Why most intraday strategies lose money after costs

Intraday trading looks cheap per trade. Multiply by hundreds of trades, compare with how small the typical edge is, and the arithmetic turns against the trader.

Vitta-Kosh team · Updated 26 September 2026 · 8 min read

This is educational content, not a recommendation. It reports a past back-test that we ran for our own research, on groups of stocks as a whole. It names no stock, offers no strategy and is not a forecast. It does not say that nobody can trade profitably; it shows how large the hurdle is.

What we tested

We tested 71 variants of six well-known intraday approaches: opening range breakout, pullback to VWAP, late-day momentum, relative strength against NIFTY 50, gap continuation and gap reversal, and VWAP mean reversion. Every variant was costed with the full broker and statutory charges and a small allowance for slippage (buying a little above and selling a little below the last price).

All 71 lost money after costs. On the training years, none of them had a result that could be told apart from chance (the best t-statistic was 0.4, where around 3 is needed). Only one family, late-day momentum, showed any edge before costs: 0.10% to 0.15% per trade. The costs of that trade were about 0.15%.

A second test: a signal scanner, replayed trade by trade

We also replayed a scanner built on two popular ideas, a short moving-average crossover plus candlestick patterns, bar by bar over 312 stocks from 25 March to 18 September 2026. It produced 1,356 trades. The profit factor is the money made on winning trades divided by the money lost on losing trades; below 1.0 the approach lost money overall.

Two lessons sit in this table. The signal itself had no edge: a profit factor of 0.95 before any friction is a coin flip that loses slightly. And friction then took it from roughly break-even to a heavy loss. On a Rs 50,000 trade the charges were about Rs 52 and slippage about Rs 50, together roughly 0.2% of the position on every round trip.

Changing the exits did not rescue it. Exit distances from 1 to 3 times the average true range, with profit-taking levels from 1.5 to 3.5 times the amount risked, all stayed below break-even, with profit factors from 0.54 to 0.66. The time of day mattered but never enough: entries between 09:15 and 11:00 were the weakest slice (profit factors 0.48 to 0.59, 58% of all trades and about 70% of the loss), and the best slice, 13:00 to 15:00, reached 0.69.

Why: the edge is smaller than the cost

Our research into how a stock's move breaks down found that roughly half of a liquid stock's move within a day is the market and its sector (see how much of a stock's move is just the market). A signal that looks at one stock in isolation is largely reacting to those shared moves, which is what a profit factor near 1.0 before costs looks like.

We then removed the market and sector parts and looked only at each stock's own movement, across 96 combinations of look-back window, time of day and holding period. There is a real, repeatable pattern: within a day, a stock that runs away from its sector tends to come partly back. But its size is tiny. The best long-short spread in the whole search was 0.06% per trade, and most were under 0.04%, against about 0.30% of costs to trade it long and short at retail rates. The edge was five to ten times too small to pay its own costs.

No choice of indicator, filter or pattern closes a gap of that size. It is a property of the costs, not of the signal.

Where the costs come from

• STT on intraday equity is charged on the sell side. It is small per trade but paid on every trade.
• Brokerage at discount brokers is usually Rs 20 per order or a small percentage, whichever is lower, so up to Rs 40 for a round trip. On small positions that flat fee is a large percentage.
• Exchange charges, SEBI fees, stamp duty and GST add a little more.
• The bid-ask spread and slippage are not on any contract note, but you pay them every time you cross the spread.

Each of these is a fraction of a percent. The problem is frequency: a strategy that trades every day pays them hundreds of times a year, while the edge it captures on each trade is of a similar size or smaller.

What this does and does not show

• It covers one market (NSE cash equities), liquid stocks, and roughly two years of data.
• It covers widely known rule-based approaches, not every possible method.
• It shows that costs set a high bar, and that the approaches tested did not clear it in this period.
• It describes the past and makes no prediction about any stock or strategy.

If you trade at all, knowing your real cost per round trip is the first number to get right. See what a share trade really costs for delivery trades, where the cost is paid once over a longer holding period.