By Shashank · Updated
1. Trade at a price you could have had
A rule that reads today's close can't also trade at that close. Test the fill at the next day's open.
2. Charge every cost
Include brokerage, tax, exchange fees and slippage, how far the price moves against you. Then rerun with worse prices and see what's left. Our guide to delivery trade costs lists each charge.
3. Compare with random picks
Run the same entry rules, stake size and costs on stocks chosen at random from the same list, many times. If random picks do about as well, the result came from the set-up. The choice of stocks added nothing.
4. Remove the biggest winners
Take out the five stocks that made the most money and rerun. If the result collapses, it rested on a few names you couldn't have known in advance.
5. Change the start date
A result that holds for only one start date is fragile. Run several start dates and look at the worst one.
6. Allow for how many versions you tried
Try enough settings and one will look good by luck. A luck-adjusted test, such as the deflated Sharpe ratio, allows for this. So keep a count of what you tried.
What is survivorship bias in a backtest?
Survivorship bias means testing only on companies that still exist, the survivors. A March 2026 preprint on the NIFTY Smallcap 250 estimated that testing only current members inflated returns by about 4.9 points a year. It's a single early study, so treat the number as provisional.
Our own price data has this gap. Delisted stocks are missing, which makes results look better than they would have been. Any backtest should tell you where it stands on this.
Questions people ask
How do you know if a backtest result is just luck?
Keep a count of the versions you tried, and use a luck-adjusted test such as the deflated Sharpe ratio. Also run the same rules on randomly chosen stocks many times. If random picks do about as well, the result came from the set-up.
Should a backtest trade at the day's closing price?
Not if the rule reads that same close. A rule can't trade at the price it is still reading, so test the fill at the next day's open.
How much does survivorship bias inflate backtest returns?
A March 2026 preprint on the NIFTY Smallcap 250 estimated about 4.9 points a year when only current members were tested. It's a single early study, so treat the number as provisional.
Does Tickjury's data include delisted stocks?
No. Delisted stocks are missing from our price data, which makes results look better than they would have been.