A rare positive in our backtests: deep gap-down buying cleared the cost wall, at retail scale only
Can an intraday strategy on Indian stocks survive its own trading costs? Most of what Strota has tested says no: the signal looks fine on paper until spreads and the price you actually get filled at swallow it. One construction did not fail. Buying Nifty 500 stocks that opened sharply below the previous close, restricted to a narrow slice of the market by daily turnover and held only to the closing bell, cleared realistic costs and stayed positive in every sub-period the test carved out. Over the tested history it produced roughly 10-13bp per calendar day at a Sharpe ratio of 1.1-1.5 — Sharpe being return measured against volatility, where above 1 is respectable — on an account of Rs 2.5 lakh to Rs 10 lakh.
That last clause is as much the finding as the return is. The result was measured on a small account and it decays as capital grows: by Rs 25 lakh the daily figure had dropped sharply, and at Rs 2 crore it was close to nothing. It describes a fixed slice of the past, reached after four rounds of trying to break it — two of which produced headline numbers that were simply wrong.
Here is what was measured. The panel held 481 Nifty 500 names, priced daily. The trigger was an opening price at or below -5% against the previous close, read out of the pre-open call auction that runs 09:00 to 09:08. Entry was that auction; the exit was the market close the same day. A stricter -7% version ran alongside. Positions were equal-weighted, capped at ten at once, each order limited to 1% of that stock's own recent median traded value.
Costs were not treated as a flat number, and that decision mattered more than anything else here. An early pass charged a flat 10bp per side and reported 48.9bp a day, a Sharpe of 2.70 and a CAGR of 207%. Those figures were fiction. The edge lives in stocks turning over between Rs 5 crore and Rs 25 crore a day — exactly where a 10bp fill does not exist. Swapping in a schedule that scales with liquidity, from 40bp a side in the thinnest names down to 6bp in the biggest and multiplied by 1.5x because the stock is gapping, turned the broad version negative.
A second trap was subtler. The long-only variant first printed 102.97bp a day at a Sharpe of 3.50, but both came from active days only, and this book sits in cash most of the time. Annualising on the days it trades inflated the Sharpe by 3.09x. Across the full calendar, the identical configuration comes out at 10.81bp a day, Sharpe 1.11. A strategy idle most sessions has to be judged against every session.
Liquidity band is the whole result rather than a refinement of it. At a 3% trigger the long side averaged +2.149% per trade in the Rs 5-25 crore band, against +0.222% among names above Rs 500 crore. Open the universe to everything liquid and the book collapses to 4.17bp a day, Sharpe 0.25, with a worst peak-to-trough fall of 75.5% — a drawdown, meaning the deepest slide from a previous high. That wide version's confidence interval includes zero and goes negative once 2020 is removed. Every narrow configuration's interval excludes zero.
Two checks separated the gap from the mere fact of being long. Resampling the returns in 20-day blocks, 2,000 times, placed the strict thin-band version between +6.33 and +19.50bp a day with 95% confidence. The other swapped the gap-downs for random eligible stocks on the same days, same trade count: the real book finished at the 100th percentile every time, z-scores of 14.6 to 17.9. The random version loses money, -3 to -25bp a day, because arbitrary intraday small-cap longs do not cover these costs.
One lucky year does not explain it. The 5% thin-band rules were positive in 11 of 11 years measured. Remove 2020 and the figure reads 9.95bp a day; from 2022 onward, 15.40; on data after August 2024, unseen when the rules were fixed, 9.05.
Buried inside the positive is a clean negative. Shorting the mirror image, stocks that gapped up, lost money with statistical significance under the same cost schedule: -0.179% per trade in the thin band and -0.340% one band above, t-statistics of -2.56 and -2.85. Whatever happens on the gap-down side does not run in reverse.
Capacity is where the thing stops. On the 5% thin-band rules an account of Rs 2.5-5 lakh earned 16.79bp a day; Rs 10 lakh, 15.40; Rs 25 lakh, 8.97; Rs 50 lakh, 5.81; Rs 2 crore, 2.39. Orders have to stay small next to a stock trading Rs 5-25 crore a day, and only so many such stocks gap down on a given morning. The measured edge also breaks if assumed slippage is understated by a factor of 2.0, the wider variant lasting to 2.5x — far thinner headroom than the large-cap gap book, which tolerated something like 5x.
The caveats deserve equal billing with the number. Survivorship comes first: the stock list is today's Nifty 500, so companies that shrank out of the index or blew up are absent from the history, and this book buys precisely the names that just fell 5-7%, making that band the most exposed of any tested. Next, the fill is an assumption — nobody has pushed a real order into a thin stock's pre-open auction under these rules, and the cost schedule is a conservative guess, not a measurement. Then the losses: worst falls of 18.5% to 25.9% on the thin band and 40.7% to 43.4% on the wider one, tail-heavy by construction, since returns arrive in a handful of violent days. Finally, it does nothing at all on 90-95% of sessions.
Put together, this is what a positive result looks like stated honestly: narrow, capacity-bound, resting on a fill nobody has placed, drawn from a panel that flatters the past. It cleared the cost wall that killed the other intraday ideas tested here, without leaning on 2020 or any single year. The research's own next step is to run it forward on paper.
More money stories
- Four attempts to beat intraday trading costs failed -- the barrier is the fill, not the signal
- A Hormuz deal cooled oil, the AI trade stumbled, and Asia's markets split in two
- A 274-point open in New York, a 544-point climb in Mumbai: the morning two markets turned green at once
- Foreign funds kept selling India. The people running the companies kept buying.
- Oil fell 5% because an attack did not happen, and that says a lot about what was in the price
- A 12.3 billion won crypto scam, a group in custody, and a news line that leaves the rest blank