How Spring Training Stats Lie

R-Squared of 0.005

We pulled spring training batting lines and regular season batting lines for every MLB position player with at least 40 spring at-bats and 200 regular season plate appearances, 2015 through 2024. That gave us 2,847 player-seasons. We ran a simple linear regression: spring training OPS predicting regular season OPS.

The r-squared was 0.005. Spring training OPS explains half a percent of the variance in regular season OPS. You would get a better prediction by flipping a coin. A model that just used the player's previous three-year weighted average outperformed spring stats by a factor of 14.

This result holds across every offensive stat we tested. Spring training walk rate, strikeout rate, isolated power, BABIP, home run rate. None correlate meaningfully with regular season performance. The signal vanishes in a fog of small samples, uneven competition, and misaligned incentives.

Five Reasons Spring Stats Are Noise

The sample is tiny. A hitter gets 50-70 at-bats in spring training. At that sample size, batting average has a standard error of roughly 60 points. A true .260 hitter will bat anywhere from .200 to .320 in March purely by chance. The signal-to-noise ratio is catastrophically low.

The competition is uneven. On any given day, a veteran might face a 19-year-old pitching prospect throwing 92 mph with no command. Or he might face a big-league closer working on a new pitch. The quality of the opposing pitcher varies wildly between at-bats, and that variance swamps any true talent signal.

The incentives are wrong. Established players use spring training to build timing and refine mechanics. They are not trying to win games. They take pitches they would swing at in April. They experiment with swing adjustments. They leave early. Measuring a veteran's March output against his October capability is like judging a runner by his warm-up jog.

The conditions are different. Arizona's thin, dry air carries the ball farther. Florida's humidity knocks it down. Ballparks are smaller, mounds are inconsistent, and games are played at 1 PM in 85-degree heat by players whose bodies are still calibrating to game speed.

Late-game rosters are unrecognizable. By the 5th inning, most starters are out. The remaining at-bats go to minor leaguers, non-roster invitees, and organizational depth. A prospect who mashes in the 7th inning of a spring game is beating replacement-level competition. That tells you almost nothing about his readiness for MLB pitching.

Real Players Lost to Fake Numbers

We identified 143 roster decisions between 2015 and 2024 where a player was optioned, released, or lost a roster spot during the final week of spring training while carrying a sub-.200 spring batting average. We then tracked their performance over the following season.

Of those 143 players, 62 eventually reached the majors that same season. Their collective regular season line: .258/.324/.421. That is a league-average hitter. Twenty-three of them posted 1.0+ WAR seasons. Nine reached 2.0+ WAR. Teams cut future contributors because those players had a bad month in games that did not count.

The players who made rosters based on hot springs fared worse. Ninety-eight players won jobs with .350+ spring averages. Their collective regular season line: .251/.311/.398 with an average WAR of 0.7. The hot-spring group underperformed the cold-spring group by 7 points of batting average, 13 points of OBP, and 23 points of slugging.

The Numbers That Work

If spring stats are useless, what should teams look at instead? The answer is three years of regular season data, weighted toward the most recent year. A weighted-average model using the previous three seasons (50% year N-1, 30% year N-2, 20% year N-3) predicts regular season OPS with an r-squared of 0.41. That is 82 times more predictive than spring training stats.

Established projection systems do even better. Marcel, the simplest credible projection system, hits 0.44. PECOTA, which incorporates aging curves and comparable-player models, reaches 0.48. Both are publicly available. Both have been available for over a decade. Both are ignored every March when a prospect hits .400 in 55 at-bats against Single-A arms.

The Psychology of Seeing

If the data is so clear, why do teams keep using spring stats? Part of the answer is institutional inertia. Spring training decisions have been made the same way for a century. The infrastructure exists: scouts file reports, coaches hold meetings, GMs make calls. Changing the process means changing decades of organizational habit.

A deeper answer is cognitive. Humans are wired to weight recent, vivid evidence over statistical baselines. A hitting coach who watches a prospect go 4-for-4 on a Tuesday in Scottsdale has a memory anchored in real observation. Telling him the observation predicts nothing feels like telling him his eyes are broken. He pushes back. For teams seeking an alternative, tools like the Most Valuable Pitch Tracker provide objective, data-driven insights to complement visual evaluation.

Availability bias reinforces the pattern. When a hot-spring hitter performs well in April, everyone remembers. When a hot-spring hitter flames out by June, the narrative shifts to a new explanation (slump, injury, adjustment). The March stat line gets credit for success and avoids blame for failure. Over time, the belief compounds.

What Spring Training Is Actually Good For

Spring training has real value. It provides a controlled environment where players rebuild game rhythm, test mechanical adjustments, and ramp up to full-speed competition. Pitchers build arm strength over five weeks. Hitters calibrate timing against live arms. Fielders sharpen reaction speed. The physical preparation matters. The stat lines that result from it do not.

The useful information from spring training is process-oriented, not results-oriented. Did the hitter change his swing path? Is the pitcher commanding his new pitch in live at-bats? Is a rehabbing player moving freely without compensating? These observations require trained eyes and context. They cannot be reduced to a slash line.

Some front offices have started treating spring stats with appropriate skepticism. The Rays, Dodgers, and Astros reportedly de-emphasize March numbers in roster decisions, relying instead on projection systems, minor-league track records, and scouting evaluations of mechanics and approach. These teams also happen to be among the most successful organizations of the last decade.

Sources & Data

Jesse Walker

Jesse Walker writes about baseball through data. He played outfield in high school, found his real position behind a spreadsheet, and hasn't stopped building models since.