What Drives MLB Run Totals: The Factors Ranked

The BullpenEdge model is a regression, which means every input has a measured effect. The cleanest way to compare them is 'how much does a typical, one-standard-deviation difference in this input change a team's expected runs?' Using the model fit on 2019–2025:

• Opposing pitchers' strikeout rate: about −5.7% (more strikeouts, fewer runs).
• Ballpark: about +4.4% per standard deviation — and Coors Field is more than three standard deviations above average.
• Temperature: about +4.3%.
• Team's own recent run scoring and the opponent's runs allowed (regressed): about +3% each.
• Lineup projected wOBA and lineup Statcast quality of contact: about +2.7% each.
• League-wide scoring level this season: about +3%.
• Wind blowing out: about +1.7%.
• Platoon advantage and team splits: under 1% each.
• Home-plate umpire: about 0.7%.
• Bullpen rest beyond who is available: well under 0.1%.

Reading the list

These are average effects. On a given night one input can dominate: a 95°F afternoon with the wind blowing out at Wrigley, or a strikeout ace against a lineup full of hitters without the platoon advantage. Each game page shows those factors for that specific matchup, and Run Environment ranks the whole slate.

How good are run projections?

Honestly, not great — nobody's are. The actual total of a single game has a standard deviation of about 4.5 runs, so even a perfect model misses by more than three runs on an average night. In our backtest the projected total missed by 3.53 runs on average, only slightly better than the 3.54 you get from the league average. Totals are the hardest thing in baseball to forecast, and we would rather say so than pretend otherwise.