A positive-expected-value screen is only as good as the fair value behind it and the freshness of the price in front of it. Four things break it: a ranking is a maximum, so it finds outlier books; a consensus over two books is noise; a stale quote ranks first forever; and one-sided markets have no fair value at all. Our own board has the same limits, stated below.
Everything below was measured on our own board before we published it. The MLB price vs fair board carries the per-book table this page argues for.
1. A ranking is a maximum, so it finds the strangest book
Sort a few hundred thousand priced combinations by how far above fair value they sit and look at the top. You are not looking at the book that prices well. You are looking at the book with the widest scatter, because a maximum is won by whoever has the fattest tail.
We measured this on our own surface before publishing it. Of the top 24 tickets on the 2026-08-14 evening board, 23 sat at one sportsbook. That book had the third-worst median price of the eleven we compare, at 5.31% below fair, and the best single price on the board, at 2.90% above it. A reader seeing only the ranking would have concluded that book was generous. Its own slate says the opposite.
This is why every ranked price list we publish carries a per-book table beside it, showing how each sportsbook's whole slate prices, not just its best quote. A tool that shows you the ranking without that table is showing you a tail and calling it a house.
2. A consensus over two books is not a consensus
Fair value is a median across the books quoting the same bet at the same number. With two books, the median is the average of two opinions and it moves the moment either one does. The cruel part is that the thinnest markets produce the noisiest fair values, and a screen sorted by the gap between price and fair value selects for exactly those markets, because that is where the estimate is most wrong.
Our board refuses to print a fair value on fewer than four books posting both sides of the same number. That floor costs the game markets almost nothing, and it removes most of the player-prop board. On the 2026-08-15 evening slate it dropped 128 player-prop bets and 11 game-market bets, each of which is two sides, and those counts are published in the diagnostics on the page rather than hidden.
3. A stale quote ranks first, forever
A price grid can honestly show a book's last quote for hours, because it prints the number and lets you read it. A screen cannot. It subtracts one book's price from a consensus of the others and calls the difference a premium, so a book that simply stopped updating will look generous every single day, and a ranking is exactly the shape that finds it.
Our board drops a game-market quote that is more than three hours behind the newest capture on the same board, before anything is ranked, and it prints the count of what was dropped and which book it came from. A screen with no such rule is quietly ranking the slowest book first.
The general form of this problem is worse than the specific one. Prices move all day. Any figure of this kind is a claim about one instant, and the instant is usually gone.
4. One-sided markets have no fair value at all
Removing a margin requires two prices. Some markets do not have two. MLB home run props are the standing example: across our own archive on 14 August 2026 the market carried 1,528 Over rows and zero Under rows, on every book and every rung. There is nothing to devig.
A tool that shows a number for such a market has estimated the missing side from somewhere, usually a model, and the resulting figure is a statement about that model rather than about the market. We exclude those markets by name and say which ones and why.
The same applies across different numbers. Over 8 and Over 8.5 are two different bets, and a total priced at 8.5 must never be compared against a consensus built at 8. Player prop ladders are worse: one book alone can post three rungs of the same player's total bases. Anything that collapses those into a single "main line" is comparing prices for bets that do not match.
What our own board cannot do
Every criticism above applies to us too, and the honest version of this page is the one that says where our own limits are.
- We are not fast. Game prices are captured ten times a day, between 9am and 7pm ET. Player prop prices are captured once, on the morning of the game, and are not refreshed afterwards, so every prop figure we publish is an opening price and is labelled as one. If you need a number that is current to the minute, our board is not that and cannot become it without a different data purchase.
- Our fair value includes the book being measured. A book's own quote is one of the quotes in the consensus it is compared against, which makes every premium we print slightly conservative. We prefer that direction of error.
- We cover the books we buy. Eleven sportsbooks quote our MLB board. A book we do not carry may be posting a better price, and we would not know.
- A snapshot is not a strategy. We publish no picks and make no claim that any figure on our board can be turned into a profit. Our own graded models trail or tie the closing market, which is precisely why no model number is allowed anywhere near these fair values.
What the board is good for is the thing it actually measures: how far a posted price sits from what the rest of the market says the bet is worth, on every side at once, with the per-book table beside it. On the 2026-08-15, 7:26pm ET board that was 6 of 148 sides at or above fair, with the median side 2.91% below. That is a description of a market, not an opportunity.
How this was measured
- The ranking finding. Enumeration of every cross-game two and three leg combination on the 2026-08-14 evening MLB board (188,752 combinations, 11 sportsbooks), each priced at each book that would take the whole ticket and compared with the devigged multi-book consensus. Produced by
mlb/mlb_parlay_board.py. - The board figures. The 2026-08-15, 7:26pm ET build, produced by
mlb/mlb_ev_board.py: 148 lined sides across 17 games and 11 sportsbooks. - The one-sided market count. Measured on our own
odds.linesarchive on 14 August 2026 across every book and every rung. - All figures come from a committed script run over a read-only database session, and are captured pre-game listed prices rather than closing lines.
Next steps
Frequently asked questions
What does +EV mean in betting?
It is shorthand for a price that pays more than the probability of the outcome justifies. The difficulty is the second half: the probability has to come from somewhere. Most screens use the devigged consensus of other sportsbooks, which means the figure is a comparison between books rather than a statement about the event.
Are positive expected value tools a scam?
Not inherently. The arithmetic is real and line shopping is real. What is frequently misleading is the presentation: a ranking without the per-book context, a fair value built on too few books, or a stale quote that ranks first because nobody refreshed it.
How many books should a fair value use?
More than two. Our board requires at least four sportsbooks posting both sides of the same number before it prints a fair value, and it publishes the count of what that floor excluded. Any tool that will not tell you how many books are behind its number is asking you to trust an estimate you cannot inspect.
Why do prop markets show bigger gaps than game markets?
Because they carry more margin. On the 2026-08-15 board the median game-market side sat 2.46% below fair while the median prop side sat 4.59% below. A larger gap on a prop is usually the market being more expensive, not a better price.
Does Nebula Insights sell a betting tool?
No. Our boards are free, we take no affiliate money, we link to no sportsbook and we publish no picks or selections. Nothing on this site is betting advice.
Related: Closing line value explained · Arbitrage guide · How often betting lines move · How to read player prop trends
Every figure on this page is a historical measurement of what already happened, produced by a committed, re-runnable script over our own odds archive. Past frequencies are not forecasts and are not betting advice: Nebula Insights publishes no picks, no selections and no recommended wagers. 18+ only.