DEV Community

Rahul Goel
Rahul Goel

Posted on

One Minute Is a Long Time: What 1-Minute Option Data Hides in Stop-Loss and Target Tests

Most retail backtests of index option selling run on one-minute bars. Each bar gives four prices: open, high, low and close. From those four numbers the software decides whether your stop-loss was hit, whether your target was hit, and at what price you got out.

The trouble is that a minute is not an instant. In sixty seconds, a NIFTY or SENSEX weekly option on expiry day can trade up, down, and up again. The bar records the extremes but not the order. If your stop sits inside the bar's range and your target also sits inside it, the data cannot tell you which one came first. The software has to guess, and it usually guesses in a way that flatters you.

This article is about that guess. It explains where one-minute data is blind, shows with simple numbers when the blindness matters, and gives a method to put a fence around the error instead of ignoring it. It is written for someone who sells index options on rules and wants to know how much to trust a stop-loss test. It does not recommend any trade, and the example figures are illustrative unless stated otherwise.

What a bar tells you, and what it does not
A one-minute option bar usually comes from last traded prices. Open is the first trade in the minute. High is the highest trade. Low is the lowest. Close is the last. It does not tell you the order of trades in between, how many contracts changed hands at each level, or what the bid and ask were at any moment.

For a short option leg, the high is the adverse extreme, since a rising option price is a loss to the seller. The low is the favourable extreme. So the high tests your stop, and the low tests your target. A simple engine checks: if high is at or above the stop price, the stop fired. If low is at or below the target, the target fired.

That works when only one of them is touched inside a bar. It breaks when both are touched in the same bar. And it hides three further things: the price you actually get at the stop, the effect of the spread, and the possibility that the high was a stray trade. The next sections take each one.

Both levels inside one bar: the ordering problem
Take an illustrative short ATM call sold at 60 points. You use a tight rule: stop at 66 (10% above entry) and target at 54 (10% below). The next one-minute bar prints open 60, high 67, low 53, close 61.

Both levels were touched. Perhaps the price fell to 53 first, hit your target and you banked 6 points. Or it spiked to 67 first, hit your stop and you lost 6 points. Same bar, opposite results, a swing of 12 points on a single trade, which is Rs 780 on one lot of 65.

Different engines resolve this in different ways, and the choice is rarely shown on the report. Three common rules exist.

The first is "target first". The engine checks the favourable level before the adverse one. This is the optimistic rule. It turns every ambiguous bar into a win.

The second is "stop first". The engine checks the adverse level first. This is the pessimistic rule. It turns every ambiguous bar into a loss.

The third is "open-distance". The engine guesses that the price visited whichever level was closer to the open first. In the example the open is 60, the stop is 6 away and the target is 6 away, so it is a tie, and the engine needs a tie-break anyway. Some engines use the candle colour: if close is above open, assume the low came first, then the high. This has some intuition behind it, but little evidence, and it is still a guess.

No rule is correct. Each is an assumption. What matters is how many trades fall into the ambiguous group, because only those trades depend on the assumption.

Ambiguity gets worse as the day goes on
Here is a point that is rarely mentioned. The share of ambiguous bars is not constant through the day. It depends on how big the bar's range is relative to the distance to your stop and target, and both of those change as premium decays.

When you express stops and targets as a percentage of premium, the rupee distance shrinks as premium falls. A 10% band around a 60-point option is 6 points each way. The same 10% around a 12-point option is 1.2 points each way. But the typical size of a one-minute range does not shrink in proportion. Option prices move in ticks of 0.05, spreads do not shrink to zero, and spot continues to move. So a one-minute range of 2 points is 3% of a 60-point option and 17% of a 12-point option.

The result: late in the session, a one-minute bar is much more likely to span both your stop and your target. If you run percentage-based rules that also apply to afternoon entries or re-entries, the ambiguous share can be much higher than in the morning. A backtest that reports one overall number hides this. A backtest split by time of day shows it.

The desk data helps frame the scale. EliteAlgo's study of 163 NIFTY and SENSEX weekly expiry sessions between January 2025 and July 2026 measured the ATM straddle premium at 15-minute marks, as a percentage of its 09:16 level. It reads 95.4% at 09:31, 89.0% at 11:01, and between 88% and 93% for the whole afternoon (for example, 89.1% at 13:01 and 90.9% at 15:01). The strike is fixed at the 09:20 level, so the straddle premium does not fall to near zero on average, because theta decay is offset by gamma. That means the average straddle stays a meaningful size all day. But individual legs behave differently: when spot trends, one leg decays to small values while the other grows. The small leg is exactly where the percentage-based ambiguity is worst.

The trigger price is not the fill price
The second blind spot is the price you receive. Most simple engines fill a stop at the stop price itself. Real stops do not work like that.

Consider the two kinds of stop order. A stop-market order becomes a market order when the trigger is hit, so you get whatever price the book offers at that moment, which in a fast move is worse than the trigger. A stop-limit order becomes a limit order, so you get at least your limit price but might not get filled at all if the price runs through. Neither gives you "the trigger price" reliably.

The gap case is the clearest. Suppose your stop is at 66 and the next bar opens at 71 because spot jumped between bars. A correct engine fills at the open, 71, not at 66. A lazy engine fills at 66 and pretends the gap did not happen. On expiry day, with fast spot moves and thin strikes, such gaps are common. Each one shifts the loss from the 6 points the rule planned to 11 points in this example, which is nearly double.

The fix is a simple fill rule for a short stop: fill price is the higher of the bar open and the trigger price, plus a slippage allowance for the part of the move that happened inside the bar. For a target, the opposite: fill at the lower of open and trigger, and do not give yourself extra for a favourable gap if you are using a limit order, since a resting limit order fills at your price (or better if the book allows) but you cannot count on a better price in a test.

Check also what the engine does with the entry. If you plan to enter at 09:20, does the engine use the open of the 09:20 bar, or the close of the 09:20 bar? The close of that bar is known only at 09:21. Using it as an entry price is a look-ahead error. It is a small one on average and a large one on the days when the first minute moves a lot.

Last-trade prices are not bid and ask
The third blind spot is the spread. A bar made from last-traded prices tells you where trades happened, not where you could have traded. A short seller sells at the bid and buys back at the ask. The gap between them is a cost you pay on each round trip, and it is missing from a bar-based test unless you add it.

For liquid at-the-money NIFTY weeklies the spread may be a small fraction of a point most of the time. For out-of-the-money hedge legs and for far strikes it can be much wider. During fast moves it widens for everyone. So the cost is not constant. It is lowest when nothing is happening and highest when your stop fires.

There is a related problem: stray prints. A high that is a single trade at an unusual price, say a small order hitting a thin ask, can trigger a stop in the test that no real stop order would have seen, since a real stop is usually keyed to the last traded price in the live market but your fill depends on the book. The reverse also happens: a real stop fires on a print that your bar's high smoothed over. For illiquid strikes, one-minute highs and lows are noisy.

When you read reviews of options backtesting software, the useful questions are the ones about these rules: which resolution rule for ambiguous bars, which fill rule for stops, and whether a spread allowance can be set. Feature lists usually skip them. Whatever tool you use, find out the answers from its own documentation before trusting a stop-loss result, and if the documentation is silent, treat the answer as unknown and test it yourself with the method below.

A quick way to size the effect on your own data: take a sample of 50 stop-outs from a test and look up the actual live fills for similar stop-outs from your own trading, if you have them. The difference between the planned loss and the fill loss, averaged, is your slippage on stops. Many traders find it is a substantial fraction of the planned stop distance, particularly when stops are tight.

Bounding the error: the three-run envelope
You cannot remove the uncertainty from one-minute data. You can measure how much it matters. The method is to run the same test three times with three sets of assumptions, and report the range instead of a single number.

Report the total result, the win rate and the drawdown for each run. The gap between the optimistic and pessimistic runs is the width of your uncertainty. If a strategy makes money in all three, the one-minute problem is not what is carrying the result. If it makes money in the optimistic run, breaks even in the middle and loses in the pessimistic one, the whole edge lives inside the assumptions, and you should not trade it on this evidence.

Here is an illustrative case. Suppose a rule set produces 200 trades over a test. Of those, 30 bars are ambiguous (15%). The average win on those trades is 6 points and the average loss is 6 points. Under target-first, all 30 are wins: +180 points. Under stop-first, all 30 are losses: -180 points. The swing between the two extremes is 360 points, or Rs 23,400 at 65 per lot. If the rest of the trades net +100 points in total, the strategy shows +280 points under the optimistic rule and -80 points under the pessimistic one. The same data and the same rules produce a winning and a losing strategy. That is the size of the thing you are hiding when you use a single rule without saying so.

Notice also that the ambiguous share, 15% in this example, is itself the most useful single statistic. If it is 2%, the problem is small and the three runs will look nearly identical. If it is 25%, your rules are too fine for one-minute data and no choice of assumption will rescue them. In that case you need finer data (tick data, or at least bid-ask snapshots) or looser rules.

How to count your ambiguous trades
This takes an hour for anyone comfortable with a spreadsheet or a few lines of code. For every trade in your test, at each bar after entry, check two conditions at once: did the high reach the stop, and did the low reach the target? Mark the bar as ambiguous if both are true. Also mark bars where the open already lies beyond the trigger, since those are gap fills and need their own fill price.

Then produce four numbers: the share of trades that ended on an ambiguous bar, the share that ended on a gap fill, the average points lost to gaps against the trigger price, and the same ambiguous share split by time of day in hourly buckets. The last one tells you whether your afternoon logic is built on guesses.

If you cannot get these numbers from your tool, export the one-minute data and compute them yourself. Most tools allow an export of the trade list but not the logic. If yours allows neither, that is itself a result: you cannot audit the test, so you cannot weigh it.

Another useful check is sensitivity to a one-minute shift. Re-run the test with entry delayed by one minute, and with every stop trigger set one tick wider. A robust result moves a little. A fragile one swings. A result that depends on a precise one-minute alignment is not telling you about the market.

What this means for studies built on 1-minute marks
The same caution applies to published statistics, including the ones used here. EliteAlgo's desk data on peak-to-close give-back for a short ATM straddle entered at 09:20 covers 159 usable sessions and finds a median give-back of 42.8% of peak profit, with 86 of 159 sessions (54%) giving back more than 30%, and 48 of 159 (about 30%) closing at a loss after showing a profit at some point. These are measured on one-minute option-chain data.

A one-minute mark can miss the true intraday peak. The best profit of the session may have occurred in the middle of a minute that closed lower. If so, the real peak is at least as high as the measured one, and the real give-back is at least as large. So the direction of the error probably works against the comfortable reading: the give-back is likely understated, not overstated. That is a modest claim and it rests on the plain fact that a bar's extremes lie outside its close. It does not turn these figures into forecasts, and the same data can say nothing about how tick-level exits would have performed.

The same logic applies to the max-pain figure: the median distance between expiry close and max-pain strike was 34.4 points (0.083% of spot) in that study, measured from an end-of-day open-interest snapshot. A snapshot is a point-in-time measure. It is a fair description of where the close sat relative to positioning, and it is silent on the path that got there.

The takeaway is not that one-minute data is bad. It is the best data many retail traders can get, and it supports many useful questions. The takeaway is that it supports some questions well (where did premium sit at 11:00?) and others poorly (which came first inside this minute?). Match the question to the resolution. Stop-loss and target ordering is the one most often asked of data that cannot answer it.

A short checklist before you trust a stop-loss backtest
Ask these questions of any test, your own or someone else's. How does it resolve bars where stop and target are both touched? What fill does it assume for a stop, and what happens on a gap through the trigger? Are spread and slippage included, and are they fixed or do they widen with speed? What share of trades end on ambiguous bars, and how does that share change by time of day? What happens to the result under the pessimistic run?

If the test can answer all five, it is worth reading. If it cannot, it is a rough sketch, which is fine as long as you do not put real money behind it.

Finally, check the regulator's and exchange's pages, such as nseindia.com, for contract specifications, tick sizes and trading hours. Tick size and expiry-day rules shape how coarse the price grid is, and a grid of 0.05 is coarse for a leg trading near 2 points, so a stop expressed as a percentage can land between ticks.

Frequently Asked Questions
Does a stop-loss work the same in a backtest as it does live?

Not exactly. A backtest on one-minute bars usually fills the stop at the trigger price, ignores the spread, and picks a side when stop and target fall in the same bar. Live stops fill at the available price, which is often worse in fast moves. Measure the gap on your own fills.

Which fires first, the stop-loss or the target, inside one candle?

One-minute data cannot say. The bar records the high and low but not their order. Run the test under both orders and report the range. If the result changes sign, the data cannot support the rule.

Is tick data necessary for options backtesting?

Not always. If your stops and targets are wide compared with a bar's typical range, one-minute data is fine. If they are tight, or you trade late in the day when premium is small, finer data or looser rules are safer.

How much slippage should I add to a stop-loss backtest?

There is no single figure. It depends on the strike, the time of day and the speed of the move. Start from your own live fills for similar stops, and use the pessimistic run to see how much the result depends on it.

Why does my live result differ from the backtest on expiry day?

The usual causes are fills at the stop that are worse than the trigger, spreads that widen in fast moves, gap openings between bars, and tie-breaks in ambiguous bars that favoured you. Count each one using the method above.

Should I use limit or market orders for stops on options?

Each has a trade-off. A market order is filled but at an uncertain price. A limit order has a price floor but may not fill when the price runs through it. For a short option, an unfilled stop is the bigger danger, so many traders accept market-style exits with a protection band. Check the exchange and broker rules for what order types are allowed on index options.

Conclusion
One-minute option bars hide the order of events inside each minute. When a stop and a target both sit inside one bar, the data cannot tell you which came first, and the software's silent tie-break can decide whether a strategy wins or loses. The problem gets bigger late in the day, when premium is small and a minute's range is a large share of it.

You can fence the error in. Count the ambiguous trades. Run optimistic, middle and pessimistic versions with fill and spread rules written down. Report the range. A strategy that survives the pessimistic run has earned some trust. One that does not has told you something useful before it cost you money.

Algo trading carries a real risk of loss, and a careful backtest does not remove it. EliteAlgo is analytical software and not an investment adviser, and nothing in this article is investment advice. The worked figures are illustrative, and the session statistics come from past data that does not predict future results.

Top comments (0)