Overnight Nifty strangles: how costs and slippage changed a 46% back-test

By TradeTech Analytics Research Desk · Research · Published 2026-10-27 · 4 min read · How we write and check

A back-test report for an overnight short strangle on the Sensex showed a gross compound return of 46% a year. Gross numbers are the beginning of the story, not the end. We replicated the idea on Nifty (and Sensex), applied standard charges and slippage, and then started a forward paper test. This is where it stands.

The report that started it

The report (generated in October 2026) covered an unhedged positional strategy on weekly Sensex options: sell a strangle (an out-of-the-money call and put) one day and buy it back the next, over two years from September 2024, on ₹7.5 lakh of capital. Headline: gross profit ₹8.53 lakh, a 65% win rate on days, gross Sharpe 3.36. Reading further on the report's own pages:

• It was gross. The report's cost section put costs at ₹4.96 lakh, 58% of gross profit, over 5,676 order fills; net profit ₹3.56 lakh and a net Sharpe of 1.38, at an assumed 0.5% slippage.
• "Repairs" lost money. Planned exits earned +₹44.86 lakh gross over 1,535 round trips; the re-strike repair exits lost ₹36.33 lakh over 898.
• The call leg did everything: ₹7.61 lakh from calls, ₹0.23 lakh from puts, in a rising market.
• Tail risk was visible: excess kurtosis 21, worst day −10.1%, nine losing trades in a row, and real fills about 1.5% worse than the minute open (3.5% in the first 15 minutes), measured over 21,011 live fills.

Our replication

We tested the same idea on Nifty 50 (464 sessions from December 2024 to October 2026) and replicated it on Sensex. One strangle per night: sell at an entry minute, buy back at a chosen minute the next session. Because the data series is relative to the at-the-money strike, the strike sold at night is not always in the next morning's series; those prices come from a standard option model fitted to the observed strikes, and we flag how much of the result depends on them (about 60% of modelled exits).

• Costs: the standard broker schedule (₹20 an order, exchange fee 0.03503%, SEBI fee, GST, stamp duty, and STT of 0.10% until 31 March 2026 and 0.15% after), and slippage of 0.5% per fill, which is 1% over the full cycle. Version 1.0 used 1.5% per fill from the report's own measurement.
• Periods: first half for development, next quarter for validation, final quarter sealed.

Results after re-costing (version 1.1)

The re-costed Nifty result stays positive under the standard costs and the lower slippage assumption, and it reconciles with the earlier version of the study. The Sensex edge is about 60% smaller and statistically weak.

Why we do not trust it yet

• It is not independent evidence. We looked at every period of this data while developing the idea, so the "sealed" period is no longer a blind test of this final version. Calling it a candidate is the honest label.
• Modelled prices. About 60% of exit prices are model-based. If the model is 10% off against us, the Sensex edge is roughly zero.
• Slippage is the swing factor. Live fills in the report averaged 1.5% worse than the minute open, three times our 0.5% base assumption. Our version 1.0 test at 1.5% per fill earned about ₹871 a night on Nifty, roughly half of the re-costed figure. Which assumption is right depends on the broker, the size and the minute; only live fills answer it.
• Tail risk. Unhedged short strangles earn small gains most nights and can lose several multiples of those gains on a gap. Whether the sample has seen enough bad nights is the key unknown.
• Regime. The sample is a rising market, and calls produced almost all the profit in the original report.

What we did next: a frozen forward paper test

We froze the specification (strikes, entry and exit times, slippage and costs) on 24 October 2026, and began a nightly paper run on 26 October 2026 (the first strangle was recorded that evening at the 15:15 open, with the buy-back due at 12:00 the next session), using only exchange price data: no orders of any kind are placed. Each evening the run records the modelled fills at the frozen assumptions and the results at 1x, 2x and 3x slippage. Rules for the test: no change to the specification during the run; results published on the strategy's own page whether good or bad; a decision on whether to keep going only after enough nights (we use a minimum of several months) to see a tail event or two.

Lessons for reading any option-selling back-test

1. Ask for the gross-to-net bridge: costs of 58% of gross are not unusual for high-frequency option strategies.
2. Ask how fills were measured. A back-test that uses the best price of each minute flatters the result; live fills were 1.5% worse in the report we examined.
3. Ask for the exit attribution (which exit type earned or lost the money) and the leg attribution.
4. Look at the worst day, not the average. Positive skew is rare in short-volatility strategies; plan for the bad night.

Limits

Our results are about what a rule did on past data with stated assumptions. The paper test uses no real orders, so it cannot capture queue position, partial fills or margin calls. The page for this study, with its trade list and daily paper results as they accumulate, is on the Strategy Review. See also how the April 2026 STT rise changed costs and how we test ideas. This is not a recommendation to sell options; short option positions can lose more than the premium collected.

Read next