An honest look at six months of 0DTE strategy data.
652 trades, 8 strategy variants, 91 trading days. The whole dataset, the methodology, and what the numbers actually say — including where the apparent edges turn out to be artefacts.
Headline numbers
Paper trades only. Every trade in this dataset was simulated against real-time market data using a paper account. No real capital was deployed. Results in live trading would differ — likely worse — due to slippage, fills, commissions, and the fact that paper executions assume idealised mid-price fills.
Outlier filter applied. Five trades on 2026-03-20 logged credits of $64–69 (vs the typical $5–9 for these strategies) and corresponding P/L of $54–60. The credits are 7–8× normal and look like a logging anomaly. They have been excluded from all analysis on this site. The pre-cleanup headline was +$261.46 — almost the entire amount came from those five trades. Full discussion of the outlier removal here.
Active vs retired strategies. Of the 8 variants tested, 5 are still actively traded (Iron Fly V1–V4 and Dynamic 0DTE) and 3 were retired at the end of March 2026 after a poor Q1 (20 Delta IC, 30 Delta IC, Gap Filter 20D). Splitting the result that way:
Active strategies combined: +$133.64 across 313 trades. Retired strategies combined: −$154.40 across 339 trades. Net: −$20.76.
Cumulative P/L over time
Daily aggregated P/L across all strategies, plotted as a running total.
The curve is not smooth. Most of the gain happens in two clusters: a small build in January and a large jump in March 2026. February and parts of December dragged the book negative. This concentration is one of the most important features of the dataset and is discussed in Caveats.
Top-line findings
What worked
- Iron Fly family — the four Iron Fly variants combined produced approximately +$112.81 after outlier removal. The single best risk-adjusted strategy is Iron Fly V1 entered 30 minutes after market open.
- Dynamic 0DTE at +30m — highest win rate in the dataset (80.5%), avg +$0.49 per trade, on the smallest sample of the profitable strategies.
- 30 Delta IC at +15m — only profitable bucket in the entire 30 Delta strategy.
What didn't
- 20 Delta IC at +60m — worst single time bucket in the entire dataset (−$52.83), driven by a single −$35 loss.
- Gap Filter 20D — supposed to filter tail risk and instead carries the worst per-trade losses in the book.
- 30 Delta & 20 Delta overall — both negative despite high win rates. Classic high-win-rate / fat-tail asymmetry.
The DST artefact — why naive analysis is misleading
One finding deserves special attention. From 2026-03-08 to 2026-03-28, US clocks moved forward but UK clocks didn't, leaving a 3-week window where the bot's UK clock times were one hour earlier than usual relative to the market open. 130 of the 652 trades fall in this window.
Without correcting for this, several strategies appear to have two distinct entry slots when in fact they only have one — and the apparent "early-slot edge" turns out to be just the strong March 2026 performance of the misaligned trades. Once normalised to "minutes after market open", the picture changes substantially.
What you'll find on this site
- Methodology — how the data was collected, the DST correction, definitions of every term.
- Strategies — full per-strategy breakdown of all 8 variants traded.
- Entry Timing — strategy × entry-time bucket analysis, normalised to minutes-after-open.
- Day of Week — performance by weekday for the top two strategies, with and without outliers stripped.
- Performance Deep Dive — streaks, drawdowns, profit factor, IV regime, day-move sensitivity, exit clustering, and autocorrelation across the top five strategies.
- Best Pick — if forced to pick one, which strategy/slot wins, and why.
- Caveats — sample size limits, paper-trading bias, March-effect, and what could go wrong.