Caveats & limitations

1. These are paper trades

Every trade in this dataset was executed against live market data using a paper account. No real capital was deployed. Paper trading systems generally:

• Assume idealised mid-price fills, where in reality bid/ask spreads on multi-leg orders can be wide.
• Do not model partial fills or slippage on rapidly moving markets.
• Do not deduct commissions or exchange fees.
• Do not model the psychological friction of real losses.

For Iron Flies in particular — where both shorts are at-the-money — the realistic per-trade slippage is likely $0.20–$0.50. On the cleaned best strategies in this dataset (avg P/L $0.31–$0.62) that is a substantial haircut — possibly compressing the edge by 30–80% or wiping it out entirely depending on broker fills.

2. Sample size is small

652 trades sounds like a lot, but they are concentrated in 91 trading days across 8 strategy variants. The best individual strategy/bucket combination has 72 trades — barely sufficient for assessing per-trade economics, and entirely insufficient for characterising tail risk.

A useful rule of thumb: to detect a 5%-difference in win rate with 95% confidence requires hundreds to low thousands of trades, not dozens. Anything you read on this site about "best strategy" should be understood as "best within the limited evidence available."

3. The 2026-03-20 outlier removal

The five excluded trades, all settled on 2026-03-20:

• Iron Fly V1 — credit $64.95, P/L +$54.95
• Iron Fly V2 — credit $65.57, P/L +$55.57
• Iron Fly V3 — credit $69.85, P/L +$59.85
• Iron Fly V4 — credit $68.53, P/L +$58.53
• Dynamic 0DTE — credit $65.75, P/L +$55.75

Combined, these five trades contributed approximately +$284 of P/L. The pre-cleanup headline was +$261.46 across all 8 strategies. After removing them, the headline is −$20.76 — meaning the entire apparent profit of the book came from these five anomalous trades. That is the most important number on this site.

The remaining 5 active strategies (Iron Fly V1–V4 and Dynamic 0DTE) are still net profitable as a group at +$133.64, with V1 individually at +$44.48. The 3 retired strategies (20 Delta IC, 30 Delta IC, Gap Filter 20D) are net −$154.40.

4. The March 2026 effect

Even with the outlier trades removed, March 2026 was a strong month for the Iron Fly family and Dynamic 0DTE. The strategies tend to thrive in the kind of volatile, mean-reverting market that March produced, so the implicit bet behind running them is that conditions like March will recur often enough to matter. That is plausible but not guaranteed.

The pre-cleanup version of this site reported March alone as carrying 77% of V1's total profit. After outlier removal, March's contribution drops sharply and the monthly profile for V1 becomes much more even (Jan +$8.87, Feb −$0.47, Mar +$20.49, Apr +$14.63, May ~$0.49). This is a healthier-looking distribution but the absolute numbers are smaller.

5. The buying power data may be incomplete

The reported average buying power for the Iron Fly variants ranges from $23 to $36 per spread. If these figures are accurate, the strategies are producing 3–5% returns on capital per trade — which would be extraordinary.

More likely, the buying power field in the data does not capture the full margin requirement — perhaps it nets the short call and short put because they are at the same strike, ignoring the wing margin. Real broker margin requirements for an Iron Fly are typically the wing width minus the credit collected, which would be considerably higher than what is logged.

This means any return-on-capital figures on this site (the ROI% column in the underlying CSVs) should be treated with skepticism. The absolute P/L per trade figures are the more reliable basis for comparison.

6. Strategy retirement bias

The 20 Delta and 30 Delta IC strategies were retired at the end of March 2026 after a poor Q1. The Iron Fly variants and Dynamic 0DTE were given more capital and continued running.

This is sensible portfolio management, but it introduces a survivorship bias into the dataset. The strategies that "survived" to be analysed in April and May are the ones that were already winning. If we had continued running the Delta ICs through April and May, their results would form a different overall picture.

7. The DST artefact (recap)

Discussed in detail on the Methodology page and the Entry Timing page. In short: 130 of the 652 trades fall in a 3-week window where US clocks had moved forward but UK clocks had not, making naïve UK-clock-time bucketing misleading.

The pre-DST-correction analysis suggested certain strategies had two distinct entry slots. The post-correction analysis shows they have one. Without the correction, it is easy to read a "morning vs afternoon edge" into the data that does not exist — what looked like a morning slot is actually the strong March 2026 subset of a single all-day slot.

8. The IV regime blind spot

This means every conclusion on this site is conditional on "high implied volatility regime." The dataset cannot tell you anything reliable about how these strategies behave when IV Rank drops below 0.35 — there are at most a handful of trades per strategy in those bins, and zero trades in the mid (0.35–0.50) bucket.

Why does this matter? Iron Flies and Iron Condors collect premium in proportion to implied volatility. In a high-IV regime, the credits are large and the structures have wider profit zones. In a low-IV regime, the credits compress dramatically and the same strikes provide much less of a buffer. A strategy that prints money at IV Rank 0.70 may struggle or lose money at IV Rank 0.20 — and the data here cannot tell you which.

If your trading environment shifts to a sustained low-IV regime, treat the historical performance shown here with caution rather than as predictive. See the Performance page for the full IV regime breakdown per strategy.

9. Market regime dependency

The research period (December 2025 to May 2026) covered a specific and fairly narrow set of market conditions. SPX had a meaningful drawdown in February and a sharp recovery in March. The strategies were not tested through:

• A sustained low-volatility environment (e.g. summer 2017 conditions).
• A multi-day VIX spike (e.g. February 2018 or March 2020).
• Pre-FOMC and post-FOMC days specifically.
• Major earnings days for index-heavy components.

0DTE strategies are particularly sensitive to single-day volatility. A strategy that looks great in calm markets can blow up rapidly in a panic open. The dataset gives us limited information about how these strategies would behave outside the specific regime they were tested in.

10. No transaction cost modelling

Commissions and exchange fees on multi-leg options trades are typically $0.50–$2.00 per trade depending on the broker and contract count. With 72 trades over 4 months for V1, this adds up to $36–$144 in fees — nontrivial against the +$44 paper profit. Returns net of realistic fees would be smaller and possibly negative, depending on broker and contract count.

11. This is not financial advice

What this analysis is good for

Despite all of the above, the research has value:

• It illustrates the importance of correcting for time-zone artefacts in any timed strategy analysis.
• It shows the gap between win rate and profitability — high win rate strategies can lose money if the loser-to-winner ratio is bad.
• It surfaces the concentration risk in any short-history dataset (the March 2026 effect).
• It provides a transparent baseline for comparison if any of these strategies is later traded in a real account.

The conclusions are honest, the data is laid bare, and the limitations are documented. That is the standard this site is held to.