POSTERIOR · MLB RESEARCH

MLB parlay model

Parlays from picks,
not from decoration.

Posterior parlay candidates are built from the day's official model selections and nothing else: two to four legs, never two from the same game, priced on an independent-probability baseline, and published only when a real sportsbook quote for the exact combination was captured. Everything else is a lesson in why most parlays are bad bets.

See today's picks Audit outcomes

Leg count

2-4

Short enough that every leg's edge still matters; long enough to price a real combination.

Source

Picks

Candidates are built only from the day's qualified official selections — no parlay-only legs.

Pricing

Captured

Official status requires a real same-book quote for the exact combination, captured at lock.

Today's parlay inputs

The leg pool, before any combining

Refreshes every 5 minutes

A Posterior parlay never starts from a slip. It starts from the day's official singles board — the picks already published on their own evidence — and combines a small number of them under rules designed to keep the independence math honest. The figures below are that pool. When the slate produces few qualified legs, the correct output is fewer tickets, or none; the combination layer has no quota. Singles live at /today.

Underlying singles record

1586-1015

61.0% across 2601 settled official picks. A parlay is only as honest as the legs inside it.

Ticket-level detail is subscriber content inside Posterior Pro. What stays public is the discipline: leg counts, construction rules, and the settled singles record underneath.

01 / Intent

Skepticism

Why most parlays are bad bets

Every leg carries the sportsbook's margin, and multiplying legs multiplies the vig. Three legs at a fair coin-flip price pay 7-to-1 only if the book shaved nothing; at typical -110 pricing the true payout should be higher than the one on the slip, and the gap grows with every added leg. Worse, bettors routinely combine legs that move together — same game, same team, same pitcher — while the payout math quietly assumes independence. The product is engineered to be fun and to lose. Any honest parlay model starts there.

Pricing

Independent baseline, then a real quote

Posterior first prices a candidate the strict way: calibrated leg probabilities multiplied together, leg decimal prices multiplied together, theoretical EV from the pair. Three legs the model rates at 55% each have a joint probability near 16.6% — a humbling number the slip never shows you. That theoretical price is a baseline, not a claim. A ticket only becomes official when an actual sportsbook quote for that exact combination was captured atomically at lock, so the published EV is measured against a price that really existed.

Correlation

One game, one leg — no exceptions

Two legs from the same game share a pitcher, a lineup, weather, and a game script; treating them as independent overstates the joint probability, which is precisely the bug that makes same-game parlays look generous. So the construction rules are blunt: at most one leg per game, at most two legs per team, and every event identity must be proven before multiplication is allowed. The same-game endpoint does not multiply marginals or apply heuristic pair multipliers — until a promoted joint simulation or an exact same-book SGP quote exists, it returns a zero-stake WAIT receipt.

Separation

Theoretical rows never touch the headline

When no book quote for a combination was captured, the candidate can still exist as a theoretical explorer row — visible as a diagnostic, staked at zero, and quarantined from every public accuracy figure. Official parlay results are computed only over tickets that carried a captured same-book quote, and exploratory rows are rolled up separately so a speculative slip can never inflate the record. The same separation applies to stake sizing: a recommended stake is a fraction derived from the captured price and the combined edge, not a uniform unit slapped on every slip.

02 / How it works

Step 01
Collect qualified legs
Only the day's official selections enter the pool — rows that already cleared edge, price-freshness, and quality gates as singles. A pick that was not good enough alone does not become good enough inside a combination.
Step 02
Diversify before combining
The pool is seeded across market families so the menu is not four copies of one idea, then filtered by the correlation rules: one leg per game, at most two per team, proven event identities throughout.
Step 03
Price on the independent baseline
Combined probability is the product of the legs' calibrated probabilities; the combined price is the product of their decimals. Candidates are tiered by joint probability — 30% and up reads STRONG, 18% and up reads LEAN — because parlay variance is harsher than singles variance.
Step 04
Publish only with a captured quote
A candidate goes official only if the exact combination was quoted by a real book and captured at lock; otherwise it stays a theoretical explorer row and out of the public headline. Each leg remains tied to its individual pick record for settlement.

Junior Caminero flipping his bat after contact

Public proof

A parlay is only as good as its weakest leg.

Posterior does not add legs just to make the slip look busy. Unqualified rows stay out.

03 / Questions

How many selections are in a Posterior parlay?

Two to four. Below two it is not a parlay; above four the compounded vig and the joint-probability collapse stop being defensible for a model whose entire premise is honest pricing. Each candidate is also labeled by risk tier — two-leg tickets as conservative, three as balanced, four as upside — so the shape of the bet is visible before the payout.

Will every slate have parlays?

No. The combination layer can only draw from that day's qualified official picks, and a thin slate often cannot field enough uncorrelated legs. In that case the correct publication is fewer tickets or none — the system has no daily quota, and forcing a combination from unqualified legs would contradict the reason the model exists.

Are same-game parlays included?

Not as multiplied marginals, because that math is wrong: legs in the same game are correlated, and the independence assumption would overstate the joint probability in both directions. Posterior's same-game boundary accepts only canonical same-event legs and returns a zero-stake WAIT receipt until a promoted joint-probability simulation or an exact executable same-book SGP quote exists to price the correlation honestly.

Why do most parlays lose money?

Three compounding problems. The book's margin is charged on every leg, so the house edge grows with the slip. The advertised payout is computed as if legs were independent, while the popular combinations usually are not. And the hit rate collapses faster than intuition suggests — even a model that is right 60% of the time per leg wins a four-leg ticket only about 13% of the time. Posterior's answer is fewer, shorter, provably priced tickets.

Where can I see today's parlay candidates?

Qualified combinations are published inside Posterior Pro alongside the singles board at https://posterior.pro/today, each tied to the individual pick records of its legs. The public side of the story is this page's discipline plus the settled singles ledger at https://posterior.pro/data — the legs' outcomes are auditable even where the tickets are subscriber content.

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.

Parlays from picks,not from decoration.