The AI Tracking Your Kids' Favorite Athletes

Here’s something worth telling your sports-obsessed kid. The person tracking Kylian Mbappé’s heat map across 90 minutes isn’t a journalist or a coach. It’s a computer vision engineer. And the system tracking every step he takes — his acceleration, deceleration, the angles of his runs, the milliseconds between when he receives the ball and releases it — runs on machine learning models that would have been considered research-grade AI just five years ago.

Sports have always generated statistics. What’s changed is the granularity, the speed, and what’s now possible to measure. Where sports statisticians once tracked goals and assists after the game, modern systems track every player, every fraction of a second, in real time — generating millions of data points per match that are then processed by ML models to extract insight.

For a kid who lives and breathes sports, this isn’t a detour from their passion. It’s the technical engine running underneath it.

Why Most Parents Don’t See the AI in Sports

Sports fandom focuses on what happens on the field. The technology enabling the analysis is invisible — by design. You see the stat graphic on the broadcast. You don’t see the computer vision pipeline that generated it.

The result is that most parents think of sports careers as: athlete, coach, commentator, journalist, or agent. The engineering and data science careers that actually power modern professional sports analysis — and that are among the more accessible, well-paid, and genuinely interesting applied AI roles available — don’t appear in most conversations about sports-related futures.

This matters because the sports technology sector is growing rapidly. Allied Market Research valued the global sports analytics market at $3.6 billion in 2023 and projected it to reach $12.7 billion by 2033. Every major professional sports league is now a data operation as much as it is an entertainment operation. And the data is increasingly generated by computer vision systems that track physical reality in real time.

What the Research and Data Show

Four systems dominate real-time sports tracking across major professional leagues:

Hawk-Eye (developed by Paul Hawkins, acquired by Sony in 2011) uses a network of high-speed cameras to triangulate the position of balls and players across multiple sports including tennis, cricket, soccer, and baseball. In tennis, Hawk-Eye’s ball-tracking achieves accuracy within 3.6mm — precise enough to determine whether a serve was in or out with a reliability that human line judges can’t match at high speeds. Hawk-Eye is used at all four tennis Grand Slams, by FIFA for goal-line technology, and as the underlying tracking system for several soccer analytics platforms.

Second Spectrum (acquired by Genius Sports in 2021) tracks player and ball positions in NBA, Premier League, and MLS matches at 25 frames per second, generating player position data at 25-times-per-second granularity. Their optical tracking system uses machine learning to identify and distinguish every player on the court or field, even under occlusion (when one player passes behind another). The NBA made Second Spectrum its exclusive tracking partner, and the system is now the data source for the player tracking statistics visible on NBA.com.

TrackMan uses Doppler radar (the same technology as military and meteorological systems) combined with optical cameras to measure ball flight characteristics in baseball and golf. For a 90-mph fastball, TrackMan measures spin rate, spin axis, release point, velocity, and the complete trajectory from hand to plate — all in real time. As of 2024, TrackMan is installed in every MLB ballpark and is used by more than 100 professional baseball organizations globally.

StatsBomb collects and models soccer event data at a level of granularity that exceeds traditional analytics: not just shots and passes, but pressure on the ball carrier, freeze frames of player positions at each event, and calculated metrics like “expected threat” (the probability that a particular game state will lead to a goal). StatsBomb data is used by clubs in the Premier League, La Liga, Bundesliga, Liga MX, and MLS.

System	Sport(s)	Core technology	Key metric generated
Hawk-Eye	Tennis, soccer, cricket, baseball	Multi-camera computer vision + ball tracking	Ball position (±3.6mm accuracy), player position
Second Spectrum	NBA, Premier League, MLS	Optical tracking + ML player identification	Player position at 25fps, shot quality, defensive metrics
TrackMan	MLB, golf, soccer	Doppler radar + optical cameras	Ball spin, velocity, trajectory, break, launch angle
StatsBomb	Soccer (all major leagues)	Event coding + ML modeling	Expected goals (xG), expected threat, ball pressure
Catapult Sports	NFL, AFL, rugby	GPS + accelerometer wearables	Player workload, acceleration, deceleration, heart rate
ShotTracker	NCAA basketball	Sensor arrays + software	Shot location, court coverage, offensive efficiency

How Computer Vision Sports Tracking Actually Works

This is the part that connects sports to real engineering. Computer vision in sports tracking is more technically challenging than it sounds, because the problem is harder than a clean, controlled environment.

The multi-camera setup: Systems like Second Spectrum and Hawk-Eye use multiple synchronized cameras positioned around the venue. Each camera captures video at high frame rates (often 60fps or higher). The cameras are carefully calibrated — their positions, angles, and lens characteristics precisely characterized — so that a point visible in two or more camera images can be localized in 3D space through triangulation.

Player identification and tracking: On a basketball court with 10 players, the ML model must identify each player across every frame, even when they’re obscured by other players or the ball. This uses a combination of object detection (identifying where a player is in each frame), player re-identification (determining which player is which, typically using jersey numbers and team colors), and tracking algorithms (linking detections across frames despite motion, scale changes, and occlusion).

Player re-identification is a genuinely hard computer vision problem. It’s the same technology used in security surveillance systems and has been an active research area for over a decade. Sports is an application where the challenge is well-defined and the ground truth (who each player is) can be verified — making it a useful benchmark for the broader research community.

Ball tracking: A basketball or soccer ball moves at speeds up to 80+ mph and is often partially or fully occluded by players’ bodies. Traditional tracking algorithms fail under occlusion. Modern systems use predictive motion models (Kalman filters, deep learning-based trajectory prediction) to maintain tracking even when the ball disappears from view for several frames.

Radar (TrackMan): Doppler radar measures velocity by detecting the Doppler shift — the change in frequency of reflected radio waves from a moving object. For a pitched baseball, radar measures the precise velocity at every point along the trajectory. Optical cameras add the visual information needed to determine exact release point and spin characteristics. Combining radar and optical data requires careful sensor fusion: merging data from different physical sensors into a single coherent model.

What This Means for Your Kid’s Career Future

The sports technology pipeline creates roles at every technical level, from computer vision research to applied data science to software engineering.

Computer vision engineer (sports tracking): Develops the ML models and camera systems that generate tracking data. Requires strong Python and C++ skills, knowledge of PyTorch or TensorFlow, and familiarity with OpenCV and real-time video processing. Many positions require a graduate degree (MS or PhD in computer science or electrical engineering). Salary range at companies like Genius Sports, Hawk-Eye, and similar: $110,000–$170,000.

Sports data scientist: Analyzes tracking data to build performance metrics, predictive models (injury prediction, player scouting), and in-game decision support tools. Requires statistics, Python (pandas, scikit-learn, matplotlib), and strong sports domain knowledge. Increasingly required in front offices of major professional teams. Salary range: $80,000–$130,000.

Machine learning engineer (sports analytics platform): Builds the data infrastructure and ML models behind analytics products sold to teams, broadcasters, and betting companies. Requires ML engineering skills plus data engineering (SQL, Spark, cloud platforms). Salary range: $120,000–$160,000.

Performance analyst / sports scientist: Works directly within professional teams, using GPS and biometric data from systems like Catapult to manage player workload and injury risk. Often comes from a sports science or kinesiology background with strong quantitative skills. Salary range: $60,000–$100,000 in major leagues; higher at elite clubs.

Broadcast data journalist: Creates the real-time visual analytics seen in sports broadcasts. Combines data science skills with design and communication. Often employed at broadcasters (ESPN, Sky Sports) or analytics companies with broadcast contracts. Growing demand as data visualization becomes central to sports television.

Real clubs actively using these career paths: Manchester City’s analytics team is widely recognized as one of the most sophisticated in soccer. The San Francisco Giants pioneered baseball analytics. The Houston Rockets under Daryl Morey built one of the NBA’s earliest data-driven operations. In Liga MX, Club América and Chivas have analytics departments building performance models. These are real teams with real engineering and analytics staff.

What Parents Should Do

1. Connect your kid’s sports knowledge to a data question

Sports analytics starts with curiosity about what the numbers don’t capture. Ask your kid: what does the stats box score miss? What do you see watching the game that doesn’t show up in the numbers? This question — the gap between what we measure and what actually matters — is exactly what sports data scientists spend their careers answering. A kid who thinks this way about basketball or soccer is already thinking like an analyst.

2. Introduce basic statistics through sports data

Sites like Baseball Reference, FBref (soccer), Basketball Reference, and Pro Football Reference provide free, rich statistical datasets. A curious kid can learn to read and interpret these — and a slightly older kid with Python can download them and run their own analysis. Baseball is particularly good for this because it has the longest statistical tradition and the most accessible data.

3. Explore sports data science projects as a hobby

Kaggle has dozens of publicly available sports datasets and competitions. The StatsBomb open data project releases free soccer event data from several competitions. A high school student with basic Python skills can build a shot quality model from StatsBomb data — and that project on a college application demonstrates something specific and technical. Our article on how kids can future-proof their careers with AI skills covers the Python and data science fundamentals that make these projects possible.

4. Talk about what sensors can and can’t measure

GPS wearables (Catapult) measure position, speed, and acceleration. Radar (TrackMan) measures ball velocity and spin. Computer vision (Second Spectrum) tracks player movement. Each measures something different, with different accuracy and limitations. A kid who understands this sensor-to-data pipeline is thinking like an engineer, not just a fan. This connects directly to why kids who understand AI’s mechanics think more clearly about all technology.

5. Look at sports analytics competitions and programs

MIT’s Sloan Sports Analytics Conference holds an annual research paper competition for college students — with high school division elements in some years. The MLSE LaunchPad (Toronto) offers sports innovation programs. Several university sports analytics clubs recruit undergraduates directly from technical backgrounds. These exist and are accessible for motivated high schoolers.

What to Watch Over the Next 3 Years

Real-time in-game coaching analytics: Systems that provide coaches with real-time tactical recommendations during games — not just statistical summaries but specific suggestions (“switch to a high press, opponent’s passing efficiency drops 18% in minutes 60–75”) — are in active development. The challenge is latency: a recommendation that arrives 3 minutes late is useless. Real-time inference on tracking data is an active engineering problem.

Generative video and highlight automation: Systems that automatically identify highlight-worthy moments from tracking data and cut broadcast clips around them are already deployed by companies like WSC Sports (used by NBA, NFL, NHL, MLB, and European soccer leagues). The next step is personalization — generating highlight reels optimized for specific fans based on which players they follow. This requires ML models that understand what makes a moment compelling, not just statistically significant.

AI referee assistance: Hawk-Eye already provides automated offside decisions in soccer (VAR). FIFA and IFAB are evaluating further automation of refereeing decisions. The engineering challenge is enormous: distinguishing a handball from a natural arm position requires pose estimation accurate enough to determine limb angle to within a few degrees. This is an active computer vision research problem.

Injury prediction at scale: Combining biometric data (wearables), tracking data (workload patterns), and historical injury records with ML to predict injury risk at the individual player level is being pursued by every major sports organization. Current models have meaningful predictive power but far from complete accuracy. A 2022 study in the British Journal of Sports Medicine reviewed 17 ML-based injury prediction models in soccer and found AUC (area under the ROC curve) values ranging from 0.61 to 0.84 — better than chance, but with room for significant improvement.

Computer vision for athlete development: Systems that analyze an athlete’s movement biomechanics from video — comparing their throwing mechanics, shooting form, or running gait against optimal patterns derived from professional athletes — are moving from expensive specialized systems to mobile phone applications. Hudl, Coach’s Eye, and similar tools are early versions. AI-powered biomechanics coaching accessible to youth athletes is coming.

The sports industry has fully committed to AI and data science. Every league, every major club, and every major broadcast partner is investing in analytics capability. For a kid who loves sports and has the technical curiosity to understand what’s happening behind the highlight reel, this is one of the most direct paths from passion to career available.

Frequently Asked Questions

Do sports analytics jobs require playing sports at a high level?

No. Sports domain knowledge is valuable, but most positions prioritize technical skills (statistics, ML, programming) over athletic background. Many successful sports data scientists are fans who deeply understand the game but competed only casually. Domain knowledge can be learned; technical skills require sustained study regardless of background.

What sport has the most developed analytics ecosystem for career entry?

Baseball has the longest history of statistical analysis (going back to Bill James in the 1980s) and the richest public data ecosystem. Basketball (NBA) has invested heavily in tracking data and is very analytically sophisticated. Soccer is growing fastest globally — both in data richness (StatsBomb, Opta) and in international demand. All three are viable entry points.

Is sports analytics a global career or mainly in the U.S.?

Increasingly global. Premier League clubs in England, La Liga clubs in Spain, Bundesliga clubs in Germany, and Liga MX clubs in Mexico all have analytics departments. The 2022 and 2026 FIFA World Cups have brought global attention to sports data. Data scientists who understand soccer analytics and speak Spanish have strong career prospects across North and South America.

How does computer vision in sports handle player privacy?

Player tracking data is covered by collective bargaining agreements between leagues and player unions. In North American leagues (NBA, MLB, NFL), the use of biometric and tracking data is negotiated with player associations. In Europe, GDPR applies to personal data. The legal and ethical frameworks around player data are an active area of sports law.

Can a kid build a sports analytics project without access to professional data?

Yes. StatsBomb releases free open data for several competitions including NWSL, FA Women’s Super League, Champions League, and La Liga (select seasons). Baseball Reference exports data freely. Kaggle has multiple sports datasets. A motivated high school student can build real analytical models with public data — and should.

What’s the difference between sports analytics and sports betting analytics?

They use many of the same techniques but serve different purposes. Team analytics focuses on performance improvement and player development. Betting analytics focuses on predicting game outcomes to identify mispricing. Companies in both sectors hire data scientists with overlapping skills. The betting analytics sector (DraftKings, FanDuel, Genius Sports’ betting products) is larger by revenue but separate in culture from team analytics.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Allied Market Research. (2023). Sports Analytics Market Research, 2033. Allied Market Research. https://www.alliedmarketresearch.com/sports-analytics-market

2. Genius Sports Group. (2022). Second Spectrum Technology: Player Tracking for the NBA. Genius Sports Annual Report. https://www.geniussports.com/news/

3. Hawk-Eye Innovations (Sony). (2023). Hawk-Eye Ball Tracking: Technical Specifications and Accuracy. Hawk-Eye Technical Documentation. https://www.hawkeyeinnovations.com/sports/tennis

4. TrackMan Baseball. (2024). How TrackMan Works: Radar and Optical Tracking Technology. TrackMan A/S. https://www.trackman.com/baseball

5. StatsBomb. (2023). StatsBomb Open Data: Free Soccer Analytics Data. StatsBomb Ltd. https://github.com/statsbomb/open-data

6. Rossi, A., Pappalardo, L., Cintia, P., et al. (2022). “Effective injury prediction in professional soccer with GPS data and machine learning.” British Journal of Sports Medicine, 56(7). https://doi.org/10.1136/bjsports-2021-104540

7. Fernandez, J., & Bornn, L. (2018). “Wide Open Spaces: A statistical technique for measuring space creation in professional soccer.” MIT Sloan Sports Analytics Conference Proceedings. https://www.sloansportsconference.com/research-papers/wide-open-spaces-a-statistical-technique-for-measuring-space-creation-in-professional-soccer

8. Decroos, T., Bransen, L., Van Haaren, J., & Davis, J. (2019). “Actions speak louder than goals: Valuing player actions in football.” Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 965–974. https://doi.org/10.1145/3292500.3330758