Options, from the first contract to the backtest

Where you are starting from

Owning the index, and what that already commits you to.

Say you own an S&P 500 index fund. You already hold a position with a known behaviour: it rises over decades, and it falls hard several times per lifetime. You have accepted that trade, and the reason it works is that you do not sell during the falls.

Everything below adds one small position on top of that. It does not replace the index fund and it is not a hedge against it. Keeping that clear is most of understanding it.

What an option actually is

A contract about a price and a date, and nothing more mysterious.

An option is an agreement between two people about a future price. One side pays for the right to do something; the other side is paid to be obliged to do it if asked.

A put is insurance on a price. Somebody who owns shares at $100 and is frightened of a fall can buy a put with a $95 strike. If the market collapses to $70, they still get to sell at $95. They paid for that, and if the market does not collapse they simply lose what they paid, the way you lose a house insurance premium in a year the house does not burn down.

The multiplier is where beginners lose the most money fastest. An option quoted at 1.20 costs 120 dollars, because one contract covers 100 shares. Ten contracts on a $90 fund is not a small position, it is $90,000 of exposure.

Buying one: the long call

A call is the right to BUY at the strike. Buy one and you have paid for that right and taken on no obligation at all. If the shares finish above the strike the contract is worth the difference; if they do not, it expires and the premium is gone.

The shape is the opposite of everything else on this page. Your loss is capped at what you paid and your gain is not capped at all. That sounds like the better side of the trade, and it is the reason most people start here and most people stop.

What makes it hard is that you need to be right twice. The direction has to go your way AND it has to happen before the contract expires. A share that rises the week after your call expires pays you nothing. Every day that passes takes value out of the contract even when the price does not move, which is the same decay that works FOR a seller working against you here.

A long call is also how the second leg of a spread is bought, so it is worth understanding even if you never open one on its own. It appears again under vertical spreads.

Selling instead of buying

The other side of that trade, and what it obliges you to.

Somebody has to sell that insurance. When you sell a put, you take the premium up front, and in exchange you are obliged to buy the shares at the strike if the buyer wants you to.

Your outcomes are not symmetrical, and this is the single most important thing on this page:

This is why the honest description of selling puts is selling insurance, not generating income. An insurer collects small premiums from many people and pays out rarely and large. That is a real business, and it is also a business that goes bankrupt when it writes more cover than it can fund.

The mirror trade: covered calls

Selling the right to buy your shares, and what it does and does not do for you.

You will meet this one everywhere, so it is worth being able to place it. A covered call is the same move pointed the other way. You already own the shares, and you sell somebody the right to buy them from you at a strike. The premium is yours straight away. If the price finishes above the strike, the shares go, at that strike.

Covered means you already hold what you might have to hand over. Selling a call without owning the shares is a different trade and an unbounded one, because there is no ceiling on how far a price can climb. Everything below assumes you own them.

That last row is the one the word income hides. Nothing about selling a call reduces how far your shares can fall. What it does is give away the top of the rise, and the top of the rise is where a long run of index returns actually comes from.

So the two trades are not opposites in the way the names suggest. Selling a put takes on a fall you did not previously own, in exchange for a premium. Selling a covered call keeps the fall you already had and sells off the recovery. One adds a risk and is paid for it. The other reshapes a risk you were already carrying.

The bargain, stated plainly: a slightly shallower worst case, in exchange for a lower return over time. That is a real trade and it is not obviously a bad one, but it is not the trade the marketing describes. The measured version, run on the same three decades as everything else here with the drawdown printed beside every return, is on the research page.

And being assigned means something different here. When a short put is assigned you buy shares with cash. When a covered call is assigned your shares leave you. If they carried a gain, that gain becomes real in that tax year whether or not the timing suited you, and it is a question about your own cost basis rather than about the option.

That is the reason many people prefer selling puts on shares they do not yet own to selling calls on shares they have held for years.

Putting both together: the wheel

The wheel is not a new instrument. It is the two trades you have just read about, run one after the other on the same shares.

Sell a put on something you would be content to own. If it expires worthless you keep the premium and sell another. If it is assigned you now own 100 shares at the strike, and you start selling covered calls against them. If a call is assigned the shares go at that strike and you are back to cash, selling puts again.

Assignment is not the failure case. That is the part people find surprising. Each leg has a defined next step, and the trade only goes badly wrong in the way a short put goes badly wrong: you are holding shares that have fallen a long way, and the premium collected is small against the fall.

Its real difficulty is bookkeeping, not strategy. One wheel across a year is a put, an assignment, several calls, perhaps a roll or two, and a final assignment. Your broker reports each of those as an unrelated event, and the cost basis of the shares moves every time you collect premium against them. Getting that arithmetic wrong makes every figure after it wrong, which is most of why this tracker exists.

Restrike folds those legs into one position and adjusts the basis as the premium lands, so a wheel reads as one thing rather than eleven. It does not tell you which strike to sell.

Assignment, settlement and expiry

What happens at the end, and the one surprise worth understanding.

If your put expires with the market above the strike, it is worthless to the buyer, nothing happens, and you keep the premium. If it expires below the strike, it settles, and how it settles depends on the instrument.

• Physically settled options on an ETF deliver shares. You wake up owning 100 shares per contract, bought at the strike, which is above where the market now is.
• Cash settled index options pay the difference in cash. No shares appear. Nothing about your holdings changes except the balance.

Early assignment

American-style options can be exercised before expiry, which surprises people. It is worth knowing what actually triggers it, because the common belief is wrong.

A put holder exercises early when the contract is deep in the money and has almost no extrinsic value left: at that point they give up nothing by acting now. An upcoming dividend makes early assignment of a put less likely, not more, because exercising forfeits the dividend. That is the opposite of the rule for calls, and it is the detail most often stated backwards.

A call holder exercises early to collect a dividend, and only when two things are true at once: the call is in the money, and the time value left on it by the eve of the ex-date is less than the dividend. They give up that time value to become the holder of record, so they do it when the dividend is worth more.

Both conditions are required. An out-of-the-money call is never exercised however little time value is left on it, because exercising buys the shares above the market price. Leaving that test out is the single most common way the rule is got wrong.

Why there is a premium at all

Implied volatility, and the gap that has historically existed.

An option's price is mostly a statement about how much the market expects things to move before expiry. More expected movement means a more valuable right, so a higher premium. That expectation, backed out of the price, is called implied volatility.

Implied volatility is a forecast. What actually happened afterwards is realised volatility. The two are rarely equal, and historically the forecast has tended to be the higher of the two: people pay a little more for protection than the protection turned out to be worth.

That gap is the entire reason a seller has an edge, and it exists for an unglamorous reason. Being wrong about a crash is not survivable for most people, so they pay to not find out. You are being paid to be the person who can afford to find out.

The edge is real and it is small. In the one strategy studied here across thirty years, a monthly at-the-money put on an index, it was worth one to two points of annual return and came with deeper drawdowns. Other strategies have not been measured that way, here or anywhere most readers can check.

Anyone describing this as a way to make a lot of money quickly is describing a different, larger position than the one that produced those figures. The numbers are here.

The Greeks, one at a time

Five sensitivities. Each answers one question about what moves your position.

The Greeks sound like a barrier and are not. Each one answers "if this changes by a bit, what happens to the option's value?" That is all they are. The signs below are for a short put, one you have sold, because it is the position where every sign is the awkward way round. Flip each of them for a long option.

The combination is the point

Read those together and the position has a shape. Negative gamma and negative vega arrive at the same moment. A sharp fall makes your delta worse exactly while fear is raising implied volatility, so the contract gets more expensive for two reasons at once, and both of them are the same event.

That is why the losing months are large rather than frequent. In the thirty-year study on this site the worst single cycle lost 20.3 times the premium it collected. Positive theta pays you a little every day. Negative gamma and vega take it back in a fortnight, occasionally.

Two legs: vertical spreads

A vertical spread is one option bought and another sold, same underlying, same expiry, different strikes. That second leg exists to put a wall at the far end of the trade.

A credit spread takes in more than it pays out. Sell a put at 95, buy one at 90, and you collect the difference in premium. The bought leg caps what the position can lose: below 90 the two move together and the damage stops. You have swapped some of a short put's premium for a floor under it.

A debit spread costs money to open. Buy a call at 100, sell one at 110, and you have paid for the difference. The sold leg caps the gain at 110 and pays for part of the call. You have swapped some of a long call's upside for a lower cost of entry.

Both are defined risk, and that is the whole point. The most you can lose is known when you open it, which is not true of a naked short put. The price of that is a capped best case, and a second commission and spread on every trade.

Two things go wrong in practice. The legs are quoted separately and filling them one at a time can leave you briefly holding just the naked one. And close to expiry a spread that finishes between the strikes assigns on one leg and not the other, which leaves you holding shares over a weekend you did not plan to.

Straddles and strangles

Everything above has a view on direction. These do not: they are positions on how far a thing moves, in either direction.

A straddle is a call and a put at the same strike and expiry. A strangle is the same idea with the two strikes apart, which is cheaper to buy and collects less to sell.

Bought, either one profits if the move is bigger than the market charged you for it. You lose if the price sits still, and you lose to time decay on both legs at once, which is why "nothing happened" is the expensive outcome rather than the neutral one.

Sold, the trade is the mirror and the risk is the thing to understand before anything else. A short strangle collects two premiums and is exposed on both sides at the same time. On the call side that exposure has no upper bound at all, unless you hold the shares. This is the structure that most often turns a long quiet run of small wins into a single loss larger than all of them.

The premium is not the point of them either. Implied volatility is what the market charges for the move; realised volatility is what actually happens. Selling one of these is a bet that the first is above the second, and there are long stretches where it is not.

Rolling, and what it really is

Rolling means closing a contract that is running out of time and opening another one further out, usually in a single order. It is the most common adjustment there is and the most commonly misunderstood.

It is two trades, not a repair. The old contract closes at whatever it is worth, which realises the loss or the gain on it, and a new one opens at today's price. A roll that takes in more than it costs is described as a credit, and that credit is real; what it does not do is undo what the first contract cost you.

Rolling to avoid assignment is a decision, not a default. A short put deep in the money can be rolled almost indefinitely, and each roll collects a little and keeps the exposure. Whether that is patience or refusing to take a loss is a question the arithmetic cannot answer for you, and it is worth deciding in advance rather than on the day.

Before tax, rolling and being assigned are close to the same trade. After tax they are not: an assignment realises a gain or a loss on shares in that tax year, and a roll does not.

Restrike keeps a rolled position as ONE position, with every contract it ran through underneath it, because a roll that reads as two unrelated trades makes the total meaningless. A great deal of the work in this tracker is that one idea.

Margin, and what actually goes wrong

Not the loss. The forced exit.

Selling a put obliges you to buy shares, so your broker requires collateral against that obligation: margin. It is not a fee. It is capital that stops being yours to use, and the requirement grows as the market falls, because the obligation is getting larger.

The requirement is not only on the option. Your broker also charges maintenance margin on the shares you already own, typically 25%, and that book is usually the larger position. An early version of the research behind this site charged margin on the put alone and treated the shares as free. Correcting that moved the survivable range down by a third.

Why sizing is the whole game

The one decision that separates the outcomes.

Nothing above is a strategy yet. The strategy is the size.

Selling one contract against a large account is nearly pointless and nearly harmless. Selling twenty is a different instrument entirely, with the same name. The same trade, at two sizes, is a modest edge or an account-ending event, and no amount of cleverness about which day to sell changes that.

The question that matters is not "will this make money", it is "how far can the market fall before somebody else closes this for me". That number is knowable in advance and it depends almost entirely on size, which is what makes it the one thing worth working out before you open anything.

For reference, the worst 30-calendar-day fall in the S&P 500 since 1993 was −32.8%, ending in March 2020. A size that does not survive that is a size that has already failed once in living memory.

How to approach a backtest

Sooner or later somebody shows you a strategy with a curve attached. Here is how to read one, written by somebody whose own backtest has been wrong more than once and had to be corrected in public.

Ask what it is being compared against

A return with no benchmark next to it means nothing. "Made 11% a year" is only interesting against what the same money would have done sitting in the index, over the same window, through the same crashes. And the comparison has to be like for like: a strategy that is only invested half the time is not comparable to one that is always on.

The trap is subtler than a missing benchmark. Any rule that trades LESS will beat an always-on rule at something, with no skill involved, because doing nothing is free. So a conditional rule has to be scored against its own frequency, not against the unconditional version.

One way is to take the same pattern of active and inactive periods and slide it around the timeline. If the real alignment is not better than most of the shifted ones, the rule found nothing.

Ask what the costs were

Most published results are gross. Commissions, the spread between bid and ask, and slippage all come out of the same premium the strategy is built on, and a strategy trading weekly pays them fifty times a year. A figure measured at zero cost is an upper bound, and it should say so.

Ask what the cost would have to be to erase the edge. That single number tells you more than the headline does, because it says how much room there is between the result and nothing.

Ask what the drawdown was, and on what basis

Never accept a return without the worst loss beside it. And check how the loss was measured: a drawdown sampled once a month steps straight over the bottom of a crash and can understate the real one by several points. Two figures measured on different bases must never be subtracted from each other, which is a mistake that has been published on this site and had to be corrected.

Ask whether it survives being cut in half

Split the sample and run the two halves separately. A rule that works in one half and not the other has usually found something about that period rather than about the market. This is the test that kills most timing rules, and it is cheap to run.

Watch for parameters that barely move. A signal that crosses its own threshold twice in thirty years is not a strategy with a rule, it is a strategy with a date, and it will look wonderful and mean nothing.

Ask what the model quietly assumed

Every backtest simplifies. The useful question is which simplification is doing the work. Options that never existed at the strike quoted, an expiry available on any day of the week, a fill at the midpoint every time, a volatility taken from an index rather than from the contract actually traded: each is defensible and each moves the answer. A backtest worth trusting names its own.

Ask whether the author has ever published a correction

This is the least technical test and the most reliable. Anybody who has run a backtest for long has found a bug in it. A body of research with no corrections in it has either been very lucky or is not looking, and the second is far more common.

The research behind this site is published with its own limitations, its costs stated as zero, and the corrections left in. Read it, including the study that concludes covered calls lose to simply holding the index.

Going deeper

Longer pieces on one question each.

• Where this money belongsAn employer match and a tax-advantaged account both beat this strategy, on arithmetic that needs no backtest. The order, and the reasoning for each step.

The words, in one place

Everything above, for looking up later.