본문 바로가기
MoneyBallScore

MLB 14 Model Factors

KBO 10 factors (FIP · xFIP · wOBA · Bullpen FIP · Recent Form · WAR · H2H · Park Factor · Elo · Defense SFR) + Statcast 4 (xwOBA · Barrel% · xwOBA-against · wOBA σ) = 14 factors.

Weight total = 100% (including home bonus 10%). Weights defined in packages/kbo-data/src/factors/mlb-base.ts. Model will update after n=150 forward cohort milestone.

Weight Summary Table

FactorCategoryWeight
SP FIPStarting12%
SP xFIPStarting3%
Lineup wOBALineup10%
Bullpen FIPBullpen10%
Recent FormForm10%
Team WARRecord8%
H2HRecord3%
PFPark4%
EloRating10%
Defense SFRDefense5%
Lineup xwOBAStatcast5%
Barrel%Statcast3%
SP xwOBA-aStatcast4%
wOBA σStatcast3%
Home Elo Bonus (+24 Elo (~ +3.4%))10%
Total100%

KBO 10 Factors (Equivalent)

The 10 factors from KBO model v1.8 mapped directly to the MLB domain. Only the data sources change — statsapi.mlb / FanGraphs MLB instead of KBO sources.

  1. 1. SP FIP (Fielding Independent Pitching)

    12%

    Category: Starting · range 1.50 ~ 6.00

    ERA estimated from strikeouts, walks, and home runs only — removes fielder influence to measure true pitcher skill. Lower is better.

    Why it matters: Starting pitcher matchup is the single biggest influence on a game — ranks #2 in 14-factor model weights.

    Source: FanGraphs MLB · KBO Fancy Stats

  2. 2. SP xFIP

    3%

    Category: Starting · range 1.50 ~ 6.00

    FIP with home run component normalized to league-average HR/FB ratio. Removes HR luck to measure true contact suppression.

    Why it matters: Eliminates HR noise from FIP → more meaningful in small samples. Supplementary layer to FIP.

    Source: FanGraphs MLB · KBO Fancy Stats

  3. 3. Lineup wOBA (Weighted On-Base Average)

    10%

    Category: Lineup · range 0.250 ~ 0.420

    Weighted average of run values for BB / HBP / 1B / 2B / 3B / HR. More accurate measure of offensive production than OBP.

    Why it matters: Lineup production = scoring capacity per game. Tied with SP FIP for top weight in the model.

    Source: FanGraphs MLB · KBO Fancy Stats

  4. 4. Bullpen FIP (Aggregate)

    10%

    Category: Bullpen · range 2.50 ~ 6.00

    Weighted average FIP for the team bullpen (all pitchers outside the starter). High leverage in later innings.

    Why it matters: Bullpen takes over after 5–6 innings — decisive in close games. High leverage factor.

    Source: FanGraphs MLB · KBO Fancy Stats

  5. 5. Recent Form (Last 10 Games)

    10%

    Category: Form · range -3 ~ +3

    Difference between last-10 win rate and season win rate. ±0 = season baseline, +0.3 = strong hot streak.

    Why it matters: Captures player condition, bullpen fatigue, lineup hot zones, and other daily variables. Fade-or-follow signal.

    Source: statsapi.mlb.com (boxscore aggregate)

  6. 6. WAR (Wins Above Replacement)

    8%

    Category: Record · range -2 ~ +12 (per player)

    Wins a player contributes above a replacement-level player. Team lineup WAR sum = total season talent.

    Why it matters: Long-term talent base. While recent form is short-term, WAR is the long-term layer.

    Source: FanGraphs MLB · KBO Fancy Stats

  7. 7. Head-to-Head (H2H)

    3%

    Category: Record · range 0.00 ~ 1.00

    Win rate in direct matchups between the two teams this season. Captures matchup quirks (ballpark / rivalry / specific pitcher vs lineup).

    Why it matters: Small sample (~13–19 games per season) but preserves matchup-specific patterns. Weight capped at 3%.

    Source: statsapi.mlb.com (schedule aggregate)

  8. 8. Park Factor (PF)

    4%

    Category: Park · range 85 ~ 115

    Run environment at the home ballpark (hitter-friendly vs pitcher-friendly). Coors Field ~115 / Petco ~95. 100 = neutral.

    Why it matters: Home ballpark adaptation = one axis of home advantage. Supplements wOBA / FIP ballpark adjustments.

    Source: FanGraphs MLB · ESPN Park Factor

  9. 9. Elo Rating

    10%

    Category: Rating · range 1300 ~ 1700

    Baseball adaptation of chess Elo — updates both team ratings after each game. Stabilizes after the second half of the season.

    Why it matters: Expresses team strength in a single metric. Information value Δ=+0.30 — highest among all factors (KBO backtest).

    Source: KBO Fancy Stats Elo · MLB FiveThirtyEight Elo (legacy)

  10. 10. Defense SFR (Skill-Free Runs)

    5%

    Category: Defense · range -30 ~ +30 (team)

    Runs saved by defense. Positive = runs prevented above average; negative = runs allowed above average.

    Why it matters: Defense component missing from FIP — SFR fills that gap. Noisy in small samples.

    Source: FanGraphs MLB Def · KBO Fancy Stats SFR

Statcast 4 Factors (MLB-exclusive layer)

Batted-ball measurement layer available from the MLB Statcast Era (2015~) — 4 factors not present in the KBO model. Detailed team measurements = /en/mlb/players.

  1. 11. Lineup xwOBA (Expected wOBA)

    5%

    Category: Statcast · range 0.250 ~ 0.420

    Expected wOBA from launch angle + exit velocity. Removes luck (defense / ballpark / weather) to measure true contact quality.

    Why it matters: Eliminates result noise from wOBA → more meaningful in small samples. Top Statcast weight in the 14-factor model.

    Source: Baseball Savant (Statcast Era 2015~)

  2. 12. Barrel %

    3%

    Category: Statcast · range 0% ~ 25%

    Barrel = minimum launch angle + exit velocity threshold = avg .500+ AVG / 1.500+ SLG expected. Percentage per plate appearance.

    Why it matters: Hard contact frequency = raw signal for home run + extra-base production capacity.

    Source: Baseball Savant (Barrel definition 2015 Tom Tango)

  3. 13. SP xwOBA-against

    4%

    Category: Statcast · range 0.250 ~ 0.420

    xwOBA of batted balls allowed by the starting pitcher. Lower = better contact suppression. Complements FIP.

    Why it matters: Batted ball quality layer beyond FIP's K/BB/HR. Key pitcher-side Statcast factor.

    Source: Baseball Savant (Statcast pitching)

  4. 14. wOBA Standard Deviation (variance)

    3%

    Category: Statcast · range 0.020 ~ 0.080

    Standard deviation of wOBA across lineup batters. Low = balanced lineup; high = stars-and-scrubs.

    Why it matters: Measures lineup depth. Same average wOBA but different σ leads to different run distributions.

    Source: FanGraphs MLB · KBO Fancy Stats

Home Advantage Bonus

Home Advantage (Elo Bonus)

10%

range +24 Elo (~ +3.4%)

Elo bonus of 24 points added to the home team = ~+3.4% win probability (Elo 400-point conversion). Covers ballpark familiarity, travel fatigue differential, and crowd factor.

Why it matters: Quantifies home field advantage. Applied consistently to every game.

Source: FiveThirtyEight Elo (MLB) · KBO own measurement +1.5%

※ These weights = MLB v1.0 (KBO v1.8 mapping + Statcast 4 additions). Will update when model evolves.

※ Weight source: packages/kbo-data/src/factors/mlb-base.ts. Shadow C learning cohort = walk-forward expanding window (milestones n=27 / 60 / 150 / 300 / 1000 / 2430).

※ KBO model = /methodology. Statcast team measurements = /en/mlb/players.