Middles & Low Hold
Where two books are far enough apart that some results win BOTH bets (Middles), and where covering both sides costs least (Low Hold). Two boards, one screen. Neither is free money.
The screen, annotated
Every column, and what its number actually means.
Two sidebar entries, one screen. Both render the same card, the same filter bar and the same stake boxes; only the kind of row differs. Everything below applies to both unless it says otherwise. The differences are small: the Low Hold board has no "Max hedge bet ($)" field, and its green badge and footer carry different numbers.
The green badge at the top-left, on a MIDDLES row: expected value as a percentage of a $100 position. Not a locked return, not the chance of winning, and not the width of the gap. The board is ranked on it rather than on window width, because a ten-point window priced badly is worth less than a two-point window priced well. Rows with negative expected value are not served at all, so every middle you can see is at worst break-even on the engine's own numbers.
The market and the pair of lines, on a MIDDLES row. Two numbers with a slash, because a middle lives BETWEEN two lines. You back Over 44.5 at one book and Under 47.5 at another. Every result inside that gap wins both bets. Middles exist only on totals, team totals and spreads; there is no gap to sit inside on a moneyline.
Each leg keeps its OWN line inside its panel, under its book's name. Read these two, not the composite above: the composite says what the position is, the leg lines say what you actually type into each book's slip. A leg showing a line you did not expect is the row telling you the book has moved.
Identical to the arbitrage board's — see that walkthrough for how the linked boxes size the position. The consequence that is specific to a middle: equal payout means a MISS costs the same amount whichever side lands, which is why the card shows one "misses" figure instead of two. On the Low Hold board the boxes start at a flat $100 rather than from a server-supplied plan, because low-hold rows ship no stake split.
The live outcome line under the two legs on a MIDDLES row, recomputed as you type. Read it right to left. "3.0 pt window" is the gap. "misses" is what you lose when the result lands outside it, which is the usual outcome. "hits" is what you collect on the rare occasions it lands inside. Before you have typed a stake the same slot shows the server's $100 plan instead — "wins both: $88.91 · misses: -$5.55".
The second footer line on a MIDDLES row, and the only place the card tells you whether the position is worth taking. "needs" is the break-even hit rate — how often the result must land inside the gap for the wins to pay for the losses. "market implies" is how often the rest of the market says it will, taken from the price ladder at both lines with the book's margin removed. Green when the second number beats the first. The EV figure is the gap between them expressed in dollars.
Replaces the "market implies" half when the ladder is too thin to price the window. The break-even is still shown; what is missing is the number you would compare it against. These rows sit at the bottom of the board and their badge reads 0.00%, which is an admission and not a measurement.
The footer on a LOW HOLD row: how far above 100% the two best prices add up. At -104 on both sides each price implies 50.98%, so the market totals 101.96% and the book's combined margin is 1.96% — against a typical 4.5%. Small is not zero. The board stops at 2.5%, and a NEGATIVE hold means the two prices have crossed, which is an arbitrage rather than a low hold.
The green badge on a LOW HOLD row, which is the hold with a minus sign in front of it. It looks like a bug and is not: lower hold is better, so the board negates it in order to sort best-first like every other board. A row whose badge is POSITIVE has a negative hold — the two sides have crossed and it is a locked position, not a cheap one.
The live outcome line on a LOW HOLD row, and the honest version of the hold: cover both sides of a -104/-104 market with $50 each and you get $98.08 back whichever side wins. Red, because it is a cost. It is a shade under the 1.96% hold because the hold is measured against the implied total and this is measured against your stake.
Opens that book at this selection and copies the leg's exact stake. Worth knowing what the two clicks look like from the book's side: two bets, two books, two DIFFERENT lines on the same game, minutes apart, on a line that has just moved. Arbitrage at least looks like two people disagreeing; a middle looks like one person with a screen, and it is the single fastest way to have an account's maximum stake cut to $20.
Every book's price on this line, in place. On a middle, open it BEFORE you place anything. It shows which other books offer the far side of the gap. That is the list you will need once the first leg is on and the second price has moved.
When this pair was first seen at these prices. A middle opens because one book has moved its line and another has not, so the clock is the lagging book's — the row dies the moment it catches up. Several minutes old means re-check both lines at both books first.
The three controls that matter
Ignore everything else on a first read.
Min edge % (in Filters)
The same box means two different things on these two tabs, and this is the one thing to learn before touching the board. On MIDDLES it is expected value per $100. Leave it blank to see everything (already floored at break-even), or set 1-2 to demand rows that pay for their own execution risk. On LOW HOLD the number being filtered is the NEGATED hold, so \"hold of 1.5% or better\" is typed as -1.5. Typing 1.5 there asks for a hold of minus 1.5% — an outright arbitrage — and returns almost nothing.
No pushes (the toggle row at the bottom of Filters)
Drops any row where a result can hand a stake back — a whole-number line on the row or on either leg, including a middle's composite \"44/47\". Leave it OFF on the Middles board. A result landing exactly on a whole-number leg pushes that leg and cashes the other, which pays better than a miss. The card's EV figure gives the row no credit for it. So the toggle removes the rows whose bad outcomes are cheapest. On Low Hold it is more defensible: a push there simply returns both stakes.
Sportsbooks (the chip list at the bottom of Filters)
Tick only the accounts you hold and can fund today. A middle needs two books to disagree about the line itself, not just the price. So the board is thin to begin with, and a row you cannot place is pure noise on it. Leave the rest of the bar alone on a first read: Segment, the dates, the event picker, and the whole hedge group.
A worked example
From a real row to a placed bet, including the stake.
Read the last three lines again, because they are the walkthrough. This is a GOOD middle — the market says 10.4% and you only need 5.9%, and $42.75 of expected value on $1,000 is a better number than any arbitrage on the board will show you. And it loses $55.48 nearly nine times out of ten. Expected value and the outcome you should expect are different things. On a middle they point opposite ways: the average is positive only because of a payout you will not see for a while. A reader who takes six of these has a better-than-even chance of being down $332.88 with nothing to show for it, and that reader has done nothing wrong. Size accordingly, count in dozens rather than in single bets, and do not start unless you can sit through the losing run — because the losing run is the product. Low Hold is the opposite shape and equally not free: covering both sides of that -104/-104 market returns $98.08 on $100 every single time. No variance, no windfall, and a guaranteed 1.92% cost.
When it goes wrong
The price moved, a leg was voided, the row vanished. What to actually do.
The first leg is on and the gap has closed. You hold $494.80 on Over 44.5 at -110, and betmgm's Under is now 45.5 instead of 47.5.
The window shrank from three points to one: only a total of exactly 45 now wins both. Do the arithmetic again before you do anything else, because a one-point window is a different bet. Open All books and look for any book still offering an Under at 46.5 or higher. A middle exists because one book lagged a move — so a second lagging book is the likeliest thing on the screen. If nothing above 45.5 is left, you have three choices. Take the one-point middle for whatever the card now says it is worth. Cover on the SAME line to go flat, paying the hold. Or keep the Over as the straight $494.80 bet it already is. What you must not do is leave it open by not choosing, which is how a position you sized as a $55 risk becomes a $494.80 one. Next time, place at the book whose line has ALREADY moved first: it is the one that will move again.
Six middles taken, six misses, roughly $330 down. Something must be wrong with the board.
Nothing is wrong with the board. At the example's 10.4% hit rate, losing six straight has probability 0.896^6 = 52% — it is the more likely of the two outcomes, and eight straight is still 42%. What to check instead is whether the rows were actually good ones: for each, did "market implies" beat "needs", and by how much? If you were taking rows where the two numbers were close, or rows showing "cannot be priced", you were taking coin flips with a fee. If the rows were genuinely +EV, the only correct response is to keep the stake exactly where it is. Raising it to recover the $330 is the move that turns a positive-expectation strategy into a bankroll failure, because the size that survives twenty misses is the whole strategy. If sitting through it is not tolerable — a completely reasonable conclusion — the Low Hold and Arbitrage boards pay less and pay steadily, and that is a legitimate trade rather than a lesser one.
The row says "no two-sided quote at both lines, so the chance of landing inside cannot be priced — ranked last", and the window is wide.
Do not read the width as the edge. The engine ranks these last on purpose and badges them 0.00%, because the only number it can offer is what you NEED and not what you get. Open All books and see whether one book prices both sides at either line. If one does, you can do what the card could not: use that book's own two prices to estimate how often the result lands inside. If nothing on the board prices both sides, skip the row. A ten-point window on an alternate line nobody else quotes is usually wide because the line is stale or wrong. That is a reason the bet may be voided, not a reason it is good.
A Low Hold row's badge is POSITIVE — it reads +0.42% and the footer says "hold -0.42%".
That is not a low hold, it is an arbitrage: the two best prices have crossed, so covering both sides returns more than it costs. The board's own subtitle says so. Treat it like an arbitrage, not a low hold. Exact stakes from the boxes, both legs inside a minute, softer book first. Expect it to vanish in seconds: it is on the Arbitrage board too, and more people watch that one. If only one leg gets on, you are back in the first situation on this list.
Common mistakes
The specific ways people lose money with this tool.
Sizing a middle the way you would size an arbitrage, because the card looks the same and the stake boxes behave the same.
An arbitrage's worst case is the profit; a middle's worst case is the whole miss, and the miss is the usual result. At the example's $1,000 you lose $55.48 about nine times in ten, and a run of eight misses — 42% likely — costs $443.84 before the first win arrives. A bankroll sized for locked returns does not survive that, and the loss that ends it is indistinguishable from all the ones before it.
Size a middle at a fraction of your arbitrage stake — a quarter is a reasonable starting point — and judge it over dozens of positions rather than one. Before placing, multiply the "misses" figure by ten and ask whether you would put that number at risk today, because ten in a row without a hit is an ordinary month. If the answer is no, lower the stake until it is yes.
Treating Low Hold as free money, or as a strategy in its own right — covering both sides of cheap markets repeatedly because the cost looks like a rounding error.
It is not a rounding error, it is a fee you pay every time. Covering both sides of a 1.96% market costs $1.92 per $100, so $10,000 of two-sided volume burns $192 with no edge anywhere in the position to pay for it. Done at volume it also produces the exact paired betting pattern a trading desk searches for, so you buy the account risk of arbitrage while collecting none of the profit.
Use Low Hold for what it is. The cheapest place to put a bet you already wanted. The cheapest place to hedge a position you already hold. And the cheapest way to turn over a bonus you have to turn over anyway. In all three the hold is a cost you were going to pay anyway and the board's job is to make it smaller. Standing alone, a two-sided cover at any positive hold is a guaranteed loss.
Sorting mentally by the "pt window" number and taking the widest gaps, on the reasoning that a bigger window is more likely to hit.
You systematically select the rows the engine already ranked last. Width is not edge: a ten-point window priced badly is worth less than a two-point window priced well, which is precisely why the board sorts on expected value instead. The widest windows are also disproportionately the unpriceable ones on thin alternate lines, where the width comes from a stale or erroneous number and the leg is a candidate to be voided rather than graded.
Read two numbers and ignore the rest: does "market implies" beat "needs", and by how much? That comparison already contains the window — a wider gap raises the implied percentage by itself — and it also contains the prices, which the width does not. The board is already in this order, so working top-down and stopping when the margin between the two numbers gets thin is the whole selection method.
Running middles at one or two soft books, several a week, and nothing else at those books.
The account is limited to roughly a $20 maximum and does not come back. Middles get there faster than arbitrage, because the pattern is obvious: two bets, two books, two different lines on one game, placed minutes apart, right after one book moved. The limit takes the strategy, every unused promotion on that account, and every +EV price you would have taken there. A middle strategy needs many accounts to run at all, so each one lost shrinks the board you can play.
Spread the volume across every book you hold rather than the two with the best lines, and make middles a minority of what each account sees. Take the promotions — a promotion-heavy profile is one of the healthiest a book looks at. Keep ordinary mainline bets and +EV plays in the mix at the same books. Prefer rows where your leg sits at the book that has ALREADY moved. Taking the new line is not the side that gets flagged.