Sports Edge Model — every game called, every call graded, in public

The Lab · Foundations · updated 2026-10-22

Start here — the Lab in plain English

New to betting, a regular, or a numbers person — here's where to begin, every term the Lab uses in one place, and the record we've published, dated, beside today's.

The short version

A betting line is a price, and the price is usually right. Everything in the Lab grows from that one idea. Pick a reading path below that matches how much you already know.

Record figures marked as today's come live from the public record, refreshed as games are graded — as of Oct 30, 2026. Numbers we published on earlier dates stay as written, dated, so you can check them against today's.

The Lab explains how betting markets work, using the same math our board runs on and the record we grade in public. Nothing here is a pick. It's written for four kinds of reader, so start wherever you fit.

Pick your path

New to betting. Start with how to read a line, then what it costs to bet, then the favorite-or-underdog question everyone asks first. What a line really says → The vig → Favorites or underdogs? → How to track your bets

You bet sometimes. You know the basics and want to know what actually moves the needle — prices, props, parlays. Which sportsbook has the best odds? → Same-game parlays → Player props → The closing line

You bet seriously. You want to know what's been tested, what failed, and what's still standing. Seven ways we tried to beat the market → Why books disagree → Does line movement mean anything? → Why we stopped posting long shots

You read regression tables for fun. Calibration, power, and grading a probability rather than a pick. How often do favorites win? → How many games until a record means anything? → The Brier score → Reading a null result

The words, once

Every term below is used the same way in every article.

Term What it means
Moneyline A bet on who wins. −150 means risk $150 to win $100; +130 means risk $100 to win $130.
Implied probability The win rate a price demands to break even. −150 → 60.0%; +130 → 43.5%. More
Overround How far the two sides' implied probabilities add past 100%. −110/−110 adds to 104.8%, so the overround is 4.8 points.
Hold What the book keeps from every dollar bet, on average, if it balances its action. At −110/−110 that's 4.5%. Overround and hold are two views of the same fee; this Lab calls the fee the vig. More
De-vig / fair probability The implied probabilities with the fee taken out, so they add to 100%. It's the market's honest opinion — and our board's number.
Edge Your probability minus the price's break-even. Real edges are a point or two, and rare.
Expected value (EV) What a bet returns on average per dollar: edge × payout. Long prices multiply small edges and small errors. More
Closing line / CLV The last price before the game starts, and how much better your price was than it. The fastest honest grade a bet can get. More
Unit A fixed slice of your bankroll, usually 1%. Units grade decisions; dollars grade your bankroll. More
Kelly (fractional) Sizing a bet by how big its edge is. We stake a fraction of it, so one wrong estimate can't hurt much.
Calibration When a model says 60%, does it happen about 60% of the time? Answerable in hundreds of games.
Brier score The average squared miss of a probability. Lower is better; a coin flip scores 0.25. More
Standard error How far a number could be off from luck alone. Two of them either side is the usual bar. More
Pre-registration Writing a test's rules down before the games are played, so nobody can choose the test after seeing the result.
Same-game parlay Several bets from one game, paid together. The legs move together, so the book prices them its own way. More
Exchange A venue where bettors take each other's sides and pay a small fee instead of a vig. We include Novig in our consensus.

The record, then and now

Every figure about our board that the Lab has published, with the date it was published, and today's live figure beside them. The board's "favorite" is the side the market gives more than 50%; "claimed" is the average probability it gave them.

PublishedGamesFavorite wonMarket saidGap (pts)
Jul 27, 202626654.9%55.8%−0.9
Jul 28, 202631154.7%55.9%−1.2
Aug 1, 202633254.5%55.9%−1.4
Aug 15, 202651955.1%56.6%−1.5
Oct 21, 20261,43958.9%58.2%+0.7
Today live1,59559.7%58.6%+1.2

The sample grew more than fivefold, and at every date we published, the gap between what the market claimed and what happened was within about a point and a half. That's the market keeping its word — and the reason this site doesn't sell picks against it. The same trail for the few picks we've published, which still prove nothing either way:

PublishedPicksRecordAvg CLVBeat the closeROI
Jul 27, 20261—+16.43%——
Aug 15, 202632–1+3.66%—+97.9%
Oct 21, 2026146–8+6.58%64%+55.6%
Today live146–8+6.58%64%+55.6%

What's running right now

Every test below had its rules written down before its games were played. Results get published whichever way they land.

Model output is informational and entertainment content, not betting or financial advice. If you bet, bet what you can afford to lose.

Track your own bets against this board

Log what you actually bet — units and dollars — and see it graded beside a model that publishes every call it makes, wins and losses alike. Free, no card.

Start a free tracker

← Back to The Lab · How the model works · The record