Expected Value, and the trap of the positive number

"+EV" is the most quoted and least understood phrase in betting. A positive expected value is not a promise of profit — it is a claim about a probability, and that claim is only as honest as the model behind it. Here is what EV actually is, and why we tie every +EV we show to a baseline that is negative.

What is Expected Value?

Expected Value (EV) is the average result of a bet if you could place it an infinite number of times. The formula is simple: EV = probability × odds − 1. If a model says a team has a 55% chance to win and the price is 2.00, then EV = 0.55 × 2.00 − 1 = +0.10 — a positive expected value of ten percent per unit staked.

A positive EV means that, according to that probability, the price is generous: you are being paid more than the risk deserves. A negative EV means the opposite. Over a large enough sample, betting into positive EV grows a bankroll and betting into negative EV shrinks it. That much is just arithmetic.

But notice what the whole calculation rests on: the probability. Change 55% to 50% and the same 2.00 price is now exactly break-even. Change it to 48% and it is a losing bet. EV is not a property of the odds alone — it is a property of the odds and your estimate of the truth.

EV is only as honest as the probability behind it

This is the part the marketing never mentions. Anyone can generate a big, exciting +EV number by simply being optimistic about a probability. The bet looks brilliant on the spreadsheet and loses money in reality, because the edge was never in the price — it was in a wrong estimate.

So the only EV worth anything is EV computed from a probability that survives contact with the market. A calibrated model — one whose "70% chances" actually happen about 70% of the time over thousands of events — produces EV you can trust. An uncalibrated guess dressed up as a model produces EV that is worse than useless, because it feels rigorous.

The honest question to ask any +EV signal is not "how big is the EV?" but "how do you know the probability is right?" If the answer isn't a track record of calibration measured against real results, the EV is decoration.

The trap: +EV against the wrong reference

Here is the subtler mistake, and it catches experienced bettors. You can compute a perfectly real +EV against a soft bookmaker's inflated price — and still have no edge, because that soft price is not where the market truly settles.

The sharp closing line is the market's best estimate of true probability. If your model shows +8% EV against a soft book, but the sharp fair price implies your selection was already correctly valued, then your "edge" is an illusion created by comparing against a weak reference. The soft book will move, or limit you, and the phantom EV evaporates.

Real EV is measured against a sharp, margin-removed reference — the same devigged closing consensus used to measure Closing Line Value. This is why EV and CLV are two views of one thing: EV is the edge you claim at bet time; CLV is the receipt that proves the edge was real. A +EV bet that consistently earns positive CLV was genuinely value. A +EV bet that consistently loses to the close was a miscalculation wearing a confident number.

Why we show EV but anchor it to a negative baseline

Our terminal displays EV on every market. It also publishes, on its front page, the uncomfortable truth that our model's own CLV across mainstream markets is negative (1X2 −0.96%, AH −3.18%). Those two facts are not in conflict — they are the whole point.

A displayed +EV is a starting hypothesis: "by our calibrated probability, this price looks generous." It is not a guarantee, because across the whole board no model reliably beats the sharp close. The edge, when it exists, lives in narrow, disciplined subsets — and the only way to know whether a given +EV was real is to audit its CLV afterward, honestly, with sample size attached. We show you the EV, we show you the baseline that keeps you humble, and we give you the ledger to check which of your +EV bets actually earned their keep.

From EV to stake: where Kelly comes in

Knowing a bet is +EV tells you to bet; it does not tell you how much. Bet too little and you leave growth on the table; bet too much and variance ruins you before the edge pays off. The Kelly criterion answers the sizing question: it stakes a fraction of your bankroll proportional to your edge and inversely proportional to the odds.

In practice, because your probability estimate always carries error, most professionals stake at half-Kelly or less — deliberately under-betting the theoretical optimum to survive the inevitable stretches where the model was slightly wrong. A tool that outputs full Kelly without a half-Kelly guardrail is quietly encouraging you to over-bet an estimate you can't fully trust.

The honest chain is: a calibrated probability produces a trustworthy EV; a trustworthy EV, sized with (half-)Kelly, produces disciplined stakes; and CLV, measured afterward, tells you whether the whole chain was actually working. Skip any link and the confident number at the front means nothing.