How it’s counted

October means every postseason game in the MLB Stats API (game type P: wild card, division series, league championship series and World Series). The regular season means every American and National League regular-season game. 2026 is left out: its postseason was still being played when this was built.

Pooling. One postseason is a few dozen games at most, and before 1969 it was only the World Series, so single-year dots are noisy. Seasons are pooled in ten-season blocks counted back from 2025 (2016–2025, 2006–2015, … 1906–1915, then 1903–1905). A pooled rate is the total over the block divided by the total, not an average of yearly rates. A block is shown when at least three of its seasons have the data.

Two comparisons. “Every game” sets October against the whole regular season. But only good teams play in October, so “Playoff teams” sets it against the same teams’ own regular seasons: for batting rates, every game those teams played (their hitting and their pitching together); for pitching usage and quality, their own staffs; for the game measures, games with at least one of them; for the better-record measure, games between two of them. The gap that survives the second comparison is October’s, not the teams’.

Noise. The error bars are two standard errors of the pooled gap, treating each team’s postseason (or each game, or each series, depending on the measure) as one independent draw. A grey number means the gap is inside that range. It is a rough yardstick, not a test of anything in particular.

Who pitches. Each pitcher’s regular-season ERA (or FIP) is weighted by the innings he threw: in the regular season by his innings for that team, in October by his postseason innings. The pitchers are the same; only the weights move. FIP follows the FanGraphs formula, with the constant worked out from each season’s AL/NL totals. Postseason innings by a pitcher with no regular-season innings are left out.

Where each measure starts

Each measure starts after the last season that lacks a number it needs, checked field by field, so a few early seasons with complete records don’t stand in for an era that mostly lacks them. A comparison needs both columns; a pooled block, at least three seasons with both.

What the build checks

The build script stops without writing anything if a check fails. It checks that hitting and pitching totals agree (runs, hits, home runs, games) for the regular season and October in every season, within 0.5% (the oldest records disagree by up to that much); that the schedule’s final scores add up to the teams’ runs (within 1% for the regular season, exactly for October) and give every postseason team its number of games; that won–lost records counted from the schedule match the standings within six team-games (a game or two in some seasons, and in 1946 and 1962 two or three games that the schedule files as regular season and the standings leave out); that every postseason series has a winner; that starters’ and relievers’ innings add up to each team’s, and its pitchers’ innings to the team’s, for every season a measure uses; that plate appearances equal AB + BB + HBP + SH + SF + CI; and that no measure has a missing season after the one it starts in. It also checks the league against Baseball-Reference: runs per game, batting average, on-base and slugging for 1968, 2000, 2019 and 2023–2025, and ERA and complete games per game for eleven seasons from 1905 to 2024.

Spot checks by hand, all matching: the 1927, 1960 and 2024 World Series (runs, batting average and slugging, from Baseball-Reference’s series pages; the 2024 series batting lines match line for line), and 2024 and 2000 regular-season strikeout, walk and home-run rates, steal attempts, sacrifice bunts and intentional walks per game against Baseball-Reference’s league averages.