Best MLB Earned Runs Matchups — Wednesday, August 19, 2026
Top spot: Jeffrey Springs
The model's pick for the quietest line of the night: Jeffrey Springs (ATH), projected around 1.8 ER against KC. A low earned-run projection is really a statement about the lineup — who it strands, who it can't string hits against.
The rest of the top of the board
- Walbert Ureña (LAA) (100) — about 1.8 ER vs HOU.
- Roki Sasaki (LAD) (100) — about 2.7 ER vs COL.
- Parker Messick (CLE) (100) — about 1.4 ER vs SF.
- Sean Newcomb (CWS) (100) — about 0.9 ER vs CHC.
- Drew Rasmussen (TB) (84) — about 1.7 ER vs TOR.
How it played out
The top 10 starts averaged 2.2 ER. Jeffrey Springs finished with 2. Every projection is graded against the box score.
Earned runs are a sequencing stat
A starter can give up six hits and one run, or three hits and four — earned runs are what happens when traffic clusters. That makes ER the noisiest of the pitcher props, and it's why the model builds the number from the collision of skills: how much traffic this lineup creates against this arm's profile, how the park treats contact, and how often this pitcher strands what he allows. One more honest wrinkle: a starter's earned runs stay his after he leaves, so a shaky bullpen inheriting his runners quietly inflates his line. Small numbers, wide error bars — respect both.
How to read the earned runs board
Each starter's 0–100 score reads his run-prevention setup tonight — the traffic this lineup figures to create against him, the park, and what he tends to strand. The projected ER beside it is the expected earned-run count for the start. Books usually hang these lines at 2.5: our number beside theirs shows which side the model leans, and the box score grades it either way.
What the earned runs allowed board is
The Earned Runs board projects how many earned runs a starter gives up vs his line. One statistical model scores every matchup on the slate the same way — a star and a backup judged on the matchup in front of them, not their name — and every number is graded against the real box score once the games go final.
How the model gets its number
It isn't a gut call or a name game. The projection is built from a few things:
- His run-prevention rate vs the lineup's hand.
- The park.
- Projected length of outing.
Those pieces combine into one number, and the model re-tunes itself weekly against how its past calls actually landed.
Is there a betting edge here?
Efficiently priced — transparency board. It's also a core input to our team run model (moneyline + team totals), where the edge actually lives.
How to use it
Read it for who's in line for a clean or rough outing; the bettable version is the team-total board it feeds.
Everything here is a research signal from the model, graded in public — not betting advice, and no outcome is guaranteed.
Learn more: How the model works · The picks story · Today's plays