OPTSELL NO EDGE

No edge at 30% / 15%: the target is not reachable with option selling in this data

Non-intraday: 17,280 variants, none reaches 30% a year (best within a 15% drawdown cap 19%); intraday: 1,296 variants, 1% profitable in development. 3 of 6 pre-registered conditions passed.

• What was tested (pre-registered, STRATEGY_OPTSELL.md): short strangles, iron condors (100 and 200-point wings) and trend credit spreads on weekly and monthly expiries, with entry timing, strike distance (0.75 to 1.5 expected moves from VIX), profit targets, stops, a close-before-expiry option and six entry filters (VIX percentile, VIX minus realised volatility, VIX level, recent move), sized by risk and margin at 3%, 5% or 8%. Real end-of-day prices for every strike from NSE (Dec 2020 to Oct 2026, 1,456 sessions), settlement at the official index value, Anand Rathi brokerage assumed ₹50 per lot per order (unconfirmed), STT (0.15% from April 2026), exchange, SEBI, GST and stamp charges, and 0.5% slippage per fill (stressed at 2x and 3x). Development, validation and a sealed final 20%.
• The best an option seller could do inside the drawdown cap was 19% a year in development and 26% in validation, by different variants; 5 variants earned 10% or more in both periods with drawdown under 15%, and none reached 15%. The 30% target needs roughly 2.5 times the leverage the best robust variants use, which pushes the calm-period drawdown to roughly 20 to 25% before any crash.
• The finalist (selected on development only: short strangle at 1.5 expected moves, entered 4 days before the weekly expiry, 2x stop, entry only when VIX is above realised volatility) earned 10.3% with a 1.9% drawdown (t 5.1, 137 trades) and then -1.2% with a 10.6% drawdown in validation. At 2x costs: 7.8% and -3.3%; at 3x: 5.1% and -5.8%.
• It is not better than chance timing: random entry days with the same structure earned a median 9.2% in development (90th percentile 12.0%) against the finalist's 10.3%. The reality check across all variants is borderline (p = 0.05): the selling premium itself is real in development, but the timing and filters add nothing, and the premium weakened in validation (a ₹76,000 loss on 7 April 2025 and a net loss through 2025).
• Crash risk is the real constraint. On the 2019-02 to 2020-11 window, which was not used for selection, plain weekly short strangles lost 11% to 24% a year with drawdowns of 30% to 45%: between 20 February and 30 April 2020 they lost ₹3.1 to 4.3 lakh on ₹10 lakh, ₹2.8 to 3.2 lakh on 12 March 2020 alone. Hedged structures lose less in a crash (the 200-point iron condor lost 15% a year with a 25% drawdown on that window) but do not pay in normal times: iron condors lost in the scored periods (100-point wings: -10.3% a year in development), because the long wings cost more than the extra edge. Keeping the drawdown under 15% through a March-2020-type move would limit position size to roughly 40% of what the calm-period results used (their crash drawdown was 30 to 39%), which scales the return to about 4 to 6% a year (our estimate by scaling, not a separate back-test).
• Intraday (1,296 variants on 1-minute data): only 1% of variants were profitable in development and 9% in validation; the median iron condor variant lost about 85% of capital a year; the best variant earned 4.1% with a 5.9% drawdown and nothing in validation. 90 to 96% of trades include a model-priced leg (the strikes sit outside the observed ATM ±2 window), which weakens but does not reverse the conclusion, consistent with earlier intraday studies (Cinder, the fused scalp's selling side, MGTS).
• Limits: income tax is not deducted; end-of-day stops cannot see intraday excursions (this flatters stop variants); margin is approximated (12% of notional for naked strangles); Anand Rathi's brokerage is an assumption; March 2020 is a stress check, not part of the scored periods; sealed period (from 2025-08-19) unopened. Nothing is paper-traded because no variant passed.