Middling — when to fish for the gap and what it costs
A middle is the most misunderstood position in betting. People describe it as covered because both sides can win, which is exactly backwards — the base case is that one side loses, and you pay the vig for the privilege of the other side winning. This lesson gives you the break-even arithmetic, a fully worked example, and an honest answer to when it is worth fishing for the gap.
What a middle actually is
You hold two positions on the same game, on opposite sides, at different numbers, where a band of outcomes wins both.
Say you take an underdog at +3.5. The line then moves and you take the favorite at -6.5. If the favorite wins by 4, 5, or 6, both positions cash. If the favorite wins by 7 or more, the -6.5 wins and the +3.5 loses. If the favorite wins by 3 or fewer, or loses outright, the +3.5 wins and the -6.5 loses.
Three outcome buckets, not two. That third bucket — the middle window — is the whole point, and everything about whether the position is good comes down to how wide the window is against what the other two buckets cost you.
The cost of the middle
Here is the piece people skip. When the middle misses, you do not break even. You win one side and lose the other, and you paid juice on both.
At -110 on each side with one unit staked per side:
- Middle hits: both win.
0.909 + 0.909= +1.818 units - Middle misses: one wins, one loses.
0.909 − 1.000= −0.091 units
That −0.091 is the cost of the middle, and it is exactly the vig on one bet. The gap does not come to you at no cost. You are paying roughly 9% of a unit, every time the window misses, to hold a ticket on the gap.
The break-even hit rate
Solve for the window probability p that makes EV zero:
p × 1.818 + (1 − p) × (−0.0909) = 0
p × 1.909 = 0.0909
p = 0.04762 — the middle needs to hit 4.76% of the time, or once in 21 attempts.
That number is the single most useful thing in this lesson, and it is entirely a function of price. Change the price and it moves hard:
| Price on both sides | Break-even hit rate | Roughly |
|---|---|---|
| -105 / -105 | 2.44% | 1 in 41 |
| -110 / -110 | 4.76% | 1 in 21 |
| -120 / -120 | 9.09% | 1 in 11 |
Look at the -120 row. Bad prices more than double what the window has to deliver. A middle you had to pay -120 on both sides to construct is a fundamentally different proposition from the same middle at -105, and members chase the second bet at a bad number constantly because the gap looks free.
A fully worked example
Take the NFL middle above: +3.5 at -110 on Monday, then -6.5 at -110 on Saturday after the line moved three points.
The window is a final margin of 4, 5, or 6 in the favorite's direction. NFL margins of 4, 5, and 6 each land in the low-single-digit percentages historically. For this example, call the combined window 11%.
- 11% of the time: +1.818 units
- 89% of the time: −0.0909 units
EV = 0.11 × 1.818 + 0.89 × (−0.0909) = 0.2000 − 0.0809 = +0.119 units
You risked two units to make +0.119, which is +5.95% ROI on the total staked. That is a genuinely strong position — better than most single bets you will find — and it looks nothing like the no-downside framing the word "middle" tends to invite.
Run the same window at a worse construction price and it degrades fast. At -120 on both sides, the break-even is 9.09% and an 11% window barely clears it.
Where middles genuinely appear
Middles are a byproduct, not a strategy. You do not go looking for them; you notice them when a position you already hold gets overtaken by the market. Three situations produce almost all of them.
A line moves through a key number after you have bet one side. This is the classic. You took an underdog at +3.5 early in the week on your own read. Injury news lands, the number runs to +6.5 or 7, and the other side is now available at a number that opens a real gap. The move was information; the middle is the residue.
Cross-book dispersion on a slow-moving market. College basketball, second-tier props, and anything on a large slate will show two-point or three-point spreads across books simultaneously. On a 100-game Tuesday, that dispersion is not an error a single book will fix quickly.
Live markets against a pregame position. You hold a pregame side, the game script moves the live number well past where you bought, and the gap is available in-game. This is the most common source in practice and the one where the hold is highest, which is why the price check matters most there.
The unifying rule: the second leg has to stand on its own. If you would not take -6.5 at that price without the +3.5 in your account, you are not middling, you are hedging badly and calling it something more flattering. Ask the question directly — is this second position, priced as it sits, one you would take cold? If no, the gap is not paying for the vig.
Practical notes
Use half-points. A middle constructed on whole numbers introduces pushes, which change the arithmetic and usually not in your favor. +3.5 and -6.5 is a clean window. +3 and -7 is not.
Size both legs deliberately. Two units of exposure is two units, and it still lives under the 5% bankroll cap. Members routinely size the second leg as if it were free.
Check the hold on the pair. If both sides came from books charging -120, you constructed a position that needs a 9%+ window. The middle calculator prices this in seconds; do not do it in your head under time pressure.
Do not manufacture middles. Deliberately taking a bad first number in order to fish for a gap later is a negative-EV habit dressed as cleverness. Every middle in this lesson starts with a position that was +EV on its own terms.
Middles are account-flagging behavior. Two-sided action on the same game at two books is a recognizable pattern. Rotate books and space the legs out.
The honest summary
A middle is a small, real edge sitting on top of two positions you should have wanted anyway. At standard pricing you need the window to hit about once in 21 tries, and a well-constructed NFL key-number middle can run two or three times that. It is not free, it is not a promise, and the moment you start taking worse numbers to create one, it stops being +EV at all.