Which Triple-A numbers travel to the majors
The finding Across 188 moves from Triple-A to the majors by 147 hitters, swing decisions, whiffs, and power carried almost intact: chase rate (r = 0.80), whiff rate (0.81), and 90th-percentile exit velocity (0.85). Triple-A wOBA carried nothing (r = 0.05). For pitchers (144 moves by 119 players), Stuff+ carried best (0.79), strikeout rate partly (0.49), and run value not at all (0.05). Even the best Triple-A number explained only about 6% of a hitter’s MLB xwOBA in his first 100+ PA, and nothing predicted a pitcher’s MLB run value.
So what
- Judge a Triple-A hitter by his swing, not his slash line. Chase, whiff, and EV90 in Triple-A are close to what he’ll show in the majors, after a predictable shift: chase rate up 3.1 points, whiff rate up 1.4, EV90 down 0.9 mph. His Triple-A wOBA tells you almost nothing.
- Expect fewer walks and more strikeouts. The typical hitter’s walk rate fell 3.4 points and his strikeout rate rose 3.2. Pitchers’ strikeout rates fell 3.2 points.
- For pitchers, start with Stuff+. It was the most portable pitching number. Triple-A run value, the closest thing here to ERA, said nothing about MLB run value.
What carries and what doesn’t
Each bar is one Triple-A number’s correlation with the same number in the majors, for the same players. A high correlation means the order of players held: the hitter who chased least in Triple-A tended to chase least in the majors. It says nothing about the level, which is the next section.
Swing decisions and contact quality kept their order best: chase rate (r = 0.80, interval 0.75 to 0.84), whiff rate (0.81), and EV90 (0.85). Strikeout rate (0.71) and hard-hit rate (0.71) were close behind; walk rate (0.57) less so. Triple-A xwOBA carried some (0.29, interval 0.16 to 0.41). Triple-A wOBA, the number closest to a batting line, carried nothing (r = 0.05, interval −0.06 to 0.16).
Pitchers looked the same way. Stuff+, which grades only the pitch itself, held its order at 0.79 (interval 0.71 to 0.85). Strikeout rate (0.49), whiff rate (0.52), and walk rate (0.41) carried partly. Run value (0.05) and xwOBA allowed (0.05) didn’t carry at all. Split by MLB role, starters (63) and relievers (70) show the same pattern; pick a role above to see it.
How much each number shifts
The table under the chart gives the typical change for the same player from Triple-A to the majors: the median, weighted so that players with more playing time count more. For hitters, the typical shifts were: chase rate +3.1 points (interval +2.8 to +3.7), strikeout rate +3.2 (+2.2 to +3.9), walk rate −3.4 (−3.8 to −3.0), hard-hit rate −3.2, EV90 −0.9 mph, and xwOBA −.036. For pitchers: strikeout rate −3.2 points and whiff rate −3.8.
Those shifts describe the players who were promoted, not a fixed exchange rate between leagues. The pitch grades show why that matters. Pitchers who moved up within a season kept about the same Stuff+ in MLB (a shift of +1.7, interval −0.4 to +3.6), while those compared a season apart graded 4.8 points higher. That gap more likely reflects a year of development, and the fact that pitchers whose stuff improved were the ones who stayed in the majors, than anything about the majors themselves.
What predicts major-league production
The second view of the chart asks a different question: which Triple-A number best predicts how well a hitter produces in the majors (MLB xwOBA) or how well a pitcher prevents runs (MLB run value)? Triple-A xwOBA alone explained about 6% of the differences in MLB xwOBA, out of sample. All ten Triple-A numbers together explained 5%, no more than xwOBA alone. Triple-A wOBA did worse than guessing the cohort average (cross-validated R² −0.025). For pitchers, nothing did: the best single Triple-A number (whiff rate) reached a cross-validated R² of 0.021, and all nine together −0.043, no better than predicting the cohort average for everyone.
That’s less bleak than it sounds. Many of these are short first stints in the majors, some as few as 100 plate appearances or batters faced, where results are mostly noise. The process numbers, which do carry from level to level, are the better guide to what a player is.
The players behind each number
Pick a number to see every player in the cohort: his Triple-A value across, his MLB value up. The dashed line is “no change”; the solid line is the fit. A fit flatter than the dashed line means the number regresses toward the middle on the way up.
Translating a current Triple-A player
Every Triple-A player with 150+ plate appearances or batters faced gets a translated line: each number run through its own fitted line from the scatter above, with a band that held 80% of the cohort. The band is wide. For strikeout rate it’s about ±5 points; for xwOBA, about ±.037, and the xwOBA line is nearly flat (slope 0.26, against 0.80 for chase rate), so a big Triple-A xwOBA translates to only a little above average. Stuff+, Location+, and Pitching+ are already on the MLB scale, and their shift is mostly the selection described above, so player pages don’t translate them. Translations are on each Triple-A player’s page and as a view in the Triple-A explorer.
Method and caveats
Data. Pitch-level Statcast from Baseball Savant, Triple-A and MLB, 2025–2026 regular seasons, aggregated to season lines in the dbt pipeline (fct_translation_lines). Stuff+, Location+, Pitching+, and RV+ use the same models and frozen scale as the pitch-grade leaderboard. In-zone means the height-based strike zone, the same at both levels and in both seasons, so chase and zone rates compare directly.
Cohorts. Next season: 2025 Triple-A (150+ PA or batters faced) to 2026 MLB (100+): 108 hitters, 95 pitchers. Same season: in 2025 and 2026, a player’s Triple-A line before his first MLB game that season against his MLB line from that game on, same minimums: 80 hitters, 49 pitchers. Pooled: 188 hitter pairs (147 players) and 144 pitcher pairs (119 players). Pitchers take their MLB role (63 SP, 70 RP, 11 Swing).
Statistics. Every pair is weighted by the smaller of its two samples. Correlations are weighted Pearson; shifts are weighted medians of (MLB − Triple-A). Intervals are 90% bootstrap intervals over players (2,000 resamples; a player in both cohorts is resampled as one). “Predicts MLB production” uses MLB xwOBA for hitters and MLB RV+ for pitchers; the cross-validated R² comes from ridge regressions (standardized inputs, penalty chosen by inner cross-validation) on 5-fold cross-validation grouped by player, repeated 20 times. Below zero means worse than predicting the cohort average for everyone.
Translation. For each number, a weighted least-squares line of MLB on Triple-A over the pooled cohort; the band is the 10th to 90th percentile of the cohort’s weighted residuals around that line. The same band applies to every player, whatever his sample.
What this can’t tell you. Only promoted players are observed, and promotion depends on things this data doesn’t see: scouting reports, injuries, roster needs, and often the very Triple-A numbers being tested. Players who struggled in a short MLB stint and went back down are in the data; players who were never called up aren’t. The samples are small, especially within a season, so read the broad pattern rather than the second decimal. The shifts mix the level with a season of development (next-season cohort) and with in-season change, and they don’t adjust for park or for the Triple-A league environment (the Pacific Coast League plays as a hitters’ league).
Code: baseball/aaa_translation/ and fct_translation_lines