Pinch Hit Grader: Methodology

Managers make the correct platoon call 64.5% of the time, for an average wOBA gain of +0.031 per substitution. Leverage weighting raises the effective run value (average leverage is 1.155, above the PA-weighted mean of 1.0), since pinch hit situations tend to arise in higher-run-expectancy states. The 2021 National League season, the last before universal DH adoption, shows the sharpest advantage: NL pinch hits for pitchers averaged +5.5 RV/100.

A pinch hit substitution is a discrete, manager-called decision. Given the current pitcher (and their handedness), is the replacement batter a better platoon matchup than the player being removed? We grade the decision quality, not the outcome: a pinch hitter who grounds out on a quality swing still receives a GOOD_PH if the wOBA matchup supported the substitution.

We do not grade whether the manager should have used a pinch hitter at all (e.g., burning a bench player in a low-leverage situation). We grade only the quality of the specific matchup chosen when a substitution is made.

Pinch hit events are not surfaced in the pitch-level Statcast CSV, because the substitution itself happens between pitches. We use two sources:

• MLBAM Stats API (/api/v1/game/{game_pk}/playByPlay): identifies each offensive_substitution event, capturing the pinch hitter ID, the replaced batter ID, and the at-bat index.
• Statcast event-level data via pybaseball: joined by (game_pk, at_bat_number) to provide base-out state, pitcher identity, and pitcher handedness at the time of the substitution.
• Season wOBA splits computed directly from the same Statcast parquets: sum(woba_value) / sum(woba_denom) grouped by (batter, game_year, p_throws). IBBs are excluded (they have woba_denom = 0 in Statcast).

Each pinch hit appearance is graded using two quantities:

The run value of the substitution is the leverage-weighted expected offensive gain:

where woba_scale = 1.157 (the standard FanGraphs factor that converts a wOBA difference to an expected run-per-PA difference) and RE_MEAN = 0.477 (the frequency-weighted mean RE across all MLB plate appearances, 2020–2026).

The grade is based on woba_gain alone, not leverage. Leverage scales the run_value magnitude but does not change whether a substitution was a good or bad matchup call:

• GOOD_PH: woba_gain > 0. The pinch hitter has a better platoon matchup against the current pitcher.
• BAD_PH: woba_gain ≤ 0. The replaced batter was at least as good a matchup.

A wOBA gain of +0.05 in a bases-loaded, 0-out situation (RE ≈ 2.24, weight ≈ 4.7×) is worth far more than the same gain with the bases empty and 2 outs (RE ≈ 0.10, weight ≈ 0.21×). Using the RE24 leverage of the base-out state captures this without requiring win-probability data, which would introduce dependency on score and inning.

The average leverage weight across all 23,946 graded pinch hit appearances is 1.155 (vs. the expected 1.0 for an average MLB PA). This confirms that managers do use pinch hitters disproportionately in above-average run-expectancy situations.

Season wOBA splits against each pitcher hand require meaningful sample sizes to be reliable. Batters with fewer than 50 plate appearances against a given pitcher hand in a season have their split replaced with the league-average wOBA for that hand and season.

Imputation rates in the current dataset (2020–2026):

The high imputation rate vs. LHP reflects two structural facts: left-handed starters appear in roughly 29% of starts (61% of PH situations involve a RHP), and pinch hitters themselves are often bench players who accumulate fewer plate appearances overall. When both the pinch hitter and replaced batter use league-average splits, the grade still captures whether the matchup was directionally correct, but with lower precision.

We aggregate by batting team × season, since the manager controls pinch hit decisions:

• n_ph: total pinch hit appearances
• good_ph_rate: fraction with woba_gain > 0
• avg_woba_gain: mean wOBA improvement per substitution
• run_value_per100: total run_value / n_ph × 100

Entries with fewer than 20 pinch hit appearances in a season are flagged Low. In practice, no team-season in 2020–2026 fell below this threshold (minimum observed: 21 PH, in the 60-game 2020 season).

Pinch hit decision quality shows weak to moderate year-over-year stability at the team level. Spearman correlation of run_value_per100 between consecutive seasons (n = 30 teams per pair, all teams with ≥ 20 PH each year):

The weak 2020–2021 and 2021–2022 pairs are structural, not random. The 2020 short season produces small per-team samples (typically about 25–70 PH vs. about 55–165 in post-DH full seasons). The 2021–2022 transition reflects the universal DH adoption: NL teams in 2021 had large, reliable GOOD_PH readings from pinch-hitting for pitchers, and those signals disappeared in 2022 when every team used a DH. In the stable full-season post-DH era (2022 onward), the mean ρ is +0.300, ranging from +0.12 to +0.44 by pair: weaker and noisier than the ESV module, but consistently positive. Interpret team-level trends across multiple seasons more confidently than any single season.

2021 was the final season in which National League teams were required to bat pitchers. This created a unique structural environment: a pinch hitter replacing a pitcher almost always represented a massive wOBA upgrade (average pitcher wOBA vs. RHP ≈ .110; league-average hitter ≈ .320). 2021 NL pinch hits for pitchers averaged +5.5 RV/100, more than twice the post-DH rate (+2.5 RV/100 across all pinch hits, 2022–2026).

This structural shift is why the 2021–2022 YoY rho is weak (+0.26): 2021 NL rankings reflect pitcher-replacement opportunities that simply no longer exist. The 2020 season adds a second confounder (60-game schedule, small samples, no fans). For cross-era comparisons, 2022–2026 is the cleanest window.

• ▲Season-average wOBA splits, not recent form. We use a batter's full-season wOBA split vs. a pitcher hand. The manager at the time of substitution may be responding to a hot streak, a platoon tendency against the specific pitcher, or recent performance that the season average does not reflect.
• ▲High imputation rate vs. LHP. 38% of pinch hitters and 53% of replaced batters use the league-average split when facing left-handed pitchers, because most bench players face LHP infrequently. In those cases the grade is driven by directional matchup logic rather than precise individual splits.
• ▲No pitcher quality or fatigue signal. We compare batter wOBAs but do not model the current pitcher's effectiveness. A manager pinch-hitting against a tiring ace vs. a fresh middle reliever may be making the same matchup call by our metric but a very different decision in practice.
• ▲No "best available" counterfactual. We grade the specific substitution made, not whether a better option was available on the bench. A GOOD_PH reflects a real upgrade over the replaced batter, but the team might have had an even better platoon option who was not used.
• ▲Weak year-over-year stability at the team level (mean ρ +0.30, ranging +0.12 to +0.44 by pair, in the post-DH era). Pinch hit decision quality has a real signal, but the sample per team per season is small (typically about 55–165 PH post-DH), and roster construction changes year to year. Multi-season aggregates are more reliable than single-season grades.
• ▲No manager attribution. We aggregate at the team-year level. Manager names require a separate lookup CSV not yet in the automated pipeline. Multi-year team entries serve as a reasonable proxy.
• ▲2021 NL entries are not comparable to post-DH seasons. The pitcher-batting era inflated 2021 NL good_ph_rate and RV/100 significantly. These entries are included for completeness but should not be compared directly to 2022-onward values.
• ▲2020 entries carry higher uncertainty due to the 60-game season (typically about 25–70 PH per team vs. about 55–165 in post-DH full seasons). These are flagged 60g in the leaderboard.