POSTERIOR · MLB RESEARCH
Back to the almanac

The Strikeout Lab

Strikeout props,
priced from the full distribution.

A pitcher's strikeout total is not a number — it is a distribution, and the Lab prices every rung of the K ladder off it. Season rate, recent form, opponent whiff and chase profiles, expected batters faced, park, and plate umpire feed a negative-binomial model whose dispersion decides how wide each pitcher's plausible night is. Every published pick settles in public, and the per-threshold record is on this page.

See today's boardAudit the accuracy

Model head

Neg. binomial

An explicit dispersion parameter carries honest game-level variance — aces have genuinely wider nights.

The ladder

3.5 – 9.5

Every threshold is priced off one distribution, then calibrated on its own settled track record.

Settlement

Public

Locked before first pitch, graded by the box score, released to the open ledger — losses included.

01 / Intent

The market

What a strikeout prop actually is

A strikeout prop is an over/under on one starting pitcher's final K count — 5.5 strikeouts means the over needs six. Books post a ladder of alternate lines around the main number, and the juice shifts along it. The bet is not “is this pitcher good”; it is “is the probability implied by this price wrong.” That reframing is the entire Lab: the model produces a probability for every rung, the no-vig consensus produces the market's, and only a meaningful gap at a fresh, deep quote publishes as a pick.

The model

Why a negative binomial, not a projection

Strikeout totals are over-dispersed count data: the same pitcher can punch out four or eleven without anything being wrong. A single projected number hides that. The K head blends the pitcher's season strikeout rate, recent form, and the opposing lineup's strikeout tendency — each shrunk toward league priors by sample size — then adjusts for whiff rate, chase rate, expected batters faced, workload, park, and umpire. The output is a full negative-binomial distribution, and each ladder threshold reads its probability directly off it. The methodology in full lives at /methodology.

The filter

Most K props never publish

The Lab prices the whole slate every day, and most of it is priced correctly — no edge, no pick. A row publishes only when the model-versus-consensus gap clears the per-market threshold at a fresh quote from an allowlisted book, with the starter confirmed. On days when nothing qualifies, the board says so instead of forcing coverage. The silence is archived with reasons, so an empty day is as auditable as a winning one.

02 / How it works

  1. Step 01

    Build the K distribution

    The negative-binomial head turns pitcher rate, opponent K profile, expected batters faced, park, and umpire into a full strikeout distribution with an explicit dispersion parameter.

  2. Step 02

    Price every rung

    Each ladder threshold (4+, 5+, 6+ strikeouts, …) reads its probability off the same distribution, then gets exact-threshold calibration on its own settled history.

  3. Step 03

    Compare to the no-vig market

    Allowlisted book prices are stripped of vig into a fair consensus. A pick exists only where the model and the market disagree by enough, at a quote fresh enough to trust.

  4. Step 04

    Lock, settle, release

    Qualified picks lock publicly before first pitch and grade off the box score. Settled rows release to the open ledger after the public delay — wins, losses, and the ladder record above.

Byron Buxton laying out for a catch in center field

Public proof

Two pitchers, same projection, different prices.

A 6.4-K mean with tight dispersion and a 6.4-K mean with a wide one are different bets. The distribution is the product.

03 / Questions

What are MLB strikeout props?

A strikeout prop is an over/under wager a sportsbook offers on a starting pitcher's final strikeout count in one game — for example, over 5.5 strikeouts needs six or more. Books post alternate lines above and below the main number, forming a ladder (4+, 5+, 6+ strikeouts and so on) with different prices at each rung.

What are the best MLB strikeout props today?

Posterior does not keep a permanent list — the honest answer changes daily with pitching matchups, lineups, and prices. The model prices every rung of every starter's K ladder each day and publishes only rows whose model probability meaningfully beats the no-vig market consensus at a fresh quote. Today's qualified board is at https://posterior.pro/today, and on days when nothing clears the bar the board says so.

How accurate is the strikeout prop model?

The settled record is public and split by ladder threshold on this page, drawn from the same delayed ledger anyone can download at https://posterior.pro/data. The fuller audit — win rate by tier, Brier score, and the calibration curve — lives at https://posterior.pro/mlb/accuracy. Posterior publishes losses alongside wins and treats calibration, not raw win rate, as the load-bearing claim.

Are the strikeout picks free?

The settled record is free: every resolved K pick releases to the public ledger after a delay, with model probability, price context, and outcome, downloadable as CSV under CC BY 4.0. Live and pending picks — the rows that have not played yet — are part of Posterior Pro. Posterior is baseball data, not a sportsbook: it does not accept or place wagers.

Why do different strikeout thresholds have different win rates?

Because they are different bets. A 4+ strikeout line settles over far more often than an 8+ line, and their prices reflect that — so win rate at one rung says nothing about another. That is why the Lab reports the record per threshold and prices each rung from the full K distribution with its own calibration, rather than projecting one number and rounding.

Posterior is baseball data, not a sportsbook. The public receipt lives at /data, the methodology lives at /methodology, and today's board starts at /today.