Five ways to build a 0.5% intraday trade, and the cost wall that stopped all of them
An operator asked Strota a blunt question: could the system generate intraday trades with targets of 0.5-1%? The appeal is obvious -- a half-percent move shows up many times a session, and a rule that takes it wins most of the time. Five backtests were built to answer it. All five failed, and they failed for the same reason, which is the only thing that makes a pile of dead tests worth writing up.
Cost is the wall, and it stands in two courses. The fixed course is the round trip itself: on a Rs 1,00,000 intraday position, brokerage, taxes and exchange charges come to 0.0824%, and that figure does not shrink when the target does. So the win rate needed just to break even explodes as the target narrows. A setup taking 0.5% or losing 0.5% needs 58.3% winners with flawless fills, and 78.3% once you allow 0.10% per side of slippage -- the gap between the price you aimed at and the one you got. Widening the target to 1% against a 0.5% stop pulls that down to 38.8%. The narrower the target, the larger the share of it that belongs to the broker.
The first study bolted small targets onto the strongest setup the project already had: buying stocks that had gapped down 5% or more while otherwise trending up, on daily bars from 2016 to 2026 across a Nifty-500 universe, 935 trades, with 0.05% per side charged for slippage. Holding to the close won 71.0% of the time and returned +2.722% per trade. Capping the exit at 1% lifted the win rate to 88.8% and cut the average trade to +0.275%. A half-percent cap won 92.5% of its trades and lost money anyway. None of the nine target-and-stop pairs beat holding.
Why? The payoff lives in the right tail. Median movement from open to close on a 5% gap-down was +2.7%, so a one-percent ceiling amputates exactly the stretch of the distribution that funds everything else. Dropping the stop did not rescue it -- stop-free versions still trailed holding by 2.4 percentage points a trade. Targets bought a win rate between 85 and 95% and paid for it in cash.
Two further studies chased intraday mean reversion. One removed the market's own move by construction and traded stocks against their peers on five-minute bars: none of its 56 configurations made money even at zero slippage. The effect was statistically real, at +0.057% a trade before costs, but against the 0.0824% charge the whole measured edge amounts to roughly one commission, and five of six years lost. The other bought dips beneath the day's volume-weighted average traded price and did worse -- all 32 configurations lost at every slippage level including zero, one of them winning 62.2% of its trades while losing 0.21% on each.
Genuine structure did appear in the fourth study, and it still could not be traded. Returns earned during the trading day drifted at -39.0% a year while returns held overnight ran +58.7%, with 41 of the 43 symbols drifting down during the session. That drift was concentrated rather than smeared: the full day lost 12.9 basis points -- one basis point is a hundredth of a percent -- and the opening thirty minutes carried 8.1 of them, about 63% of the day's drift inside 8% of its length.
Then the arithmetic. Roughly 11 basis points of gross daily drift, against 8.2 basis points of cost, leaves nothing: exactly 1 of 144 tested cells came out positive, and only with slippage set to literally zero. Out of sample the sign flipped outright, to +11.5% a year.
Run purely as a control, the fifth study proved the most useful. It tested breakouts from the opening range in both directions and, as expected, produced nothing deployable: -0.010% a trade on the up-break, -0.046% shorting the down-break, -0.319% fading it. The value was in the autopsy. A stop order resting at the range high gets filled, systematically, near the top of the bar that triggers it. Measured against a fill at a bar's open, that cost -0.340% a trade, with gross falling from +0.515% to +0.172%. It is about four times the entire fee load, it is structural, and any entry that triggers through a level pays it.
One of the project's own beliefs died here too. Earlier work had suggested that buying weakness works while buying strength fails; the fade side of this test was the single worst configuration in the whole study, below every draw of its random-entry benchmark, while both momentum directions sat at the top of that same benchmark. On like-for-like fills the down-break bar is a marginally better moment than a random one -- by an order of magnitude too little to offset facing the wrong way. Not chasing strength is a claim about timing, not about direction.
A positive average trade also turned out not to mean a positive account: the equal-weight daily book lost for every in-sample configuration at every slippage, because busy trending days carry the pooled number. Day selection, not the signal, did the work.
In practice, small fixed targets here are not mis-set, they are the wrong instrument. They shrink what a trade can win while the cost of taking it stays put, and every intraday effect measured in this project is worth between 2 and 16 basis points before costs. Only one exit rule came out ahead of its alternatives anywhere in the five studies: holding to the close. A small target can only bind on a strategy whose per-trade edge is far bigger -- here, only the gap book at +0.7% to +1.5% a trade -- and there it merely reduces variance, at a cost of roughly 2.4 percentage points of expectancy. That describes past data; it is not a recommendation.
The limits are worth stating. Three of the five studies leaned on a 43-name universe originally chosen for volatility, so they are thin on breadth; a larger set of 352 symbols exists and follow-up work should use it. The +58.7% overnight leg is far too big to believe for mid-cap names and is probably inflated by the fact that those names were picked after the event, a caveat that travels to any overnight-hold idea built on the same list. And the gap-down edge decays hard with liquidity, from about +3.17% a trade in the smallest turnover band to +0.14% among the largest on only 57 trades, so numbers drawn from thinly traded names are partly a fantasy about fills.
More money stories
- Strota backtested its own Darvas box screen: no edge once you pay the fees
- Four attempts to beat intraday trading costs failed -- the barrier is the fill, not the signal
- A rare positive in our backtests: deep gap-down buying cleared the cost wall, at retail scale only
- Gold crosses Rs 1.5 lakh per 10 grams, and silver is the tighter story
- 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