Working the Positive EV board
Which prices on the board are better than the bet is actually worth, and how much of your bankroll each one justifies.
The screen, annotated
Every column, and what its number actually means.
The edge: how much of the stake this price is expected to return, on average, above breaking even. It is a statement about the price and never about this particular bet, which you will still lose about half the time.
A badge on rows below 1%. The engine detects down to 0.5% net of fees. These rows are real, but they are line-shopping money rather than anything to size up. They only appear once you lower the min edge filter yourself.
The engine has decided this edge is larger than book disagreement explains — usually a stale line or a fat-fingered price — and the badge pins to the book it blames. Such prices are normally corrected within minutes, and taking them repeatedly is the fastest way to have an account restricted.
Sport, the two teams or the proposition's title, the start time, and the market with its line. For a player prop the title names the player, and the market names what is being counted.
When the engine first saw this price. Rows are removed as prices move. So an old detection time on a row still here means the book has held the number — either a slow book, or a market nobody else is pricing.
The whole bet, in the order you will place it: selection, book, price. The book's logo sits beside its name, and the Bet link beneath opens that book with the selection loaded where the book supports it.
The chance the bet actually has, with a book's margin removed, and the same figure as decimal odds. Everything on the card is measured against this. Tap or hover it for how it was built — a sharp reference or a de-vigged cross-book consensus, the de-vig method, and the confidence tier.
Pre-filled with a fractional-Kelly stake computed from the bankroll and risk level in your Settings, scaled down by the engine's confidence in this particular edge and hard-capped at your own maximum. Hover the `?` for how this row's figure was sized. It is a suggestion, and the box takes whatever you type over it.
On exchange and prediction-market rows only. Kalshi and Polymarket charge per contract, so the headline edge on the card is already net of that charge and this line shows what was taken out. The fee bites hardest on longshots, which is exactly where a gross edge looks most attractive.
Files the bet with its price, stake and the edge it carried at placement. This is what lets the tracker later compare the price you took with where the market closed, which is a readable signal in dozens of bets where profit takes hundreds.
The three controls that matter
Ignore everything else on a first read.
Min edge %, inside Filters
It defaults to 1, which is the board most people should read. The engine detects down to 0.5% net of fees. Typing 0.5 opens a thin band that is genuinely positive, badged small edge so it cannot be mistaken for something to bet full size. Raising it is the more dangerous direction — see the pitfalls.
Books, inside Filters
Select only the books you hold accounts at. Everything else on the board is a bet you cannot place, and a feed of unplaceable rows trains you to scroll past the ones you could have taken. Books that are not licensed where you are appear in a separate reference-only group, so you can see the price without mistaking it for one available to you.
Max quote age (s), inside Filters
Keeps rows where at least one book confirmed its price that recently. The edge here exists because a book is slow. The gap between a price being published and you clicking it is where the edge goes. A 60-second cap is the difference between reading live prices and reading history. The rest of the bar — presets, the event picker, segment, min and max odds, the date range — narrows the board and can wait.
A worked example
From a real row to a placed bet, including the stake.
Eighty-seven cents. That is what a good bet here is worth, and the arithmetic that follows is the whole strategy. Two hundred bets like this expect $175. One standard deviation across those two hundred is about $435. Finishing two hundred bets down $260 is an ordinary outcome for a method that is working exactly as described. If that would make you stop, size smaller or use a different screen — the suggestion is also scaled down further on low-confidence rows, and capped at whatever maximum you set.
When it goes wrong
The price moved, a leg was voided, the row vanished. What to actually do.
The +105 is gone by the time the book's slip opens, and it is now +100.
Price the new number against the same fair figure before clicking: 0.502 × $100 − 0.498 × $100 = +$0.40 per $100, which is 12 cents on a $30 stake. The break-even price is the fair decimal on the card, 1.99 — about +99 — so +100 is technically positive and, at 12 cents, not worth the click. Anything below +99 is a losing bet. Do not take a worse price because you had already decided to bet.
The row has disappeared from the board while you were on the book's site.
That is the system working: rows are removed when the price they described stops existing. Go back, refresh, and look for the same market. Do not assume the bet is still there. The book may have moved, or the fair number may have moved to meet it — the same bet becoming a bad one. Tightening Max quote age makes this happen less often by showing you fewer rows that were already dying.
The book accepts the bet at a lower price than the card showed.
Turn off accept any odds changes in the book's own settings. With it on you commit the stake and the book picks the price afterwards, which on a 2.91% edge is enough to remove the entire reason for the bet. If the slip re-prices anyway, compare it with the fair decimal and decline rather than shaving the edge.
The book refuses your $30 and offers a maximum of $8 on that selection.
Take the $8 if the edge holds, and read the refusal as information: that account is being managed, and the ceiling usually falls further rather than lifting. Move volume to your other books, keep ordinary bets and promotions on the account, and shift size to the exchanges, where no one has a position to protect.
Common mistakes
The specific ways people lose money with this tool.
Raising min edge to 5% so that only the strong bets show.
Above about 4% the board fills with stale and mistaken prices — the ones books void, and the ones that get accounts restricted. The ordinary 2-3% rows that make the money disappear from view. Thirty bets a week at $30 and 2.91% expects about $26; three bets a week expects $2.60, and the difference is not skill.
Leave the floor at 1% and take volume. If a row does show 6% or more, read the mispricing badge and decide deliberately whether the account risk is worth it, rather than filtering yourself into a board made only of those.
Raising the stake after a losing run, to get it back.
A 50.2% bet loses five in a row 3.1% of the time — roughly one five-bet stretch in thirty-three, so it will happen to you this month. Doubling to $60 does not change the edge at all; it doubles the standard deviation over the next two hundred bets from about $435 to about $870, on an expectation of $350.
Size from the bankroll, not from the last result. If the bankroll is down, half Kelly on the smaller bankroll is already a smaller bet, which is the whole point of sizing that way.
Deciding after forty bets whether the strategy works, by looking at profit.
At $30 and 2.91%, forty bets expect about $35 with a one-standard-deviation swing of about $195. The noise is roughly six times the signal, so the profit figure is not evidence of anything — and people who read it as evidence quit a working method or double a broken one.
Log every bet and read closing line value instead. If the prices you took keep beating where the market settles, the method is working whatever the balance says this month, and that signal separates from the noise in dozens of bets rather than hundreds.