Transparency

How EdgeBot works.

You should be able to see what EdgeBot saw, when it saw it, what the market was doing, and what happened. This page explains every rule behind the numbers on this site.

Where the numbers come from

How an edge is calculated

For every game EdgeBot builds a fair price: the chance of each outcome with the bookmaker's cut removed. It comes from the sharpest prices available (Pinnacle, plus any exchange whose market is tight), blended with EdgeBot's own rating model for that sport.

Then it compares the fair chance with what Hard Rock pays. Edge = fair chance × payout − 1. An edge of +5% means that, over many bets like it, you'd expect about $5 back per $100 bet. A play is only published when the edge clears the bar for its sport (3%+ for model picks).

Probability, edge and confidence are different things. Probability is the chance the bet wins. Edge is how much better the price is than that chance deserves. Stars show the size of the edge. A 40% longshot can carry a bigger edge than a 70% favorite.

What gets thrown out: when the sharp market is far away from EdgeBot's number (more than 15 points of win probability), the market almost always knows something the model can't see: a late scratch, a pitcher change, a game already in progress. Those are never published, however big the edge looks.

Tested and untested sports: every sport's model is tested on past seasons against real closing lines. Sports that haven't tested profitably yet are labeled untested on the member board and are never used for the free pick.

How picks are published

How picks are graded

Units and ROI

A unit is 1% of a bankroll. EdgeBot sizes each play with a quarter of the Kelly formula (bigger edge, bigger bet), up to 2 units on sides and totals and 0.5 units on player props and first-half totals, with limits of 3 units on any one game and 10 units in a day. Each desk's report card (below) then scales its bets up or down. Units won is the sum of every graded pick's profit. ROI is units won divided by units risked.

Beating the close (CLV)

The closing price is the market's final, most-informed number right before the game starts. If you consistently bet at better prices than the close, you have an edge, whatever the short-term wins and losses say. EdgeBot measures every side and moneyline against the sharp closing price (Kalshi's market at kickoff) and reports the average as Beat the close. Above zero over a few hundred bets is the strongest evidence of a real edge.

Every pick is checked before it's a bet

Before a pick is published, and again every 30 minutes until kickoff, it runs a checklist:

A pick that passes everything is VERIFIED (full size). One where a second opinion disagrees is PARTIAL (half size). Anything waiting on a starter is PENDING, and anything that fails a hard check is BLOCKED and never a bet.

EdgeBot grades itself

Every morning EdgeBot grades each desk (one sport and one kind of bet, like NFL price edges or WNBA player props) on its last 120 days, using two honest tests: did its bets beat the sharp closing price, and did they win as often as the model said they would? A desk that passes gets bigger bets. A desk that fails gets smaller bets and has to show more edge before anything is published. A desk with fewer than 15 graded bets is still learning, and nothing changes: that's too few to tell skill from luck. The grades only size new bets. Nothing on the ledger is ever removed.

DeskGradeRecordUnitsBeat the closeWon vs predictedBet size
Loading the report card…

Grades: STRONG bets ×1.5 · BOOST ×1.25 · NORMAL · CUT ×0.5 and +2% edge required · PROBATION ×0.25 and +3% edge required · LEARNING (not enough bets yet).

Postponed games and voids

What this site is not

EdgeBot is not a sportsbook. It never takes bets, never holds money and cannot guarantee a profit. Sports betting carries real financial risk, and even a real edge loses often in the short run. Please read Responsible Gaming.