PaperTrade

We tested the most popular intraday setup in India. It loses money.

Buy the day’s biggest gainers at 9:16, sell the biggest losers, exit when the move slows. Every trading channel in the country teaches some version of it. We backtested it on 3,016 real trades across 208 F&O stocks and two years of one-minute data. It wins 46% of the time and loses money in every construction we tried.

What we tested

Every trading day from October 2024 to August 2026, we ranked all 208 F&O stocks by their opening gap, took the three biggest gainers and three biggest losers, and entered at 9:16 in the direction of the move - the setup as it is usually taught.

Prices are the exchange’s own one-minute bars. A leg only counts as tradeable if it actually traded that minute. Two years, 505 sessions, 3,016 trades.

The result

At a one-to-one risk and reward, where the win rate is the whole answer, it won between 45.7% and 48.4% depending on how wide the stop was. Never above 50%, which is break-even before costs. With brokerage and STT it is worse.

The exits made no difference. A momentum-decay rule, a slower one, and simply holding to 9:30 all produced a negative average. Nor did the selection: ranking by gap, by gap times relative volume, by gap against the stock’s own volatility, or taking only gaps above 5% - all four lost.

The most uncomfortable number: a randomly chosen mid-ranked stock did BETTER than the day’s biggest gapper in two of the three tests. Being the biggest mover carried no information at all.

How an 85% win rate loses money

We were asked to build a strategy with a high win rate, so we did. Same trades, same entry - we only moved the target and the stop. At a 10% stop and a 0.15% target it wins 85.6% of the time. It loses about ₹45 per trade after costs.

Nothing improved. A tiny target and a wide stop makes wins frequent and small, losses rare and enormous. That is the mechanism behind every high win rate you will be shown, and it is why a win rate quoted without its average loss tells you nothing.

Watch what happens as the win rate climbs: 45.7% loses ₹117 a trade, 64.1% loses ₹117, 78.7% loses ₹63, 85.6% loses ₹45. The win rate quadrupled. The money never turned positive.

We tested selling premium too

Credit spreads on NIFTY weeklies, 98 real expiries, actual option closes, costs and slippage charged. Bear call spreads at 350 points out won 83.6% of the time and lost ₹188 a week. Iron condors won 64.4% and lost ₹593.

Most lost before costs were even applied, which means there was no edge for brokerage to eat. And the one strategy that did make money - selling calls - made it because NIFTY fell 6.7% over the test window. Selling puts, the mirror bet, lost in both periods. That is a short position in disguise, not a premium edge.

Why we are telling you this

PaperTrade sells a practice simulator. It would be easier to put "83% win rate" on the strategy cards and let you find out the rest yourself. Instead every strategy in the app shows what it measured next to the base rate - how often the stock simply fell anyway - and four of the six say plainly that they were no better than that.

A simulator cannot make you profitable. What it can do is let you find out that a setup does not work using virtual money, over an afternoon, instead of finding out with real money over a year. That is the entire product, and this page is the clearest example of it we have.

The caveats, because they matter

This is two years of Indian data, not a universal law. It says the setup did not work on these stocks over this period at this resolution - it does not prove no intraday strategy can work.

One-minute bars are the finest history available, and the moves resolve faster than that. Perfect hindsight over the same trades captured 0.9% per trade; every rule we tried captured about 2% of that. A tick-level version is genuinely untested.

And the costs are modelled, not measured. Historical option data carries no bid-ask spread, so slippage is an assumption stated alongside every result rather than something we could observe.