Protective Put or Stop Loss: What Downside Protection Actually Costs

Two instruments, one job, and one difference that matters

Both positions start from the same place. You own something, you have decided how much of a fall you are willing to sit through, and you want the rest of the downside to belong to somebody else. The stop loss does this by leaving a standing instruction: if the price touches this level, sell. The protective put does it by buying a contract from a counterparty who agrees to make up the difference below a strike.

The mechanics of the put itself are covered in depth in our guide to hedging, which works through what a protective put does to a payoff, why the premium raises your breakeven, and how a rolled put compounds into a drag through a calm market. That page is the parent here and this one does not repeat it. It also sets the put beside a covered call, which is the comparison people usually reach for and the wrong one for this question, because a covered call is not downside protection at all. The comparison nobody has run is the one a holder actually faces: put or stop.

The mechanics of the stop are equally well covered. That a stop is a trigger rather than an order, that a released order walks the book, and that no instruction can execute inside a price range where nothing traded, are all worked through in our pages on how a stop order actually behaves and on what a gap is. This page takes all of that as read. Its job is the part those pages leave open, which is how much any of it is worth in rupees, set against what the other instrument charges for the same protection.

One structural difference drives most of what follows, and it is worth stating plainly before any numbers appear. A put settles against a fixed strike on the closing level of the underlying. The contract does not care what route the price took to get there, and there is no execution to get wrong. A stop is an instruction to trade, and an instruction can only be filled at a price somebody is willing to trade at. Everything else on this page is a consequence of that one asymmetry.

The model, stated so it can be disputed

The exercise uses a simulated index-like series rather than a live instrument. That keeps every assumption visible, makes the run reproducible, and avoids implying that any particular Indian security would have behaved any particular way. A live series would give one history; the point here is the distribution of histories, which is exactly what a single history cannot show you.

The one modelling choice that matters more than all the others is that each day is built in two pieces. An overnight move happens first, carrying part of the day's variance plus occasional jumps, and then a continuous session runs from the open to the close. That split is the whole reason the exercise is worth running. A model where price moves continuously through the night cannot represent a gap, and a model that cannot represent a gap will conclude that a stop loss is a perfect floor, which is the error this page exists to price.

Two of those rows deserve a note because they are where a sceptical reader should push. The overnight share is an assumption, not a measurement, and it is the single input the conclusion is most sensitive to, which is why it is swept across a wide range later on rather than asserted once. The zero drift is a choice rather than a claim: the honest position is that this page has no view on what any market returns, and the effect of holding a different view is computed rather than argued.

The run also checks itself before it reports anything. Ten controls run first, including one that switches the gap channel off entirely and confirms the stop then fills exactly at the trigger every time, and a matching one that turns the gap channel violently up and confirms the measured shortfall moves by roughly fifteen times. A check that has never failed on a case you know to be broken is not a check.

The payoff, computed rather than sketched

Start with the diagram every options explanation opens with, and then notice what it cannot show. At the horizon, a position with a put attached has a known shape: below the strike the option makes up the difference, so the loss stops descending, and above it the put is worthless and you keep the move less what you paid. Here the put struck 8 percent below entry cost 0.6027 percent of notional, or about ₹10,848 illustrative. That premium moves the breakeven up by the same 0.6027 percent and puts the floor at 8.6027 percent below entry, about ₹1,54,848 illustrative, which is the protection level plus the price of the protection.

The stop cannot be drawn the same way, and the reason is the point of the figure.

A payoff diagram is a function of one variable, the closing price, and it works for a put because a put genuinely is such a function. Applying the same diagram to a stop quietly assumes the thing that is in question. A stop's result depends on the path, and the moment you accept that, the clean kinked line has to be replaced by a distribution. The band in the figure is that distribution, and in the region above the trigger it hangs a long way below the line, because some of the paths that ended there had been sold out near the bottom first.

This is not an argument that the stop is worse. In the region well below the trigger the stop's median outcome is slightly better than the put's, for the simple reason that the stop holder did not pay a premium. What the figure establishes is narrower and more useful: the two instruments cannot be compared on a payoff diagram at all, because only one of them is a payoff. Comparing them requires running the paths.

The same paths, run three ways

Here is the central experiment. Take the 200,000 simulated paths and run three regimes over each one: hold with no protection, hold with a stop at the trigger, and hold with a put struck at the same level. Nothing differs between the three except the protection, because the price paths are literally the same arrays.

The averages are almost identical and almost zero, which is what a driftless model is supposed to produce and is the first sign the machine is behaving. The unprotected mean and the put mean both came in at 0.0001 percent of notional and the stop at a loss of 0.0076 percent. If you were choosing on expected outcome alone you would be choosing between things that are the same, which is precisely why expected outcome is the wrong sole criterion for insurance and why so many comparisons of this kind reach a confident and meaningless conclusion.

The tails are where they part. The unprotected position's worst five percent averaged a loss of 15.72 percent, and its worst single path lost 34.45 percent. The stop's worst five percent averaged 9.40 percent and its worst single path lost 17.55 percent, which is more than double the eight percent it was set to prevent. The put's worst five percent averaged 8.6027 percent, its first percentile was 8.6027 percent, and its worst single path out of 200,000 was 8.6027 percent. Those are not three numbers that happen to be close. They are the same number, because a contract does not have a distribution of outcomes at its floor.

The spread of outcomes fell in the expected order too, from 7.99 percent unprotected to 7.53 with the stop to 7.06 with the put. The share of paths that finished with any loss at all went the other way, from 51.25 percent unprotected to 52.83 with the stop and 54.25 with the put, which is the premium and the slippage showing up as slightly more frequent small losses. That is the honest shape of the trade in both directions: protection makes the bad outcomes less bad and the ordinary outcomes marginally worse, and any description that only mentions one of those is selling something.

The whipsaw, counted, and the surprise inside it

The complaint every stop-loss user recognises is being taken out just before the price turns around. It is worth counting rather than grumbling about. In the base run the stop fired on 29.69 percent of paths, and on 47.18 percent of those firings the market finished back above the trigger before the horizon ended. That is 14.01 percent of the entire sample: roughly one path in seven where the holder was sold out and then watched the level recover. When it happened, the average distance between the exit and the closing level was 3.95 percent of the position, about ₹71,116 illustrative.

Those numbers move a great deal with where the trigger sits, and the direction is the one you would guess. Placing it three percent below entry produced a stop that fired on 69.39 percent of paths and left 32.89 percent of all paths sold out before a recovery. Placing it fifteen percent below entry cut that to 2.05 percent of all paths, at the cost of accepting a fifteen percent fall before the instrument does anything at all. That exchange rate between whipsaw and depth is the real content of the tight-against-wide argument, and our page on choosing a stop distance covers the discipline side of it.

Now the surprise, which took some checking before it was believed. Adding up what the whipsaw cost across every path gives about ₹9,502 illustrative per path on average. Adding up what the stop saved on the paths where the market kept falling gives about ₹10,848. Those very nearly cancel, and the second figure is not approximately the put's fair value, it is exactly the put's fair value, to the rupee.

That identity is not a coincidence and not a fluke of the seed. Any path that finishes below the trigger must have crossed the trigger to get there, so the set of paths where the stop saved something is the same set where the put paid something, and the amount is the same amount. The two instruments are removing the identical expected downside. And on a series with no drift, the times a stop takes you out before a recovery and the times it takes you out before a further fall offset one another, because that is what it means for a price to be as likely to go one way as the other from wherever it currently is.

What survives the cancellation is only the part that never had a chance to cancel: the imperfection of the exit. The execution shortfall came to about ₹1,486 per path on average, and the charges added roughly ₹567 more once you weight them by how often the stop actually fired. The total expected cost of the stop was about ₹706 per cycle on an eighteen lakh position, with a standard error of ₹109 on the pre-charge component and a ninety-five percent interval that includes zero. The put's expected cost at fair value was ₹28, which is its charge stack and nothing else, because a fairly priced contract returns its premium in expectation.

The uncomfortable implication is worth stating rather than hiding. On this model, at zero drift, neither instrument costs very much in expectation, and a page that told you stops are expensive because of whipsaw would be wrong. The whipsaw is real, it is frequent, it is painful, and in the average it is paid for by the times the stop was right. What it is not is free, because the exit is imperfect, and what it never was is a floor.

The gap is not a cost, it is a hole in the floor

This is the section the whole exercise was built for. That a stop cannot execute inside a gap is not news, and it is explained properly on our pages about gaps and slippage. What none of those pages does, and what the argument has always needed, is a number: given that a stop cannot cover a gap, how much of the downside does it actually stop?

Of the 59,371 paths where the stop fired, 25.27 percent were filled at an opening price that was already below the trigger. The market never traded at the stop level on those paths, so there was nothing for the instruction to hit. Those exits averaged 8.96 percent below entry against a trigger set at 8 percent, and the worst of them filled 17.55 percent below entry. The intended floor was breached by more than nine percentage points of notional on a single path, on a model with a modest one jump per quarter.

Averaged across every firing, the exit landed 0.2781 percent below the trigger, or about ₹5,005 illustrative. That is the honest headline number for what a stop's floor is worth: it holds to within a few basis points most of the time, and when it fails it fails in the direction and at the moment that hurts most. The median shortfall was 0.046 percent and the ninety-ninth percentile was 3.96 percent, which is a distribution with a very long right tail dressed up as a small average.

The lower panel of the figure is the part that should change how the trade-off is framed. As the share of daily variance arriving overnight is raised from 20 percent to 70 percent, with the total volatility of the series held exactly constant, the expected cost of the stop barely moves at all. It stays a few hundred rupees throughout, wandering with Monte Carlo noise. What moves is the tail: the stop's worst one percent goes from a 9.17 percent loss to a 16.35 percent loss, while the put's stays between 8.60 and 8.67 percent across the entire range.

So gap risk does not show up as a cost. It shows up as a failure of the thing you thought you had bought, and it is invisible to any comparison that stops at the average. This is also why the two instruments are not substitutes even though their expected costs match. One of them has a floor whose value is fixed by contract. The other has a floor whose value depends on how the market happens to be delivering its variance that year, which is not something a holder gets to choose or, in advance, to know.

The bill nobody adds up

There is a second asymmetry, and unlike the first it has nothing to do with market behaviour. It is written into the statute, and it runs the other way from what most people assume, because options carry a reputation for being expensive to trade.

The full exit stack on a delivery position of ₹18,00,000 came to ₹1,908.93 illustrative, of which the transaction tax alone is ₹1,800. Buying the put cost ₹28.09 illustrative all in. The transaction tax on one stop exit is about 64 times the entire cost of establishing the hedge. Expressed against the position, the stop's exit is 10.61 basis points and the put's purchase is 0.156.

The reason is structural rather than a quirk of rates. A stop protects you by selling the position, so the tax base is the position. A put protects you by buying a contract, so the tax base is the contract, and the contract costs a small fraction of the exposure it covers. If the stop is re-entered rather than left in cash, the round trip adds stamp duty on the way back in and comes to ₹2,269.86, or 12.61 basis points, every time the trigger is touched. Holding the same exposure through futures rather than delivery reduces but does not remove the asymmetry, since the transaction tax on a futures exit still falls on notional.

This matters most for exactly the traders who are most likely to prefer a stop, because a tighter trigger fires more often and every firing pays the stack again. It is a small number set against a large position, and it is also the one component of the whole comparison that is certain, known in advance, and completely unaffected by how the market behaves.

A year of protection, in one honest unit

Four cycles of the base case gives a year. The put, rolled quarterly at fair value, has an expected annual cost of about ₹112 illustrative, which is four charge stacks, against a total premium outlay of ₹43,391, or 2.41 percent of notional, that is expected to come back in claims. The stay-out stop has an expected annual cost of about ₹2,824 illustrative. Both figures carry a few hundred rupees of Monte Carlo error and neither is far from zero on a position of this size.

The re-entry variant deserves its own row, because leaving the position in cash for a quarter is not what most people do. Re-entering five days after each exit produced 0.327 round trips per cycle on average, and 2.93 percent of paths went through the cycle twice or more. Its expected annual cost was about ₹3,617 illustrative, most of which is charges. But the important number is not its cost, it is what it bought: its worst five percent averaged a 15.07 percent loss against the unprotected 15.72 percent. The re-entry stop removed almost none of the tail while paying the most for it.

The last row is the one that reframes the question. Nobody buys protection to raise their average outcome, so pricing these instruments by expected cost alone answers a question no holder is asking. Pricing them by what they charge per rupee of tail actually removed gives a ratio that is stable, computable, and much more informative, and on this model the put is the cheaper way to buy tail reduction unless you are paying a long way over fair value for it.

Where the answer changes

Everything above rests on two assumptions that a reader may reasonably reject: that the underlying has no drift, and that the put can be bought at fair value. Both are levers, and both are computable, so rather than defend them the run sweeps them.

Take the second lever first, because it is the one the put loses on. The fair value of the put is not a cost in expectation; only the amount paid above fair value is. At zero drift, the break-even overpayment against a stay-out stop was 1.3 percent, which is to say the two instruments are level and any meaningful premium over fair tips it to the stop. Real options are not usually available at flat-volatility fair value: the run itself found that the jumpy series makes the eight percent put 2.3 percent richer than a flat-volatility model would price it, equivalent to about 0.12 volatility points, and a real market skew is typically wider than that. Someone who consistently pays several volatility points over realised for downside protection is paying for it, and this model says so.

Now the first lever, which runs the other way and runs harder. A stop takes you out of the market and a put does not, so every point of drift you believe in is drift the stop forgoes while it sits in cash. At two percent assumed drift the break-even overpayment rises to 16.2 percent, at four percent it is 28.5 percent, at eight percent it is 54.6 percent, and at twelve percent it is 99.8 percent, at which point you could pay double the fair value of the put and still be no worse off than the stop. Against the re-entry stop, which is only out of the market for five days at a time, the same figures are much lower, rising from 6.8 percent to 26.0 percent across the same range.

Volatility, which is the lever people expect to matter, mostly does not. Across a sweep from 10 to 32 percent annualised the stop's hit rate climbed from 9.82 percent to 60.88 percent of paths and the put's fair value rose roughly thirty-fold, but the two costs tracked each other and the break-even overpayment stayed in single digits throughout. Higher volatility makes both instruments busier and the put dearer in absolute terms without changing much about which one is the better buy. The share of firings that gapped through the trigger held near a quarter at every volatility, and the share that recovered held near 47 percent, which suggests both are properties of the process rather than of its scale.

So the honest answer to which is cheaper is that it depends, and the dependency is specific. It depends almost entirely on what you believe the underlying earns while you are not holding it, and on how far above fair value you are paying for the contract. It depends much less than expected on volatility, and not at all on the gap channel, which changes the reliability of the stop without changing its price.

What you can actually buy in India

All of the above assumes the put exists. For most positions held by most people it does not, and this is where the comparison stops being interesting and starts being decided for you.

Options are listed on a limited set of instruments, and the minimum value of one contract has been held by regulation inside a band of roughly ₹15 to 20 lakh. The consequence is arithmetic rather than opinion: a holding worth less than the smallest available contract cannot be hedged with a matched put at all. Someone holding ₹3,00,000 of exposure has a choice between no put and a put covering five or six times their position, which is not a hedge but a separate directional bet stapled to the holding. The floor under a ₹3,00,000 position simply is not purchasable, whereas a stop on it costs nothing to place.

The lot sizes behind that band are revised periodically as index levels move, in both directions, so any specific contract size quoted on a page like this would be stale before long. The history is worth knowing as history: contract sizes were raised sharply in November 2024 when the minimum contract value band was lifted, and were revised back down again from January 2026 as index levels rose and pushed contract values above the top of the band. Our page on how lot sizes are set covers the mechanism. What matters here is the band, because the band is what determines whether your position can be matched.

Liquidity is the second filter, and it bites hardest exactly where a holder most wants protection. Index options trade actively. Single-stock options away from the largest names often do not, and a wide spread on a thinly traded contract can cost more on entry than the whole charge advantage computed earlier. A hedge that exists on a screen but cannot be bought near its value is not available in any sense that matters, and none of the arithmetic on this page survives a spread of several percent.

There is also a specific case where the put is unavailable at precisely the moment it would be worth most, which our page on circuit limits works through. A stock that can lock at a price band does not have derivatives listed on it, and a market-wide halt closes the derivative market in the same coordinated action that closes the cash market. The put fails in both cases, for opposite reasons. That is a genuine limitation of the instrument and it belongs in any honest account of it.

These constraints sit inside a wider context that is worth stating plainly on any page touching retail options: about 93% of individual traders in equity derivatives made net losses over FY22 to FY24, with aggregate net losses exceeding Rs 1.8 lakh crore (SEBI, September 2024). Nothing on this page is an argument for taking on derivative exposure. It is an account of how one particular defensive structure behaves compared with the alternative most people already use, and the availability constraints above mean the comparison is simply unavailable to a large share of the holdings it would apply to.

When each one is the better instrument

The computation supports some conclusions cleanly, refuses to support others, and the difference between the two is worth being explicit about.

The last row is the one worth carrying away. A great deal of writing on this subject asserts confidently that stops are expensive because of whipsaw, or that puts are expensive because the premium usually expires worthless. This run says both claims are close to empty on their own. The whipsaw is offset by the saves, and the expired premium is what a fair price for insurance looks like. Neither observation, on its own, prices anything.

What the comparison is actually for

It would be reasonable to reach the end of this and conclude that the two instruments are more or less equivalent, and for the average outcome on a driftless model that is very nearly true. It is also the wrong lesson, because the average outcome is not what a floor is for.

The useful reframing is that these two things are not competing versions of the same product. One is a contract that transfers a defined region of the distribution to somebody else in return for a price agreed in advance. The other is a plan to sell, which works as well as the market lets it work on the day, and which happens to be free to place, universally available, and unusable as a floor in exactly the conditions that produce the losses people most want floored. Knowing which of those you hold is worth more than knowing which is cheaper, because they fail in different places and only one of the two failures is bounded.

The habit worth taking from the exercise is smaller than the exercise. Any protection you place can be priced, and pricing it takes an afternoon rather than a career: state the floor, state what you pay for it, count how often it would have fired on your own history, and count how often the market opened straight through the level you were relying on. The answer will be specific to your instrument and your holding period, and it will not match anyone else's. What generalises is the discipline of computing it rather than assuming it, which is the substance of the method we teach.

Every figure comes from a single seeded simulation that reproduces identically on each run, written in Python with numpy only. Each of 200,000 paths is built day by day in two pieces: an overnight move carrying 40 percent of the day's variance plus a compound Poisson jump component, then a continuous session from the open to the close whose low is drawn from the exact minimum law of a Brownian bridge rather than from a grid of sub-steps. An earlier version of the model did use a grid, and it silently handed the stop about 12.5 basis points per exit that it had not earned, because a price sampled at intervals can slip past a trigger between samples. The exact sampler removes that bias and a control asserts it is gone.

The jump component is sized so that switching it on does not change the volatility of the series, so the gap sweeps redistribute variance between channels rather than adding any. The put is valued at the average of its payoff across all paths, which is its fair value at a zero interest rate, and a flat-volatility Black Scholes price is reported alongside for reference. Cost is defined as the difference in mean outcome against the unprotected regime on the identical paths, so the Monte Carlo error largely cancels, and the residual standard error is reported wherever a cost is quoted. Charges use statutory rates read from primary sources dated 2026, with brokerage and the depository fee stated as illustrative because they are commercial rather than statutory. Ten controls run before any result is produced, including a negative control that removes the gap channel and a matching positive control that amplifies it, so that a clean number means something.

All results are illustrative and simulated. They are not a track record, not a forecast, and not an indication of what any instrument would produce in a live account. The purpose is to demonstrate the structural difference between a contractual floor and an instruction to trade, which is a property of the two mechanisms rather than of any particular market. The model asserts no drift and no view on any index; where drift matters, the sensitivity is computed and shown instead.

• Hedging, and what each hedge really costs →
• How a stop order actually behaves →
• Gaps, and the band that never traded →
• Slippage and the order book →
• How lot sizes are set →
• What implied volatility is pricing →