POSTERIOR · MLB RESEARCH
Back to the almanac

Closing line value tracker

Did the model beat
the later market?

Closing-line value answers one narrow question: was the price Posterior locked better than the price the market offered at first pitch? It is not a trophy and it is not profit — it is evidence about whether the model is finding numbers before the market finishes moving. This page shows the real counts.

Open public ledgerSee settled returns

Signal

CLV

Market movement after lock is tracked as an audit signal where a closing snapshot exists.

Outcome

W/L/V

Wins, losses, and voids still decide settlement. CLV does not replace results.

Calibration

Brier

Probability quality is evaluated with calibration and Brier-style evidence.

01 / Intent

Definition

Entry price versus closing fair

When a pick locks, Posterior stores the exact quote: book, price, and timestamp, frozen in the decision archive. Near first pitch, the market's closing state is captured and stripped of vig to produce a closing fair probability. Textbook CLV is the comparison of those two numbers. If you took -110 and the fair close says the same side should have been -130, you captured price — regardless of what happens in the game.

Honesty

CLV and ROI are different questions

Realized ROI measures what settled picks returned at the prices that were actually available. CLV measures whether those prices beat the closing number. They can disagree for weeks: a model can consistently beat the close and still lose money while variance runs against it, or win money at prices that were worse than the close and be living on borrowed time. Posterior's own audit found the two figures telling different stories, which is exactly why both are published instead of averaging them into one comfortable number.

Why it matters

The close is the market's final answer

By first pitch, lineup news, weather, and sharp money have finished repricing the market, so the closing number is the best public estimate of true probability. A model that never beats it is a reader of the news, not a forecaster; one that beats it at a meaningful rate is finding prices before the information arrives. That is why Posterior tracks whether each locked price beat the number at first pitch — it is the closest thing betting has to a point-in-time honesty test, and it cannot be gamed by cherry-picking outcomes.

Limits

What CLV cannot tell you

Two rows in the sample are not two independent experiments: picks from the same game share information, and a week of closes can move together when one news pattern repeats. The count above is the whole graded population — no minimum-streak filter — but even a few hundred picks is a small sample for a two-point edge, and the roll-up excludes every market the model declined. Posterior treats a strong CLV run as a reason to keep auditing, not as permission to relax the gates, and the settled record at /mlb/accuracy remains the final word.

02 / How it works

  1. Step 01

    Record the locked view

    At publication, the model probability, entry price, book, and quote timestamp are frozen into an immutable archive. The entry side of the CLV comparison can never be revised to look better later.

  2. Step 02

    Capture the close

    Near first pitch, a closing snapshot records the market's final pre-game state and converts it to a no-vig fair probability. Rows without a trustworthy closing snapshot stay out of the CLV sample instead of being estimated.

  3. Step 03

    Settle the outcome

    The game grades the row won, lost, or voided. Settlement answers 'did it win?' — a different question from 'was the price good?' — and both answers are kept.

  4. Step 04

    Audit the full pattern

    Nightly, the compute job rolls CLV up across the resolved population. The public record keeps CLV, Brier score, win rate, and realized ROI visible together, so no single flattering metric can carry the story.

Shohei Ohtani swinging during an MLB at-bat

Public proof

A good process can still lose a game.

That is why Posterior separates market movement, calibration, and final settlement instead of pretending one metric explains everything.

03 / Questions

What is closing-line value?

Closing-line value compares the price you captured when you bet against the price available when the market closed, usually at first pitch. If a pick locked at a better number than the close, it has positive CLV. Because the closing price is the market's most informed estimate, consistently beating it is the standard evidence sharps use to show a real edge rather than a lucky stretch.

Does positive CLV guarantee profit?

No. CLV is process evidence, not a settlement outcome. A pick can beat the close and lose the game — over a small sample that happens constantly, which is why Posterior publishes CLV next to realized ROI, win rate, and Brier score instead of presenting it as a profit claim. The ledger at https://posterior.pro/data carries the settled results.

Why do Posterior's CLV and ROI figures disagree sometimes?

Because they measure different things. Price capture compares the entry price to the closing fair number; realized ROI compares entry price to what actually happened. Short windows routinely produce one positive and one negative. Persistent divergence in either direction is informative: beating the close without profit suggests variance; profiting without beating the close suggests the run may not be repeatable.

How often is CLV computed?

A nightly compute job grades every resolved pick that has a usable closing snapshot, after the day's closes are captured. The figures on this page are that roll-up — the same canonical population behind the public accuracy record, not a separate, friendlier sample.

Can I verify the record myself?

Yes. The delayed public CSV at https://posterior.pro/data/picks.csv lists resolved rows with model probability, market probability, grade, price, and outcome. CLV inputs live in the same archive discipline: the locked quote and its timestamp are frozen at publication, so the entry side of the comparison cannot be back-fitted.

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.