Value notes
Before you trust a big edge: the five-minute check
A computed edge of +8% should trigger verification, not excitement. The four checks that separate a real mispricing from a data error.
Your numbers say the price beats the fair line by 8%. That is not a green light; it is a question. Real edges of that size are rare on sharp markets, and a computed edge that large usually means the computation, not the market, is wrong. Four checks, five minutes, before the stake.
Check the market is complete
The fair price comes from removing the margin across every outcome, and a missing outcome inflates the fair prices of everything left. If your soccer market was read without the draw, the home and away fair prices are computed against a 1X2 sum that is missing a third of itself. The margin removal guide is explicit that removal needs a complete market, checked 2026-10-12, and a big edge is the first place that rule bites.
An edge that appears when you add a missing outcome was never there.
Check the prices are fresh
A stale fair price is the commonest source of phantom edges. If the reference price is from an hour ago and the market has moved since, your edge is measuring the move, not the mispricing. The same applies the other way: if your book's price is the stale one, the generosity you computed is a price that no longer exists.
Compare timestamps before you compare prices. Both sides of the EV comparison should be minutes old at most, and the drop feed exists precisely because minutes matter.
Check you are comparing the same thing
Same event, same market, same period. A first-half line against a full-match fair price, a main line against a spread at a different point: these produce dramatic edges on paper and nothing in reality. Verify the market definition character by character before you believe the number.
Check the direction of the surprise
The EV calculator example sits at +3.80%, checked 2026-10-12, and working value edges live in that neighborhood. When your computation says +8% or more, ask which side made the error before asking how much to stake: your model's probability, the fair-price removal, or the price you typed. If all three check out, the edge is real enough to size carefully, and the Kelly guide is the place to size it.
Big edges are rare enough to deserve suspicion. Verify the market, the freshness, the definition and the direction, and then the number has earned your stake.