How every DiamondIQ number is computed

Standard public WAR framework: sum the run contributions above replacement, then divide by runs-per-win. We build it from components we control.

• Batting runs (park-adjusted wRAA): (wOBA − lgwOBA) / wOBA-scale × PA, with wOBA from the season line using published linear weights stored in our league_constants. The hitter's wOBA is park-neutralized first (see park factors below).
• Baserunning (v1): stolen-base runs only, SB × 0.2 − CS × 0.4 (published linear weights). Extra-base advancement is future work.
• Positional adjustment: the published per-position values (C +12.5 … SS +7.5 … DH −17.5 per 162 defensive games), prorated by the player's games at each position from the league fielding appearances.
• Fielding: runs-denominated defense from Baseball Savant — Outs Above Average (the leaderboard's own fielding_runs_prevented, ≈ 0.8 runs per out) for non-catchers, plus catcher framing runs. Players with no qualifying Savant data (DH-only, small samples) are shown as an explicit omission, never a fabricated zero.
• Replacement: the ~570-WAR-pool convention, ≈ 20 runs per 600 PA — a rate, so it's season-length independent.
• Runs per win: computed from the live run environment, 9 × (leagueR / leagueIP) × 1.5 + 3 (≈ 9.9 in 2026).

FIP-based runs above replacement (public methodology):

• FIP: (13·HR + 3·(BB+HBP) − 2·K) / IP + league constant, where the constant centers FIP on league ERA.
• Park adjustment: FIP is divided by the hitter's-park multiplier so a pitcher in a tough run environment gets credit.
• vs. replacement: measured against a replacement level blended between starter (≈ .380 team) and reliever by the pitcher's starter share, scaled by innings and converted to a runs scale over runs-per-win. Reliever leverage is simplified in v1 (flat) — a known limitation.

WAR vs. PitchIQ. These answer different questions. PitchIQ is a percentile quality index (how good, including Statcast arsenal); WAR is accumulated value (how much, given innings). A dominant arm with 40 IP posts a high PitchIQ but modest WAR — a durable league-average workhorse can out-WAR him. Both are real; neither replaces the other.

Our own, computed from MLB game logs — not a third-party feed. For each park we compare total runs per game (both teams) in that team's home games to its road games over a rolling 3-season window, which cancels out the home team's own quality. The raw ratio is regressed toward 1.00 by sample size, then the 30-park set is renormalized so the league mean is exactly 1.00 by construction. A hitter plays ~half at home, so a player's park multiplier is (parkFactor + 1) / 2. Coors sits ≈ 1.14; T-Mobile Park ≈ 0.89.

Both are 0-100 percentile indices vs. the qualified population (50 = league average, 80+ = elite). BatIQ blends park-adjusted results (wOBA), Statcast contact quality (xwOBA / hard-hit / whiff), and plate discipline (BB%−K%), regressed toward 50 for small samples. PitchIQ blends K-BB%, run prevention, and Statcast arsenal quality. Unlike WAR, they don't reward playing time — they measure per-opportunity quality.