TTP TAKE THE PITCH MLB · PAPER TRIAL
ONE LOOP · NIGHTLY · NO HUMAN JUDGMENT CALLS
THE METHOD
Everything published here comes out of one automated loop. This page explains it in plain English — and defines every term the ledger uses.

START WITH THE MARKET, NOT THE MODEL

Sportsbook prices are the output of a market that punishes bad numbers with real losses. The engine polls nine legal books, strips each one's vig — price both sides of a coin flip at −110 and each implies 52.4%, summing to 104.8%; the extra 4.8% is the book's cut — and takes the median. That's the consensus: the smartest freely available estimate of every probability.

WHAT AN EDGE IS

When the model's probability beats what the best available price implies, the difference is an edge. Not a prediction the bet wins — most positive-edge bets still lose. It's a claim the price is too generous, the way a 60-cent coupon on a fair 50-cent coin flip is worth collecting even though tails still comes up half the time.

WHY ALMOST EVERYTHING IS A PASS

Disagreements below 3% are ignored as noise. Markets with too few books, stale prices, or one-sided quotes are set aside rather than trusted. Passing is the engine's default output, on purpose: a model that finds edges everywhere has not found edges — it has found a bug in itself.

THE PAPER TRIAL

Rules committed before signal one, unadjustable mid-trial:
1. Every 3%+ disagreement is logged before first pitch, priced at the best number, zero dollars attached.
2. Every signal is graded twice — box score, and closing line.
3. At 200 graded: avg CLV above +1.5% → real stakes. At or below zero → retired, publicly. In between → back to the shop.

WHY CLV OUTRANKS WON–LOST

A won-lost record over a few hundred bets is mostly luck; plenty of bad processes run hot for months. CLV asks a harder question: did the market, once it absorbed every bet and every piece of news, move toward the model's number or away from it? Beating the close consistently is the one result that's very hard to fake — which is why this ledger is judged on CLV first and the record second.

GLOSSARY

implied probability
The chance a price implies: 1 ÷ decimal odds. −110 implies 52.4%. Book margins are baked in, so it slightly overstates the truth.
vig
The book's built-in margin — the excess over 100% when you sum a market's implied probabilities. The fee for being allowed to bet.
no-vig
Implied probabilities rescaled to sum to exactly 100% — the price with the book's cut removed. The market's honest opinion.
consensus
The median no-vig probability across every book quoting the same market at the same line. Treated as the market's estimate of truth.
edge
Model probability minus the implied probability of the best available price. Positive edge = the price pays out too often to be fair.
EV
Expected value: average profit per unit if the identical bet repeated forever. Positive EV is the only reason to bet; it guarantees nothing once.
closing line
The final price before first pitch — the market's most informed number, having absorbed every bet, injury report, and lineup card.
CLV
Closing line value: logged decimal ÷ closing decimal − 1. Consistently positive CLV is the strongest evidence a process sees something real.
unit
The fixed reference stake every result is measured in. Units keep the ledger about the process, not about anyone's money.
fractional Kelly
Betting a fixed fraction (a quarter here) of the growth-optimal Kelly stake — less growth, far less risk of ruin, and humility about the edge estimate.
paper trading
Logging every signal with no money attached and grading it as if real. The honest way to test a process before anything is risked.
calibration
Whether stated probabilities match reality over volume: of all the moments a model says 60%, about 60% should come true.
TTPUPDATED JUL 10 2026 · 14:59 UTC · REGENERATED BY THE PIPELINE
Research publication, not betting advice. Variance dominates short samples. 21+ where legal. If gambling stops feeling like entertainment, call or text 1-800-GAMBLER.