April 3, 2026 — MLB Weather Features

This update adds real-time weather data to every PropsBot MLB player prop prediction.

On this Page

What Changed

PropsBot’s MLB models now account for game-time weather conditions. Temperature, wind, humidity, and precipitation are now included in every MLB player prop prediction.

Stadium-Relative Wind

Raw wind data is converted into a stadium-relative direction by comparing the wind bearing against each ballpark’s home-plate-to-center-field orientation. The result is one of six categories: blowing out (aids fly balls), blowing in (suppresses offense), or blowing across to either side. Wind blowing out at Wrigley is a very different signal than wind blowing in — the models can now tell the difference.

Dome Detection

Eight MLB venues have domes or retractable roofs — Chase Field, Tropicana Field, Rogers Centre, loanDepot park, Minute Maid Park, American Family Field, T-Mobile Park, and Globe Life Field. For these games, outdoor weather data is ignored — the models learn that weather doesn’t matter when the roof is closed.

Precipitation

A binary flag identifies games played in rain or snow conditions, which can affect grip, footing, and overall offensive output.

Updated Markets

All 18 MLB markets in PropsBot receive weather features — both batter props (Hits, Runs, RBIs, Home Runs, Total Bases, etc.) and pitcher props (Strikeouts, Hits Allowed, Earned Runs, etc.). Unlike the Batter vs. Pitcher features added last month, weather affects everyone on the field.

Why This Matters

Weather is one of the most well-documented environmental effects in baseball. Cold temperatures reduce ball flight distance, wind blowing out at Wrigley can turn warning-track flyouts into home runs, and humid conditions affect pitcher grip. Research consistently shows that wind speed and direction can shift home run rates significantly at outdoor stadiums. Previous PropsBot MLB models were blind to all of this — now they’re not.

What Didn’t Change

  • NFL, NBA, and NHL models are unaffected
  • All other MLB inputs — rolling stats, pitcher strength, platoon splits, BvP H2H, performance states, game odds — unchanged
  • Model architecture and validation process remain the same