A great Sharpe ratio on a tiny edge

The trap

A backtest reports a Sharpe ratio of 3 or more, or a return-to-drawdown ratio near 10. Both look like the signature of an outstanding strategy. Neither says how many rupees the strategy makes, and neither changes if you trade one lot or a hundred.

How it fools you

Ratios divide profit by risk. Shrink the position and both shrink together, so the ratio stays the same. That is useful for comparing strategies, and dangerous when the profit per trade is tiny:

Costs don’t scale the same way. Brokerage is charged per order, not per rupee of profit. A trade that makes 5 points per unit pays the same ₹20 an order as one that makes 500.
Slippage is measured in ticks, not percent. A backtest that charges “0.5% of the premium” on a ₹10 option is charging 5 paise, one tick. A real fill on a cheap, fast-moving option can easily be half a point or a point away.
A big capital base hides small numbers. Quote the drawdown as a percentage of ₹10 lakh and it looks negligible. So does the return.

When the edge is a few points, the ratio describes the shape of the equity curve, not whether the strategy can pay its own costs.

A real example

A rule we came across in strategy listings online: on every SENSEX weekly expiry day at 10:30, sell the call and the put that are the cheapest still priced at ₹10 or more, with a 9-point stop on each leg, and close whatever is left at 15:20. One lot (20 units).

Data: 1-minute SENSEX option prices, 156 expiry days from August 2023 to August 2026. The two options together brought in a median of about 23 points, and were each about 850–930 points from SENSEX. The median trade made 3.5 points before costs.

Cost model	Avg / trade	Win %	Sharpe	Return ÷ worst drawdown	Per lot, per year
No costs	+8.8 pts	92%	4.3	8.4	₹9,100
0.5% of the premium as slippage	+8.6 pts	92%	4.2	8.1	₹8,900
Charges at one lot, no slippage	+4.0 pts	44%	2.2	2.7	₹4,200
Charges + 0.5 pt per leg each side	+2.0 pts	42%	1.1	1.2	₹2,100
Charges + 1 pt per leg each side	0.0 pts	40%	0.0	0.0	₹0

Charges are built up from the charge schedule at ₹20 brokerage per order, one lot per order: about 4.8 points per trade, of which 4.7 is brokerage and its GST. Sharpe is on daily profit, annualised, with no risk-free rate.

Three things happened as the costs became realistic:

The “flattering” cost model changed almost nothing. Charging slippage as a percentage of a ₹10 premium cost 0.2 points a trade. The ratios barely moved, which is exactly why it feels safe.
Flat costs ate the edge. At one lot, brokerage alone is a point per unit per order, four orders a trade. That is more than half the gross profit. The win rate fell from 92% to about 44%, because most “wins” were smaller than the cost.
The ratios collapsed with it. Sharpe went from 4.3 to zero, return-to-drawdown from 8.4 to zero: at a point of slippage a leg, the strategy made nothing in three years.

Now the rupees. Even with no costs at all, one lot made about ₹9,100 a year against roughly ₹2.5 lakh of margin: 3.6% a year, below a deposit. Measured against a ₹10 lakh account, the worst drawdown was 0.1%, and so was the “risk”. Subtract a 6.5% deposit rate on the margin and the Sharpe ratio is negative before any cost is charged.

How to spot it

The average trade is a few points, but nobody shows it next to the cost per trade.
Slippage is charged as a percentage of a cheap option’s price.
Brokerage is missing, or charged as if every order were ten lots.
Returns and drawdowns are quoted as a percentage of a large account that the strategy barely uses.
The Sharpe ratio is calculated without a risk-free rate on the capital the trade locks up.

How we guard against it

Every result on the bench shows the average trade in points next to its cost in points, uses flat slippage per leg for options, and states the lot size per order behind the brokerage figure. Capital is the margin plus the worst drawdown, and the return on it is compared with a deposit. See Methodology → Costs and Wrong costs, in either direction.