The Iron Condor: A Structure That Wins Most Months and Can Still Lose Overall

Why the win rate is the most persuasive number in options

Every trading statistic has a speed. The share of positions that finished in profit is the fastest of them: it is available after the first expiry, it updates every month, and it needs no arithmetic to interpret. Expectancy is the slowest. It is an average over a distribution whose left tail is rare by design, so a sample large enough to measure it honestly takes years to accumulate. When the fast number and the slow number disagree, the fast one is the one you spend the intervening years believing.

The iron condor sharpens this to a point, because it is the rare structure whose most likely outcome is also its best outcome. A directional position that works has a spectrum of good results and a spectrum of bad ones. A condor held to expiry has essentially two: the underlying stays inside the sold strikes and you keep the entire credit, or it does not and you give back a multiple of it. In the worked structure below, 78.4 percent of simulated expiries paid the full credit and 14.9 percent cost the full maximum. There is very little in between. That two-point shape is exactly what makes the position feel less like a bet and more like a process with an occasional interruption.

The arithmetic that punctures this is not complicated, and it is worth writing out because everything else on the page is a consequence of it. Expectancy is the win rate multiplied by the average win, less the loss rate multiplied by the average loss. Set that to zero and rearrange, and the average loss at which a strategy exactly breaks even is the win rate divided by the loss rate. At three wins in five that ratio is 1.5. At four in five it is 4.0. At nine in ten it is 9.0. At nineteen in twenty it is 19.0. The win rate does not tell you whether a structure makes money. It tells you the loss size at which the question becomes interesting, and nothing more.

Our guide to measuring whether an edge exists makes this argument in general terms and it applies to any strategy. What is specific to the condor, and what this page is for, is that the structure is engineered to raise the win rate, and every mechanism that raises it raises the required loss-to-win ratio in lockstep. You do not get to move one without the other. Selling further from the market buys a higher share of profitable expiries and pays for it with a smaller credit against an unchanged wing, which is the same trade written in different notation. That is not a flaw in how condors are usually taught. It is what a fairly priced option market means.

The structure, computed rather than described

Everything below comes from one seeded model, and the model is stated in full so it can be checked or disputed. The underlying is a broad index at an illustrative level of 24,000. Implied volatility is 14 percent annualised, the horizon is one month of 21 sessions, and the risk-free rate and dividend yield are both set to zero so the expected move is centred on the current level. One standard deviation of the monthly move under those assumptions is 970 index points.

Every leg is priced with the Black-Scholes model at that same 14 percent, and the simulation that follows then draws the underlying from that same distribution. This is the single most important choice on the page and it is deliberate. Pricing a structure at the volatility it goes on to experience makes it actuarially fair by construction, which means its expectancy before costs is not approximately zero, it is exactly zero. That removes the argument. Whatever the win rate turns out to be, we already know what the structure is worth, so the two numbers can be examined separately without one contaminating the other.

On lot size, a note is needed. The minimum contract value for index derivatives was moved into a band of 15 to 20 lakh rupees, and exchanges revise lot sizes periodically to keep contract value inside that band as index levels move. Lots were raised across the board in November 2024 and revised down again from January 2026 as indices rose. Quoting any specific lot as a present fact therefore goes stale quickly, and our note on lot sizes in futures and options covers the mechanics. The arithmetic here derives the lot from the band instead: at an index level of 24,000 the band implies between 63 and 83 units, and the worked example uses 65 units, giving a contract value of 15.6 lakh rupees. That number is illustrative and is used only so the rupee figures have something concrete behind them.

Two features of that diagram deserve to be said out loud, because a payoff picture is very good at showing shape and very bad at showing likelihood. The first is the asymmetry: the flat top is worth 2,347 rupees and the flat bottom costs 10,653, so the structure risks 4.54 units to make one. The second is that the profitable zone looks enormous. A band of more than ten percent of the index level, over a single month, is a wide target, and that width is the entire source of the high share of profitable expiries. The two observations are the same observation. You are being paid a small amount to accept a large loss in a case that is unlikely, and the market has priced the unlikelihood.

If you want to change the strikes, the wing widths or the premiums and watch the shape move, our options payoff calculator plots any combination of legs and solves the breakevens numerically. This page is not about the shape. It is about what the shape hides.

The win rate and the expectancy are almost strangers

Simulate 400,000 expiries of that structure against the distribution it was priced from and two numbers come back. The share finishing in profit is 79.71 percent, against an exact analytic value of 79.78 percent. The mean result is minus 10.91 rupees per lot, with a Monte Carlo standard error of 7.62 rupees, which is about one and a half standard errors from a true value that we already know is exactly zero. The simulation is confirming the arithmetic rather than discovering anything.

One further number from that run is worth pausing on. The standard deviation of a single expiry is 4,819 rupees, which is 2.05 times the maximum the structure can make. The dispersion of one month's outcome is twice the size of the best possible month. Any process where that is true will take a very long time to reveal its average, and in the meantime the only thing on display is the win rate.

Now the part that makes the two numbers separable. Across those 400,000 expiries the average win was 2,328 rupees and the average loss was 9,199 rupees, a loss-to-win ratio of 3.952. The breakeven ratio implied by a 79.78 percent win rate is 3.947. They agree to three decimal places, and that is not a coincidence or a tuned result. It is a definition. A fairly priced structure has zero expectancy, zero expectancy means the loss-to-win ratio equals the win rate divided by the loss rate, and so the market hands you a payoff ratio that exactly cancels whatever win rate you selected. If you want the higher win rate, the price of it is already set.

The figure above is the argument in one picture, and it is worth reading slowly. The gold line is five iron condors on the same index at the same volatility, differing only in how far from the market the sold strikes sit. The share of profitable expiries runs from 58.0 percent at the nearest strikes to 95.1 percent at the furthest. That is a 37 point range, which is about as wide as this design parameter can move. The expectancy across all five, before any charge, is identically zero, and after the charge stack the whole line sits between minus 232 and minus 192 rupees. Thirty-seven points of win rate bought 41 rupees of movement in the thing that actually matters.

The coral line does the reverse. It holds the share of profitable expiries roughly constant near 80 percent by adjusting the strikes for each volatility regime, and lets the realised volatility move instead. At 10.5 percent realised against 14 percent implied the structure is worth plus 1,507 rupees per lot after charges. At 18.5 percent realised it is worth minus 1,555. Two points of win rate, 3,062 rupees of expectancy. A horizontal line and a vertical line, which is what statistical independence looks like when you draw it.

Read the two ratio columns against each other and the mechanism is visible. In block A they are equal at every row, because every row is fairly priced. In block B and block C they separate, and the sign of the separation is the sign of the expectancy. When the delivered loss-to-win ratio comes in below the breakeven ratio the structure was worth holding, and when it comes in above it the structure was not. In the unquiet block C row the breakeven ratio was 3.84 and the delivered ratio was 10.18. That row still finished in profit on 79.33 percent of expiries.

So what does decide expectancy? The gap between the volatility priced into the premium and the volatility that actually turns up, less the cost of getting in and out. That gap is a real phenomenon and it is the honest case for the structure: option premiums have historically embedded a margin over subsequently realised volatility, which is why a position that is short volatility is not obviously a losing proposition. But it is a position on that specific gap. Our note on implied volatility covers how the priced number is derived and why it is not a forecast. Nothing about the share of profitable expiries tells you which side of that gap the next month falls on, which is precisely what the coral line in the figure demonstrates.

A long run of good months is the default, not the evidence

Take the base structure, deduct the charge stack computed later on this page, and its per expiry profitable share becomes 79.67 percent with an expectancy of minus 205.58 rupees per lot. Run 200,000 independent five-year paths of 60 monthly expiries each and watch what a negative expectancy actually looks like from the inside.

The first losing expiry arrives, on average, in the fifth month. The longest run of consecutive profitable expiries within a five-year path averages 13.1 and has a median of 12. Fifty-seven percent of paths contain a run of twelve or more, which is a full calendar year in which nothing went wrong. Thirty-one percent contain a run of fifteen or more. Nine percent contain a run of twenty or more, which is nearly two years. The standard extreme-value approximation for the longest run of successes gives 13.04 against the simulated 13.11, so this is not an artefact of the simulation. It is what runs of a repeated event with a high success probability do.

The lower panel of that figure is the part that changes how the upper panel should be read. After a single expiry, 79.8 percent of the simulated accounts are showing a profit. After three expiries, 55.0 percent. After six, 36.5 percent. After twelve, 41.0 percent, and after sixty, 38.3 percent. The mean five-year outcome is a loss of 12,335 rupees per lot, the median is a loss of 11,011, and the worst one in twenty is a loss of 75,912.

Notice that the decline is not smooth, and the honest explanation is more interesting than a smooth curve would have been. Whether an account is ahead depends on how many maximum losses the accumulated credits can absorb, and that is a step rather than a slope. At the six expiry mark, five clean months net 10,711 rupees and a single maximum loss costs 10,858 all in, so one bad month is already enough to put the account behind and the checkpoint lands in an unlucky part of the arithmetic. By twelve expiries the accumulated credits can absorb one maximum loss and the share ahead recovers to 41.0 percent. Publishing the wobble rather than smoothing it is the point: the shape of the thing is lumpy, and a reader looking at a tidy monotone curve would be looking at a drawing rather than a computation.

What all of this describes is a feedback loop pointing the wrong way. The evidence that arrives monthly is the win rate, and the win rate is uninformative. The evidence that would settle the question is the expectancy, and it needs a sample long enough that by the time it is legible the position has already been sized off the wrong number, usually upward, because a year of clean months is the most natural argument in the world for doing more of the same. Our guide to risk of ruin works through the general streak arithmetic and why a run of any length is weaker evidence than it feels. The condor simply supplies an unusually long one.

This is also the point at which the base rate belongs. SEBI's study of individual traders found that about 93% of individual traders in equity derivatives made net losses over FY22 to FY24, with aggregate net losses exceeding ₹1.8 lakh crore (SEBI, September 2024). That finding is not a statement about any particular structure, and it would be dishonest to present it as one. It is a statement about what a population of participants produced across a segment where the visible feedback is fast, the decisive feedback is slow, and the gap between the two is exactly where the arithmetic on this page lives.

Eight chargeable events against one net credit

Cost models for options usually describe a round trip on one instrument. A condor is not one instrument. It is four legs to open and, if it is closed rather than left to settle, four more to close, and every statutory levy in the stack is computed on the leg it applies to rather than on the credit the four legs jointly produce. That distinction is the whole of this section.

The rates used are the verified ones. Securities transaction tax on the sale of an option is 0.15 percent of the premium and falls on the seller, a rate set from 1 April 2026 by section 159 of the Finance Act 2026 amending serial 4 of the section 98 table. Securities transaction tax on an exercised option is 0.15 percent of intrinsic value and falls on the purchaser. Stamp duty on options is 0.003 percent on the buy side under Article 56A of Schedule I to the Indian Stamp Act 1899. Goods and services tax at 18 percent applies to brokerage and to exchange transaction charges, and not to securities transaction tax or stamp duty. For the exchange transaction charge, the verified figure available is 3,250 rupees per crore of premium turnover on index options, set by an Indian exchange notice of 27 September 2024, effective 1 October 2024, and that is the rate used here. Brokerage is a commercial charge rather than a statutory one, and the model assumes an illustrative flat 20 rupees per executed order.

Applying that stack to the four legs produces the first result. The four opening premiums sum to 165.61 points, which at the illustrative lot is 10,765 rupees of premium turnover, against a net credit of 2,347 rupees. The levies are computed on the larger number. Turnover is 4.59 times the credit, because the two bought legs cost money and the two sold legs bring it in, and the charges do not net off the way the premium does.

The totals are smaller than that multiple suggests, and the reason is instructive. Opening the position costs 108.49 rupees, of which securities transaction tax is 9.83, the exchange charge 3.50, stamp duty 0.13, brokerage 80.00 and the tax on brokerage and the exchange charge 15.03. Closing it costs a further 96.53. A full eight order round trip is therefore 205.02 rupees against a 2,347 rupee credit, which is 8.73 percent of the credit. Left to settle rather than closed, only the opening stack is paid and the figure is 108.49, or 4.62 percent.

Ninety-two percent of that round trip is brokerage and the tax on it. The statutory levies, the part that scales with the size of the premium, come to under 16 rupees. The per order charge, the part that scales with the number of legs, comes to 160. This is the arithmetic that decides whether a multi-leg structure is viable at retail size, and it does not appear in any single-leg cost model. At an illustrative 10 rupees per order the round trip falls to 110.62 and at 40 rupees it rises to 393.82, so the entire cost profile of the position is set by a number that has nothing to do with options.

What that does to the breakevens is almost nothing and what it does to the year is not. Deducting the full round trip moves the breakevens from 22,764 and 25,236 to 22,767 and 25,233, narrowing the profitable zone by 6.31 points out of 2,472. The share of profitable expiries barely registers the change. But twelve expiries of the same stack come to 2,460 rupees, which is 1.05 times the entire maximum profit of a single month. Against an expectancy that was exactly zero before charges, that is the whole of the expected annual result: a loss of slightly more than one perfect month.

There is a further effect that runs the other way from intuition. As the sold strikes move further out the credit collapses much faster than the charge stack does, because most of the stack is per order. In the five structures of the earlier table the round trip cost ranges only from 232 to 192 rupees, while the credit falls from 5,332 to 514. The charge stack takes 4.4 percent of the credit on the nearest structure and 37.3 percent on the furthest. The version of the structure that produces the highest share of profitable expiries is also the version where costs consume the largest fraction of what it produces.

One honest gap. The SEBI turnover fee is verified for the cash segment at 10 rupees per crore under regulation 41 of the SEBI (Stock Brokers) Regulations, 2026, but the corresponding row for the derivatives segment could not be verified from a primary source, so it is omitted from this model entirely. Every cost figure above is therefore an understatement rather than an overstatement, which is the direction an unverified omission should err in.

What happens at settlement, and the charge nobody expects

Index options in India are European in style and cash settled, so a condor built on an index never produces a delivery obligation. Whatever happens, the four legs resolve into a cash difference. This is not true of single-stock derivatives, which are physically settled and therefore carry an obligation sized to the full contract value rather than to the premium. That is a materially different exposure and it is already treated in depth in our guide to options selling and risk management in India, which sets out the settlement framework and the writer's position under it. There is no reason to rebuild it here, and a condor on a single stock inherits every one of those mechanics on both of its sold legs.

On the index side, the resolution is simple in the good case. If the index settles between 22,800 and 25,200 all four legs expire worthless, nothing is exercised, no exit orders are placed, and the result is the credit less the opening stack of 108.49 rupees. That case is 78.4 percent of expiries in the model, which is why the round trip figure of 205.02 overstates the typical cost and the figure of 108.49 understates the cost of the expiries that actually go wrong.

The uncomfortable case has a detail in it that is easy to miss. Securities transaction tax on an exercised option is charged on intrinsic value and is payable by the purchaser, not the writer. On your two sold legs, an exercise charge falls on whoever is long them. On your two bought legs, the wings, you are the purchaser. The leg whose entire function is to cap your loss is therefore the one that bills you at settlement when it finishes in the money, and the bill grows with the size of the move even though the payoff has stopped moving. At the maximum loss point the bought put is exactly 200 points in the money and the charge is 19.50 rupees per lot. If the index settles at 22,000 the intrinsic is 600 points and the charge is 58.50. At 21,000 it is 1,600 points and 156.00 rupees. The payoff diagram is flat across all of those levels. The charge is not.

The other thing that changes past the sold strike is margin. A condor's two sold legs require initial margin that expands as the underlying moves against them, and the expansion arrives at exactly the moment the position is least comfortable to hold. The wings reduce the requirement relative to naked short options, which is a genuine benefit and one of the real reasons defined-risk structures exist, but they do not remove the dynamic. The margin-expansion mechanism, including how a requirement can grow faster than the position loses, is worked through in the same guide linked above.

Rolling a tested condor: help, or a deferral with a fee attached

When the underlying approaches a sold strike, the standard taught response is to adjust rather than accept the outcome: close the tested side and reopen it further away and further out in time, often for an additional credit, so that no loss is recorded today. The claim made for this is that it converts a losing position into a live one. Whether it does is a question a model can answer, so here is one.

The setup runs 200,000 paths of 42 sessions each, with the underlying drawn from the same distribution used throughout. Both arms carry two months of exposure and both pay their charges, so the comparison is like for like. The arm that does not adjust holds the original condor to settlement and then opens a fresh one at whatever level prevails, holding that to settlement too. The arm that adjusts closes all four legs at their marks on the first session on which the index touches either sold strike, pays a full exit stack, immediately opens a new condor centred on the level at that moment with a full month to run and the same strike distance, pays a full entry stack, and carries that to settlement. Paths where no sold strike is ever touched behave identically in both arms.

A sold strike was touched on 35.0 percent of paths. The first useful number arrives before any decision is taken. At the moment of the touch, the four legs mark at 103.3 points against the 36.1 points that were collected, so the position is already showing an average loss of 4,574 rupees per lot. This is worth absorbing on its own terms. Touching a sold strike is not a warning that a loss might be coming. It is a loss of roughly twice the credit, already incurred, whatever is done next.

On the tested paths the two arms produce genuinely different distributions and identical averages. Leaving the position alone returned an average of minus 4,608 rupees over the two months and rolling returned minus 4,727, a difference of 119 rupees on the tested subset and, across all 200,000 paths, minus 41.69 rupees with a standard error of 10.65. That difference is the one extra charge stack. Before charges the two arms have the same expectancy by construction, because a replacement condor struck at the same distance from a new level is priced on exactly the same terms as the one it replaced. Rolling does not add value. It cannot, unless the volatility you are selling into is priced differently from the volatility you sold before, and nothing in the act of rolling makes that true.

What rolling did change was everything except the average. The share of tested paths finishing the two months in profit fell from 36.7 percent to 2.0 percent, because closing the position converts a paper loss of 4,574 into a booked one, and a fresh condor whose entire upside is 2,239 rupees cannot repay it. The middle half of outcomes compressed hard, from a range of minus 8,700 to plus 4,380 down to minus 4,716 to minus 1,693. And the left tail genuinely improved: the worst one in twenty moved from minus 20,836 to minus 15,764, because exiting at the touch caps the damage below the full maximum loss the unadjusted position could still run to.

That is the honest answer, and it is neither of the two stories usually told. Rolling is not the repair it is presented as, because it leaves the average untouched and adds a charge stack. It is also not the pure error the sceptical account makes it out to be, because truncating the left tail is a real thing to have done. Across the tested paths rolling produced a better result than leaving the position alone 48.0 percent of the time, which is as close to a coin flip as a decision gets. What it does with certainty is convert a resolved position into an unresolved one and extend the time at risk.

The recovery arithmetic is where the escalation usually begins. A booked loss of 4,574 rupees against a best possible clean month of 2,239 needs three consecutive perfect expiries to erase, and the probability of three consecutive full-credit expiries in this model is 48.3 percent. The alternative route back is size: repaying it in a single cycle requires 2.04 times the original position. Both of those are worth stating plainly, because the roll is rarely the last decision in the sequence. It is usually the first one.

The model has limits and they should be named. It rolls once, at the same distance, at the same size, and assumes volatility is priced identically at the roll. Real adjustments often go out in time as well as in strike, sometimes go up in size, and are frequently made when implied volatility has risen, which changes the premium available. Rolling into higher implied volatility collects more, and if that volatility does not materialise the roll is genuinely better than this model says. Rolling at larger size is worse than this model says, in exactly the proportion of the size increase.

What the structure is, and what it is not

An iron condor is a well-behaved instrument. Its risk is bounded, its arithmetic is transparent, its worst case is a number you can write down before you open it, and none of that is in dispute. Our framework for thinking about options positions treats defining the maximum loss as a gate that has to be passed before anything else is considered, and the condor passes it cleanly. What this page has tried to establish is a narrower and more specific claim: that the number the structure is most often recommended on is the one number that carries no information about whether it is worth holding.

The uncomfortable conclusion is the one the computation actually produced, so it is the one that gets published. A structure priced at the volatility it then meets is worth exactly nothing before costs and slightly less than nothing after them, and it will show a profit in four expiries out of five while that is true. Every path in the simulation was generated by a model with a negative expectancy, and more than half of those paths contained a full year without a single losing month. Neither of those facts is a criticism of the instrument. They are both properties of arithmetic that would apply to any structure with the same shape.

What follows from that is not a rule about what to hold. It is a rule about what to measure. If the only number you are tracking is the share of expiries that finished in profit, you have chosen the one statistic that a fairly priced short-volatility structure is guaranteed to make look good, and you will keep collecting it for as long as it takes the slower number to arrive. Track the loss-to-win ratio against the ratio your win rate requires, because those two cross at a computable point and the crossing is the whole question. Track the charge stack per leg rather than per position. Track the gap between what volatility was priced at and what it turned out to be, because on the evidence of the table above that gap is doing all of the work. Those are habits rather than techniques, and they are the substance of the method we teach.

Every figure comes from a single deterministic simulation, seeded so it reproduces identically on each run. Each of the four legs is priced with the Black-Scholes model at 14 percent annualised volatility over one month of 21 sessions, with the risk-free rate and dividend yield both set to zero. The terminal distribution the structure is then simulated against is the same lognormal the prices were derived from, which makes the expectancy before costs exactly zero as a matter of construction rather than as a result. Win rates and expectancies are computed analytically from the lognormal distribution function and cross-checked against Monte Carlo runs of 400,000 expiries for the base structure, 200,000 for each scenario, 200,000 five-year paths for the streak arithmetic and 200,000 paired 42 session paths for the adjustment model. Statutory charge rates are taken from the primary instruments cited in the charge section; the derivatives-segment SEBI turnover fee could not be verified from a primary source and is omitted, so cost figures understate rather than overstate. Brokerage is a commercial charge and the 20 rupees per order used here is an illustrative assumption, not a rate anyone is obliged to charge.

All results are illustrative and simulated. They are not a track record, they are not a forecast, and they are not an indication of what any structure would produce in a live account. The purpose of the exercise is to demonstrate the relationship between the share of profitable expiries and the expectancy, which is a property of the arithmetic rather than of any particular market or instrument. Nothing on this page is a recommendation to enter, hold or avoid any position.

• A framework for options positions, gate by gate →
• Plot any combination of legs →
• Measuring whether an edge exists →
• Risk of ruin and the length of a normal losing streak →