Understanding what the batter and pitcher projection numbers mean — and how to use them for smarter analysis
Updated daily throughout the 2026 season
⚠️ Important: These projections are for informational purposes only. They are not a guarantee that tonight's projection will hit — baseball outcomes carry significant variance even with accurate models. Please review the accuracy stats on the linked pages to understand current performance before using these projections.
OddsRX tracks how accurately the model projects individual player stats each game — both for batters (Hits, Home Runs, Extra Base Hits, Walks, RBIs) and starting pitchers (Innings Pitched, Runs Allowed, Strikeouts).
For every game with a lineup projection, we compare what the model predicted for each player to what actually happened in the box score. Those comparisons accumulate into accuracy statistics that show where the model is reliable — and where it tends to go wrong.
Results are broken into three views: Overview (league-wide averages per metric), By Team (which teams are easier or harder to project), and By Player/Pitcher (individual accuracy with full search).
"On average, how many units off was the projection?" An MAE of 0.8 on Hits means the typical projection landed about 0.8 hits away from actual. Lower is better. This is the primary accuracy summary — it measures the size of misses without caring about direction.
Whether projections consistently run too high or too low. Positive bias = model overprojects. Negative bias = model underprojects. You can have decent MAE but still lean in one direction — and that lean matters when comparing projections to prop lines.
The average projection vs the average real-world result across all graded games. A quick sanity check — if projected averages run significantly higher than actual averages, expect positive bias. These should converge over a full season as the model calibrates.
How many games are in the sample for this player or team. Under 10 games, treat the numbers as preliminary — they can swing on a single outlier. Results stabilize meaningfully around 20–30 games per player.
Projected hits per game based on the batter's grade, the opposing pitcher's grade, platoon splits (vs LHP/RHP), and park factors. The highest-volume stat and typically the first to stabilize. An MAE of 0.5–1.2 is a reasonable baseline as the season matures.
HR projections are naturally low-frequency — most players project under 0.25 HRs per game. MAE will look small in absolute terms but large relative to the projection. Bias matters more here: consistent overprojects on HRs often signals an issue with power assumptions in the grade model.
Doubles + triples + home runs combined. A useful middle ground between raw hits and HR projections — captures extra-base ability without the volatility of HR alone. More stable than HRs, less stable than hits.
Walk projections lean on the batter's plate discipline grade and the pitcher's control tendencies. More stable than hits because they're less affected by defensive variance. Negative bias (model over-projecting walks) is common early in a season before current-year tendencies are established.
The hardest batter stat to project. RBIs depend heavily on lineup context — runners on base, batting order spot, and situational hitting. Expect the widest MAE and most variable bias, especially early. RBI accuracy improves as more lineup context accumulates across the season.
Hits and Walks tend to be the most stable projections and the best for comparing to market prop lines. HR and RBI carry the most variance — treat those as directional signals rather than precise forecasts.
Projected innings based on the starter's grade and historical workload patterns. IP is the foundation of pitcher projections — when IP is off, runs and strikeouts tend to drift too. Positive bias (over-projecting innings) typically means the model underestimates how quickly a starter gets pulled due to command issues or a blowout score.
Projected runs against based on the opposing lineup grade, platoon matchups, and park factors. The most consequential pitcher projection for game-level outcomes. Negative bias (under-projecting runs) is common when opposing lineups outperform their grades — which happens regularly early in the season before grades stabilize.
Projected Ks based on the pitcher's strikeout rate grade and the opposing lineup's contact tendencies. Strikeouts tend to be one of the more consistent projections because pitcher K-rate is a sticky skill. MAE in the 1.5–2.5 range per start is typical.
Strikeouts tend to be the most projectable pitcher stat — K-rate is one of baseball's most stable skills. Runs allowed carries the most variance and can swing significantly based on sequencing and hot streaks that no model can fully anticipate in advance.
The By Team tab breaks out accuracy by team, helping you spot which teams are consistently easier or harder to project. Some teams have stable, predictable lineups — same nine guys, consistent batting order, no platoon decisions. Others rotate frequently, use aggressive matchup management, or have unpredictable rest patterns.
When a team shows higher MAE across multiple stats, it's a signal that player props involving that team carry more uncertainty than average. Use the team view as a quick reliability check before comparing projections to market lines.
The By Player / By Pitcher tab is where this becomes most actionable. Search for any batter or pitcher to see their individual accuracy history across all graded games.
Some players are easier to project: everyday starters with stable roles, consistent batting order spots, and predictable usage. Others are harder: platoon players, injury-managed situations, or starters on pitch count restrictions.
Hits and Strikeouts are typically the most consistent categories. RBIs and Runs Allowed carry the most variance and are best used as directional context rather than precise targets.
These projections are most useful when you combine them with the accuracy data — a player with low MAE and low bias is a much stronger comp to a market line than one with high variance or a known directional lean.
This page is a transparent performance tracker showing average error (MAE), directional lean (bias), and projection vs actual over the course of the season. It is not a guarantee that tonight's projection will hit. Baseball outcomes carry significant variance even with accurate models.
The page is designed to answer: When the model is wrong, how wrong is it — and does it lean high or low? That context makes every projection more useful.
