Expiry-Day Theta Calendar

The trade. On a day when an expiry is expiring, sell that 0DTE option and buy the next expiry at the SAME strike, a couple of strikes out of the money. The 0DTE leg's remaining premium is destroyed by the clock -- that part is a certainty, not a forecast -- while the later expiry keeps most of its value. The spread you own therefore widens into the close.

What it costs. Gamma. If the underlying travels well past the short strike, the 0DTE leg gains intrinsic faster than the back leg does. Strike distance is the risk dial, and the table below shows the tail is small but real.

The surprise. Entering EARLY beats waiting. Entry at 09:40 earned roughly 14x per trade what a 15:00 entry did, and a 15:30 entry actually loses money, because by then the premium you were hoping to collect has already decayed — somebody else collected it. "It goes to zero after 2pm" is true, but you have to be holding it while that happens, not arriving at the end.

The trap this backtest caught. Sweeping the strike further out of the money makes the numbers look progressively amazing — 76% wins and +$601 per trade at 6 strikes OTM. It is not real. Splitting the P/L by leg shows that at 3+ strikes OTM the short leg gains value instead of decaying, and the long leg supplies all of the profit: that is a directional bet on a 1-day option, not decay harvesting. The apparent improvement is also selection bias — a far-OTM 0DTE strike only passes the liquidity gate on days the underlying travelled toward it, so the deep offsets quietly keep only the sessions that were going to pay (124 of ~800 at 6 strikes out). Every table below therefore carries Front Gave and Back Gave: if Back Gave dominates, it is not this strategy.

Entry 09:40 → exit 15:55, 1 strike(s) out of the money, back leg 1–1 DTE, fill assumption 0.35. Exit is deliberately before the bell: an unclosed short 0DTE leg is assignment risk and the last prints are the least reliable quotes of the day.

Three actual positions from the backtest, reconstructed from the 5-minute chains. The amber line is the 0DTE option that was SOLD; the green line is the next-day option that was BOUGHT; the shaded gap between them is what the spread is worth. Watch the amber line fall to the floor while the green one holds — that gap widening is the entire trade. The third example is a loser, included because a page that only shows winners is advertising.

The entry time, strike offset and back expiry were chosen by sweeping the whole history, which is fitting. This runs the same fixed settings on the later period separately. It holds up — but notice the mechanism drifts: the in-sample half is nearly pure decay, while the later half takes more of its profit from the long leg moving. The decay component keeps the same sign in both halves; its share of the profit does not.

Each row is the same trade opened at a different time of day and always closed at 15:55. The edge decays with the clock you are trying to sell, and a 15:30 entry is a loser.

Read Front Gave down this table. It collapses to zero and then goes negative, while Back Gave takes over: past 2 strikes OTM this is no longer a decay trade, however good the P/L column looks.

The nearest expiry is the best long leg: the cheapest debit for the same decay collected, so the best return on capital.

This is the table that separates this strategy from the other two: it never goes negative. Even paying the entire quoted spread on every leg both ways it still makes money, because the short 0DTE leg usually expires worthless and costs nothing to close.

Click any row to expand the same picture for that trade — the two legs and the stock, minute by minute. The paths load on demand from a sidecar file so the log itself stays small; if that file cannot be reached the row simply says so and the table keeps working.