Same-day SPX credit spreads did not survive the test

What we tested

Six hundred trading days sampled from May 2022 (when daily SPX expirations began) through October 2026, using intraday NBBO quotes. Short strike at 10 or 20 delta, a wing about 1% below, every leg quoted inside the same 20-second window, credit taken as short bid minus wing ask, held to the SPX close with no stop, fees on every leg. Entries at 9:45 and at noon Eastern, plus a 10-delta iron condor.

The strike picker hit its targets (achieved deltas 0.102 and 0.202), so the results are not an artifact of picking the wrong strikes.

What it earned

Win rates looked excellent: 93.5% for the 10-delta morning spread, 93.9% at noon. Net of fills and fees, the edge was about 0.1% to 0.4% of the collateral per trade. Maximum drawdown ran from 1.5 to 4.6 times the collateral of a single spread, and full-width losses happened on 0.2% to 2.0% of days depending on the cell.

Sized so the worst drawdown was 20% of capital, the morning cells earned 1% to 3% a year and the noon spread about 12%. Two of the five years were negative or flat in every cell, and the noon cell's total came mostly from two of those years.

Why noon did better, and why that is not an edge

The noon entry had a quarter of the morning cell's full-loss rate (0.2% against 1.0%) because three hours to the close carry less gap risk than six. The trade is shorter, not smarter. Five of the six worst days in every cell were the same handful of dates, and none of them was visible at 9:45.

The deeper point is the same as for weekly puts and for stop-losses: the market prices the variance premium fairly closely, the edge is the tail you carry, and recycling the collateral every day multiplies the number of gap days you are exposed to, not the edge. A stop does not help either, because the losses are gaps.

What it changes in practice

A 94% win rate on 0DTE spreads is not evidence of a good program. Look at the drawdown in units of collateral, the number of losing years, and how much of the total comes from a few days. On our measure the 30 to 45 day put credit spread was a better use of the same capital, with about nine entries a year instead of 252. Our lessons on stop-losses and weekly options cover the related evidence.