. 传统 Traditional 线下球探带笔记本看球 · 试训 · 集锦 · 经纪人网络 Scouts with notebooks at matches · trials · highlight reels · agent networks 2B 选材平台 Scouting platforms 01 Wyscout / StatsBomb★ 视频/数据选材，数千场比赛按位置/年龄/风格筛选（Hudl 旗下）Video/data scouting, filtering thousands of matches by position/age/style (under Hudl) 02 aiScout 远程试训，与切尔西等俱乐部合作，用手机视频比对成功球员画像 Remote trials with clubs like Chelsea, matching phone video against successful-player profiles 03 NCSA / Hudl Recruit / SkillCorner NCSA / Hudl Recruit / SkillCorner NCSA 做美国校园招募，Hudl Recruit 青训下沉；SkillCorner 从转播源提物理数据评估跑动 NCSA for US college recruiting, Hudl Recruit for youth; SkillCorner extracts physical data from broadcast to assess movement B6 Officiating, Fairness & Integrity Officiating, Fairness & Integrity 裁判 · 公平 · 诚信 Officiating · fairness · integrity 从判罚追踪扩展到半自动越位、电子司线、3D 球员化身，再到反假球/反作弊的诚信监测。技术最成熟、准确性要求最高，也是 AI 最早规模落地的场景之一。From officiating and tracking to semi-automated offside, electronic line-calling and 3D player avatars — extending to integrity monitoring against match-fixing. The most mature tech, the highest accuracy bar, and one of AI's earliest at-scale deployments. 传统 Traditional 主裁/边裁肉眼 · 即时回放 · 早期 VAR 人工画线 · 人工监测异常投注 Referees' eyes · instant replay · early VAR drawing lines by hand · manual watching for betting anomalies 2B 判罚 / 诚信系统 Officiating / integrity systems 01 Hawk-Eye (Sony)★ 鹰眼判罚系统（属索尼），NBA×Sony Hawk-Eye 光学追踪战略合作，网球电子司线金标准 The Hawk-Eye officiating system (owned by Sony); NBA × Sony Hawk-Eye optical-tracking partnership; the gold standard for tennis line-calling 02 FIFA 半自动越位 + 3D 化身 FIFA semi-auto offside + 3D avatars 2026 世界杯 1248 名球员全部 3D 扫描（每人约 1 秒），越位判罚 + 更真实的 3D 动画回放 At the 2026 World Cup, all 1,248 players 3D-scanned (~1s each) for offside calls and more realistic 3D animated replays 03 Sportradar UFDS AI 反假球/诚信监测：2025 年监控 1M+ 场次、标记 1116 场可疑比赛，FIFA 诚信合作延至 2031。判罚的防守面 AI anti-fixing / integrity: monitored 1M+ events in 2025, flagged 1,116 suspicious matches, FIFA integrity deal extended to 2031. The defensive side of officiating B7 Content · Broadcast · Fan Content · Broadcast · Fan 内容 · 直播 · 粉丝 Content · broadcast · fans 把一场比赛拆成海量个性化内容——从赛后长视频变成每人一个故事。让低级别、校园、群众赛事也能拥有类转播素材，可接广告、赞助、会员。Slice one match into vast personalized content — from a post-game long-form to a story per fan. It gives lower-league, campus and grassroots events broadcast-like footage, monetizable via ads, sponsorship and memberships. 传统 Traditional 编辑团队赛后剪辑 · 人工导播 · 人工字幕 · 专业摄制团队才能直播 Editors cutting after the game · manual directing · manual captions · only pro crews could broadcast 2B 自动内容 / 集锦 Auto content / highlights 01 WSC Sports★ 525+ 客户含 NBA/PGA/NHL，一个 NBA 季后赛生成 67,000 个个性化集锦；视频周转从 2 小时降到 5–10 分钟 525+ clients incl. NBA/PGA/NHL; 67,000 personalized highlights in one NBA playoffs; turnaround cut from 2 hours to 5–10 minutes 02 Sony Beyond Sports 从判罚追踪扩展到数据驱动的沉浸式和动画转播 From officiating tracking to data-driven immersive and animated broadcasts 2B AI 摄像机 / 直播 AI cameras / streaming 01 Pixellot / Veo★ AI 自动直播，让校园/青训/低级别比赛拥有类转播素材；Pixellot 年产约 150 万场，Veo 覆盖 2.5 万+ 俱乐部、融 $80M AI auto-broadcast giving campus/youth/lower-league games broadcast-like footage; Pixellot ~1.5M games/yr, Veo 25k+ clubs, $80M raised 02 Hudl Focus AI 自动跟拍摄像头，无需摄像师 An AI auto-tracking camera, no operator needed B8 Sponsorship & Fan Monetization · New Sponsorship & Fan Monetization · New 赞助衡量 · 粉丝互动 · 幻想体育 Sponsorship measurement · fan engagement · fantasy 补图：内容之外的两条变现线——用 CV 量化赞助曝光价值，以及幻想体育/预测/粉丝 Chatbot 这一注意力与付费的大面。A fill-in: two monetization lines beyond content — using computer vision to quantify sponsorship exposure value, and the fantasy/prediction/fan-chatbot surface of attention and spend. 传统 Traditional 人工估算赞助曝光 · 关注度靠事后调研 · 论坛聊球 · 纸质竞猜 Estimating sponsor exposure by hand · post-hoc surveys for attention · forum banter · paper pools 2B 赞助价值衡量 Sponsorship valuation 01 Relo Metrics★ NVIDIA 算力驱动的计算机视觉，量化 logo 曝光时长与媒体等值价值，2025 扩展到 F1 NVIDIA-powered computer vision quantifying logo exposure and media-equivalent value; expanded to F1 in 2025 02 Blinkfire Analytics 社媒 + 转播赞助价值实时衡量，赞助商归因 Real-time sponsorship valuation across social + broadcast, with sponsor attribution 2C 幻想体育 / 预测 / 粉丝 Fantasy / prediction / fans 01 DraftKings / FanDuel★ 每日幻想体育巨头，AI 做阵容推荐与个性化；也是投注入口（见变现层）The daily-fantasy giants, AI for lineup recommendations and personalization; also betting entries (see the monetization layer) 02 Sleeper 幻想体育 + 预测市场（接入 Kalshi），社交化玩法，累计融资约 $68M Fantasy + prediction markets (Kalshi-integrated), social gameplay, ~$68M raised 03 俱乐部超级 App / 粉丝 Chatbot Club super-apps / fan chatbots 生成式 AI 粉丝助手、个性化推送与二创，正成为标配 Generative-AI fan assistants, personalized pushes and remixes are becoming table stakes B9 Venue & Event Operations Venue & Event Operations 场馆 · 赛事运营 Venue · event operations 智能排班、观众动线、票务定价、赞助归因、场馆安全，也包括社区场馆的经营 SaaS——怎么经营一门运动生意这一整层。动态票务与安检等归在此。Smart scheduling, crowd flow, ticket pricing, sponsorship attribution, venue security — plus the community-venue operating SaaS: the whole 'how to run a sports business' layer. Dynamic ticketing and security fold in here. 传统 Traditional 人工检票排班安保 · 固定票价 · 人工赞助售卖 · CRM 表格 Manual ticketing, scheduling and security · fixed pricing · hand-sold sponsorship · CRM spreadsheets 2B 大型场馆 / 赛事 Large venues / events 01 动态票务定价 Dynamic ticket pricing★ 2026 世界杯采用动态定价，按需求/天气/买家画像实时调价 The 2026 World Cup uses dynamic pricing, adjusting in real time by demand/weather/buyer profile 02 Fastbreak AI 赛程与差旅优化，NFL 2025–26 赛程由 AI 生成 Schedule and travel optimization; the NFL 2025–26 schedule was AI-generated 2B 社区场馆 SaaS Community-venue SaaS 01 Wodify / Mindbody / Glofox Wodify / Mindbody / Glofox★ 健身房/工作室一体化管理：账单、排课、CRM、AI 会员流失预测 All-in-one gym/studio management: billing, scheduling, CRM, AI churn prediction 02 三体云动 / 青橙 / 微健 Trainbody / Qingcheng / Weijian 国内无人场馆 SaaS：人脸识别 0.3 秒入场、防尾随、月付自动续费、无人值守 Chinese unmanned-venue SaaS: 0.3s face-recognition entry, tailgating prevention, auto-renewing monthly billing, no staff 03 Peloton IQ / Perch 智能器械 + 重量房 VBT，rep 追踪 + form 反馈（Perch 面向高校/职业队重量房）Smart machines + weight-room VBT, rep tracking + form feedback (Perch targets college/pro weight rooms) ⚑ 合规敏感层 · 仅描述全球市场，不构成投注推荐⚑ Compliance-sensitive layer · a descriptive global map, not a wagering recommendation 版图 D · Map D · 数据变现与投注层 Data monetization & betting layer 官方数据版权是整套追踪生态背后的营收引擎——追踪技术存在，一部分正是为了喂这门数据生意。AI 在此既做进攻（定价、微盘、球员盘），也做防守（反假球、负责任博彩）。把它作为明确标记的子层而非并列版图，因为它是骑在采集/数据版图之上的变现应用，且法律高度地区化。Official-data rights are the revenue engine behind the whole tracking ecosystem — the tech exists partly to feed this data business. Here AI plays both offense (pricing, micro-markets, player props) and defense (anti-fixing, responsible gambling). It's a clearly-flagged sub-layer, not a co-equal map, because it rides on the capture/data maps and its legality is highly jurisdictional. 合规红线：**体育投注在中国大陆基本非法，仅国家体育彩票（1994 年起）合法，境外/线上投注被禁并处于持续打击中；澳门为独立博彩法域。**美国 2018 年 PASPA 废除后按州合法化。本层为全球市场结构描述，不构成任何投注推荐。Compliance red line: **sports betting is largely illegal in mainland China — only the state Sports Lottery (since 1994) is legal; offshore/online betting is banned and under active crackdown; Macau is a separate gaming jurisdiction.** In the US it's legal state-by-state after the 2018 PASPA repeal. This layer describes global market structure only and is not a wagering recommendation. #### 营收引擎 · 官方数据版权 Revenue engine · official-data rights 联盟把独家数据卖给分销商，再转售给博彩 Leagues sell exclusive data to distributors who resell to sportsbooks NBA × Sportradar NBA × Sportradar 约 $10 亿+、含约 3% 股权（2023–2031），球员追踪数据打包~$1B+, incl. ~3% equity (2023–2031), player-tracking data bundled NFL × Genius Sports NFL × Genius Sports 现金 + 股权，全期或达约 $10 亿，独家 Next Gen Stats + 官方投注源 Cash + equity, up to ~$1B over its life; exclusive Next Gen Stats + official betting feed 结论 Takeaway 版权即护城河，追踪技术只是被打包进去的使能器 Rights are the moat; tracking tech is just the enabler bundled inside #### 进攻 · AI 定价与盘口 Offense · AI pricing & markets AI-in-sports 约 $3.1B(2026)、约 30% CAGR AI-in-sports ~$3.1B (2026), ~30% CAGR 动态赔率 / 微盘 Dynamic odds / micro-markets Sportradar 2024 上线 AI 微盘（下一投/下一分），秒级重定价 Sportradar launched AI micro-markets in 2024 (next pitch/point), sub-second repricing 微投注 Micro-betting Simplebet 2024 年 8 月被 DraftKings 以约 $1.5 亿收购；nVenue 接棒实时概率 Simplebet acquired by DraftKings (~$150M, Aug 2024); nVenue picked up live probabilities 球员盘 / 个性化 Player props / personalization Swish Analytics 做赔率起盘与盘口生命周期，运营商做个性化 Swish Analytics for odds origination and prop lifecycle; operators for personalization #### 防守 · 诚信与负责任博彩 Defense · integrity & responsible gambling 每个进攻用例都配一个防守孪生 Every offensive use has a defensive twin 反假球 Anti-fixing Sportradar UFDS 2025 标记 1116 场可疑比赛、84 国、125 项处罚 Sportradar UFDS flagged 1,116 suspicious matches across 84 countries, 125 sanctions in 2025 负责任博彩 AI Responsible-gambling AI FanDuel AceAI 检测问题博彩并转人工；Flutter 2025 投入 $158M；ROGA 行业数据共享 FanDuel's AceAI detects problem gambling and escalates to humans; Flutter invested $158M in 2025; ROGA shares industry data 生物数据权 Biometric-data rights 球员工会推动数据/投注使用的同意与分成，隐私是新战场 Players' unions push for consent and revenue share over data/betting use; privacy is the new battleground 版图 C Map C 中国市场 · 产业链上中下游 The China market · the value chain 前面以欧美为主，这一张补上被系统性低估的中国市场。按 36 氪产业链框架分上游（资源供应）、中游（健身服务）、下游（后端需求），并标出各层 AI 渗透状态。绿点＝AI 已深入，褐点＝刚起步，灰点＝基本空白。The maps above skew Western; this one fills in the systematically underrated China market. Following 36Kr's value-chain frame — upstream (supply), midstream (services), downstream (demand) — with each layer's AI penetration marked. Green = deep AI, brown = emerging, grey = mostly blank. 上游 Upstream 资源供应方 Resource suppliers 器械 · 教培 · 课程研发 Equipment · training · courseware 健身器械 Fitness equipment Life Fitness、Technogym、Impulse 英派斯、舒华、Merach 麦瑞克、Keep 智能硬件 Life Fitness, Technogym, Impulse, Shua, Merach, Keep hardware **AI 已深入：**器械巨头集体 AI 化（详见下方），Merach 走硬件+内容+AI，CES 2026 首发大模型对话跑步机**Deep AI:** equipment giants going AI en masse (below); Merach runs hardware+content+AI, debuting an LLM-chat treadmill at CES 2026 教育培训 Education & certification 赛普健身、ASCC、体适能培训机构 Sunpig, ASCC, fitness-cert institutions **基本空白：**教练认证培训仍以线下面授为主，AI 渗透最浅的一层**Mostly blank:** coach certification is still in-person; the shallowest AI layer 课程研发 Course R&D LesMills 莱美、AIDOING、各健身房自研课程 LesMills, AIDOING, in-house gym programs **刚起步：**AI 辅助编排课程内容，但好课程仍高度依赖人的经验**Emerging:** AI assists course arrangement, but a good class still leans hard on human experience 中游 Midstream 健身服务提供方 Fitness-service providers 线下健身房 · 线上平台 · 场馆 SaaS Gyms · online platforms · venue SaaS 线下新型健身房 New-model gyms 乐刻（500 店/320 万用户/月付制）、超级猩猩、Keepland、光猪圈 Leoao (500 stores/3.2M users), Supermonkey, Keepland, Punch Gym **刚起步：**智能门禁、按次付费、数字化运营；乐刻做运动产业数字中台**Emerging:** smart entry, pay-per-visit, digital ops; Leoao building a sports-industry data mid-platform 传统健身房 Traditional gyms 威尔仕、一兆韦德、英派斯、古德菲力、银吉姆 Wills, Physical, Impulse, Goodfeel, Yinjm **刚起步：**疫情后被迫线上化，引入智能场馆、直播引流，但组织惯性大**Emerging:** forced online post-pandemic, adopting smart venues and live-stream funnels, but heavy organizational inertia 线上平台 · 自研模型 Online platforms · own models Keep（4 亿注册/约 3000 万月活）、咕咚、每日瑜伽、悦跑圈 Keep (400M registered / ~30M MAU), Codoon, Daily Yoga, Joyrun **AI 已深入：**Keep 自研 **Kinetic.ai** 运动健康垂类大模型（行业首发），AI 教练卡卡横跨计划/实时语音/饮食识别/复盘**Deep AI:** Keep's in-house **Kinetic.ai** vertical model (an industry first); AI coach 'Kaka' spans planning / live voice / food recognition / review 无人场馆 SaaS Unmanned-venue SaaS 三体云动、青橙、微健、乐刻 17 练平台 Trainbody, Qingcheng, Weijian, Leoao's 17练 **AI 已深入：**对标 Wodify/Mindbody——人脸识别 0.3 秒入场、防尾随、无人值守、月付自动续费**Deep AI:** the Wodify/Mindbody analog — 0.3s face entry, tailgating prevention, unmanned, auto-renewing billing 下游 Downstream 后端需求服务方 Downstream demand 运动服饰 · 营养代餐 Apparel · nutrition & meal replacement 运动服饰 Sportswear 安踏、李宁、FILA、lululemon、MAIA ACTIVE、粒子狂热 Anta, Li-Ning, FILA, lululemon, MAIA ACTIVE, Particle Fever **基本空白：**AI 在设计/供应链/营销后端有用，但与运动过程关系弱，非本图核心**Mostly blank:** AI helps in design/supply chain/marketing, but is weakly tied to the act of playing — not this map's core 运动营养 · 代餐 Sports nutrition · meal replacement 汤臣倍健、康比特、薄荷健康、ffit8、超级零 By-Health, Kanbeite, Boohee, ffit8, Chaojiling **刚起步：**薄荷健康的 AI 饮食管理口碑好，与训练闭环营养环（版图 A-08）打通是机会**Emerging:** Boohee's AI diet management is well-regarded; wiring it into the training-loop nutrition stage (Map A-08) is the opportunity ◆ 版图 C 最值得下手的切口 · 拍类运动专项 AI 空白验证◆ The sharpest opening in Map C · racket-sport AI whitespace, verified 羽毛球 / 乒乓球 / 匹克球——空白部分成立，且正在快速被补 Badminton / table tennis / pickleball — the gap is real, and closing fast 结论：不是纯空白，而是部分空白、快速收敛。关键玩家是 2025 年新创的小球圈（前字节/DJI/蔚来自动驾驶工程师创立，四拍通吃，CV+云端 LLM+3D 人体重建，智能网球拍量产在即）——它既验证了机会，也意味着一个通用 CV 引擎可能抢在单项专才前面。Verdict: not a pure blank but a partial one, converging fast. The key player is Xiaoquanquan (founded 2025 by ex-ByteDance/DJI/NIO self-driving engineers; covers all four racket sports; CV + cloud LLM + 3D body reconstruction; a smart tennis racket nearing mass production) — which both validates the opportunity and means a general CV engine could pre-empt sport-specific entrants. 羽毛球 Badminton 真空白 Real gap 羽毛球追踪是**已解决的研究问题**（TrackNet 系列约 97%），却没打包成大众消费产品；中国仅多练、小球圈等新创在补。机会最实。Shuttle tracking is a **solved research problem** (TrackNet ~97%) yet unpackaged into a mass consumer product; in China only new startups (Duolian, Xiaoquanquan) are filling it. The most concrete opportunity. 乒乓球 Table tennis 结构性空白 · 被硬件占据 Structural gap · taken by hardware 需求被**发球机器人**（庞伯特 Pongbot，CES 2026，轨迹分析、Aura 机 $499 于 2026 年 7 月发货）与国家队生物力学（商汤 InnoMotion、上体体育大模型）占据；消费级手机视觉 App 空白部分是结构性的——球太小太快、室内近距，市场选了硬件。Demand is taken by **ball-launch robots** (Pongbot, CES 2026, trajectory analysis; the $499 Aura ships Jul 2026) and national-team biomechanics (SenseTime InnoMotion, Shanghai Sport Univ's LLM); the consumer phone-vision app gap is partly structural — the ball is too small and fast, indoors and close-range, so the market chose hardware. 匹克球 Pickleball 最干净的空白 Cleanest whitespace 中国爆发式增长（2025 年 5000+ 球场、约 500 万参与、约 ¥18 亿市场、约 47% CAGR），美国 AI 已成熟（SwingVision、PB Vision×DUPR），但**本土几乎零专项 AI**——年轻人群 + 已被美国验证的付费 + 无本地在位者。Explosive China growth (2025: 5,000+ courts, ~5M participants, ~¥1.8B market, ~47% CAGR); US AI is mature (SwingVision, PB Vision × DUPR), yet **almost zero local specialized AI** — a young base + monetization already proven in the US + no local incumbent. ◆ 单开一条：产业巨头的 AI 化◆ A line of its own: how the industry giants are going AI Technogym AI Ecosystem 40 年器械老厂，2026 年 3 月发 agentic AI 生态：3800 万用户数据、生成式 AI Coach、给教练的 AI Assistant（编课省 80%+ 流失预测），米兰冬奥官方供应商 A 40-year equipment maker; launched an agentic AI ecosystem Mar 2026: 38M user datasets, a generative AI Coach, an AI Assistant for trainers (80% less programming time + churn prediction). Official Milan Winter Olympics supplier Life Fitness Halo Cloud Halo Fitness Cloud 做设备预测性维护；Connect 开放 API 接入 100+ 第三方 App 与可穿戴，走开放高速公路路线对打 Technogym 的围墙花园 Halo Fitness Cloud for predictive maintenance; Connect opens an API to 100+ third-party apps and wearables — the 'open highway' against Technogym's 'walled garden' Keep 的自研模型 Keep's in-house model 中国唯一有运动健康垂类自研大模型的平台（Kinetic.ai），既是线上平台又向硬件延伸，是消费端巨头自建模型的样本 China's only platform with an in-house vertical sports-health model (Kinetic.ai); both an online platform and a hardware play — the sample of a consumer giant building its own model