PropsBot added several new MLB batter feature sets covering contact quality, hit-luck adjustments, deeper pitcher matchup info, and game context from the batter’s own perspective. Five batter markets have been promoted to use the new feature set.

Models now see how hard each batter actually hits the ball, in addition to whether he gets hits. Features include exit velocity, hard-hit rate (contact at 95+ mph), expected batting average (xBA, derived from launch speed and angle), line-drive rate, and contact rate on swings. A batter making consistent loud contact but running into bad luck looks very different from one making weak contact — the models now have indicators for this.

Batting Average on Balls in Play (BABIP) and per-game hit rate are now tracked separately. BABIP isolates skill from luck on balls put in play, which is one of the best leading indicators of regression — overperformers tend to cool off, underperformers tend to bounce back.

The opposing-pitcher feature set gains strikeout rate (K%) and expected batting average allowed (xBA). The platoon-split feature set adds wOBA — weighted on-base average — split by pitcher handedness, alongside the existing AVG/OBP/SLG splits. wOBA captures the full offensive profile in a single number, weighting walks, singles, doubles, triples, and home runs by their actual run value.

Batter models now see implied team totals and spread from the batter’s own team’s perspective, not just from a home/away split. A batter on a 5.5-run-implied-total team is in a very different environment than one on a 3.5-run team — the model now sees the right number for the right team automatically.

Five batter markets have been promoted to use the new feature set as their champion:

• Total Bases
• RBIs
• Walks
• Singles
• Runs Scored

The new features were tested across all 12 batter markets, but only those five showed clear improvement over the previous champions. The rest keep their prior models for now. Pitcher markets are unaffected — these features describe the batter or the matchup he faces, not the pitcher’s own work.

Traditional batting stats (AVG, hits per game) tell you what happened. Quality stats (xBA, hard-hit rate, exit velocity) tell you what should have happened — and what’s likely to happen next. Combining the two lets the models distinguish hot streaks and cold streaks from real improvement and decline.

The pitcher K% and xBA additions, plus the team-perspective game context, sharpen the matchup picture around every at-bat.

• Pitcher prop models — unchanged
• NFL, NBA, NHL models — unaffected
• Other MLB inputs (BvP H2H, weather, performance states, game odds totals/spreads, rolling stats) — unchanged
• Model architecture and validation process — same as before