Most intraday ideas we tested did something real. Stocks that had run up hardest on heavy volume in the first half hour did, on average, lag the market for the rest of the day. The effect showed up however we sliced it, with a t-statistic above 9. If you stop reading a backtest at the t-statistic, that looks like money.
Then you look at the size. Shorting the most stretched stocks returned 0.24% to 0.37% a day before costs. About 0.18% of that was simply the market falling on most days in the sample: a random short would have earned it too. What was left for actual skill was roughly 0.06% to 0.18% per trade.
Whether that is a business or a hobby depends entirely on what one trade costs.
What one intraday trade costs
Take a ₹50,000 margin trade at 5× leverage: a ₹2,50,000 position, bought and sold the same day. These are the charges at the rates in force when we measured them:
| Charge | How it's charged | ₹ |
|---|---|---|
| Brokerage | ₹20 cap, both legs | 40.00 |
| STT | 0.025%, sell side | 62.50 |
| Exchange fees | 0.00297%, both legs | 14.85 |
| Stamp duty | 0.003%, buy side | 7.50 |
| SEBI fee | 0.0001%, both legs | 0.50 |
| GST | 18% of brokerage + exchange + SEBI | 9.96 |
| Total | 0.054% of the position | 135.31 |
That is less than many people expect. At this size the ₹20 cap makes brokerage a small part of the bill, and going bigger barely helps: at ₹10 lakh the charges are still about 0.040%. Charges are paid on the position, not on the margin, so leverage does not make a trade any cheaper per rupee traded.
The ₹20 cap works the other way for small trades. On a ₹20,000 position, the ₹40 of brokerage alone is 0.2%, larger than every real effect we found.
Rates change. Check the numbers against your own broker's charges calculator before you rely on them.
The cost nobody prints: the spread
Every time you buy at the ask and sell at the bid, you pay the gap between them. It is not on the contract note, so it is easy to forget. It can't be read straight off 1-minute candles either. We estimated it for every stock each morning from the previous day's 1-minute closes (Roll's estimator), so the figure is known before the market opens.
| Stocks, by daily turnover | Typical turnover | Spread, round trip |
|---|---|---|
| Most thinly traded tenth | ₹2.6 cr | 0.099% |
| Middle tenth | ₹24 cr | 0.073% |
| Most liquid tenth | ₹416 cr | 0.030% |
Add that to the 0.054% of charges and a realistic round trip runs from about 0.08% on the most liquid stocks to 0.15% or more on thin ones. It is wider again on a stock that is moving fast, which is exactly when most intraday strategies want to trade.
Why the spread decides everything
We scored 684 versions of one strategy family: different filters, stops and time windows. The edge and the t-statistic do not depend on cost, because each version and its comparison pay the same charges. Only the net result moves. Here is how many of the 684 were profitable after costs, for different spread assumptions:
| Cost assumed per round trip | Net positive | …and statistically solid |
|---|---|---|
| Charges only (0.054%) | 179 | 14 |
| + 0.05% spread | 63 | 13 |
| + 0.10% spread | 16 | 6 |
| + 0.20% spread | 1 | 0 |
Between a 0.05% and a 0.20% spread, the answer goes from "a dozen strategies work" to "nothing works". Same trades, same data. The spread is the most important number in an intraday backtest, and most backtests don't have it at all.
What to do with this
- Measure the spread per stock, per day, from data you had before the trade. One flat number for every stock is wrong in both directions. Ours charged the most liquid stocks about three times their real spread and undercharged the thin ones.
- Subtract the market. If a random short over the same minutes also made money, that part is not your edge.
- Always report the net. A positive edge with a negative net result is not a trade. It is an interesting fact about the market.
- Size is the only lever on cost you control. On small positions the fixed ₹20 brokerage eats the edge. By about ₹2.5 lakh it barely matters, and the cost that's left is the spread, which only liquidity changes.