AI Predicts Tommy John Surgery 100 Days in Advance

The stakes are immense in professional baseball. A pitcher undergoing Tommy John surgery faces an average recovery period of 20.5 months. For the team, the financial hit is just as significant with an estimated $1.9 million per pitcher in salary and lost value.

The deeper I dive into this research, the more I wonder: is UCL injury just the luck of the draw? With so many variables at play (mechanics, workload, genetics, fatigue) it often feels impossible to predict or prevent.

But what if it isn't?

What if teams could detect the warning signs not just pitches, but months in advance? What if an algorithm could analyze a pitcher's data and flag the subtle, invisible signs of a ligament under strain?

A groundbreaking new study leveraging deep learning and a massive set of in-game pitching data (5,537,981 pitches 😯 )suggests this future is closer than we think, offering a potential breakthrough in predicting and preventing baseball's most devastating injury.

A 100-Day Warning Shot

The study's most significant contribution is a deep learning model that can predict a pitcher's risk of needing TJS up to 100 days before their last game prior to surgery. This is a monumental leap forward from previous approaches. Earlier video-based analysis could only make predictions very close to the injury event (sometimes within just 30 pitches) which seems to be far too late for any meaningful preventative action.

This 100-day window is the game-changer according to the authors. It aligns almost perfectly with the typical 3-month (approximately 12-week) non-operative rehabilitation period prescribed for partial UCL tears. For the first time, this technology gives teams a practical and actionable timeframe. An early warning at 100 days could potentially allow a team's medical and coaching staff to intervene with rest, targeted therapy, or mechanical adjustments, potentially allowing the ligament to heal and avoiding surgery altogether. It shifts the focus from reacting to an injury to proactively preventing it.

Teaching an AI to "See" Patterns in Pitching Data

So, how does the model achieve this? The researchers' most successful tool was a Vision Transformer (ViT), a type of AI typically used for advanced image recognition, like identifying objects in a photo. This might seem counter-intuitive for analyzing pitching data, but the method seemed to be effective.

Researchers converted the raw, time-series pitching data (dozens of metrics for every single pitch thrown over months) into single-channel images. In essence, they turned a spreadsheet of numbers into a visual pattern. This allowed the ViT model to do what it does best: "SEE" and identify complex patterns and subtle interactions between different pitching metrics over time. This method proved remarkably successful, demonstrating a high degree of accuracy in distinguishing between pitchers on a path to injury and their healthy counterparts using only widely available in-game data.

Subtle Mechanical Flaws That Predict Injury

A prediction is only useful if you can understand what's behind it. While the Vision Transformer was better at spotting who is at risk, the researchers used a separate regression model combined with Explainable AI (XAI) to understand why and when.

This second model, designed to predict the exact timeline to injury, allowed them to pinpoint the specific mechanical red flags it was seeing. They discovered three measurable changes in a pitcher's mechanics, primarily centered on their four-seam fastball, that consistently signaled an impending UCL tear.

• A Lower Release Point: The model found that the vertical release point of the four-seam fastball (Release_Pos_Z_FF) consistently decreased as the pitcher got closer to the date of their injury.

• A Changing Spin Axis: The spin axis of the four-seam fastball (Spin_Axis_FF) became significantly more horizontal as the injury neared, indicating a change in how the pitcher was releasing the ball.

• A Dip in Velocity: The speed of the four-seam fastball (Release_Speed_FF) showed a clear trend of decreasing closer to the injury event, a classic sign that something is physically wrong.

The critical insight is the biomechanical reason why these signs matter. The study concludes that a lowered release point is directly associated with increased elbow valgus torque. In plain terms, the AI discovered the data-driven fingerprint of a pitcher whose mechanics were placing an unsustainable strain on their elbow ligament, causing it to gradually fail… potentially haha.

Is This A New Era of Injury Prevention?

This research signals a potential paradigm shift in how baseball teams manage pitcher health. By analyzing standard in-game data with AI, teams can now receive substantial advance warning (over three months) that a pitcher is at high risk for Tommy John surgery. The algorithm identifies subtle mechanical flaws invisible to the naked eye, creating an opportunity for early, targeted interventions that could save careers and millions of dollars.

But there's a caveat: this technology is only accessible to organizations with the resources to build these models, hire the right people, and maintain the massive datasets required. For now, the competitive advantage belongs to MLB teams with deep pockets haha.

The ethical considerations are equally important. While this technology empowers coaches to proactively manage workload and mechanics, it must be wielded carefully. Teams can't risk derailing a player's career or pulling them from competition based solely on an algorithm's prediction. The key will be integrating these data-driven alerts with the expert judgment of coaches, trainers, and medical staff with appropriate ethical guardrails using AI as a powerful tool, not a replacement for human decision-making.

As this technology becomes more refined, it raises a fascinating question:

• Will we see a day when a pitcher's career is saved by an algorithm flagging risk that no human eye could detect?

• And how will teams balance the precision of data science with the irreplaceable human element of coaching and player development?