Research

What the strategy is

The whole thing, before any of the numbers.

You own an S&P 500 index fund. Once a month you also sell a put on that index, a contract that pays someone else if the market falls below a set price before a set date. They pay you up front for it. If the market drifts sideways or up, which it usually does, you keep that money. If it falls hard, you pay out more than you were paid.

That is the whole thing. You are selling crash insurance to people who want it, on an index you already own. The contract runs about a month, you buy it back when it has made half its money, and you sell a fresh one at the current price.

Why there is anything to collect

Option prices are set by implied volatility, the market's estimate of how much things will move. Over 34 years that estimate came in higher than what actually happened 83% of the time, by an average of 3.7 volatility points. It was positive in 33 of 34 years; 2008 was the exception.

People overpay for crash protection for the same reason they overpay for every other kind of insurance: being wrong about it is not survivable, so they pay to not find out. Selling into that gap is the entire edge. It is well documented, it is not a secret, and it is not large.

What it adds, and what it costs

Three things it is not

It is not a hedge. It loses money when the market falls. That is when the puts you sold pay out, and it happens at the same moment your shares are dropping. If you want protection, this is the opposite of it.

It is not income. The premium is not a dividend or a yield. It is payment for a risk you accepted, and roughly 28 cents of every dollar collected survived after the losing months were settled.

It is not timing. Every rule tested for when to sell, high VIX, an inverted volatility curve, a drawdown, price below the 200-day average, turned out to be noise once the results were adjusted for the fact that frightening periods pay more premium anyway. There is no clever version of this.

Who should not run it

• Anyone with unused 401(k) or IRA space, which beats this by a wide margin and cannot host it anyway, since those accounts are not permitted to borrow.
• Anyone who would sell during a drawdown. This deepens the worst one you will sit through, and a position closed at the bottom converts a loss you would have recovered from into one you keep.
• Anyone who needs the money within ten years.
• Anyone who wants an exciting return. One to two points a year is the honest number, and no amount of optimisation moved it.

What 8,424 cycles actually say

Monthly at-the-money short puts on SPY, 1993 to 2026, closed at 50% of maximum profit.

The overlay sells one at-the-money put a month against a portfolio you already hold, buys it back at half its maximum profit, and resizes off whatever the account is worth that day. Run across every start day in the sample, that is 8,424 overlapping cycles.

The losses are rare and large rather than frequent and small. That is the whole reason sizing dominates every other decision. Thirteen percent of cycles lost money, and the worst one lost 17.6 times what it collected. A position sized off the typical month is not sized for the month that matters.

The 28 cents figure is worth reading carefully. It is what was retained per dollar of premium sold when closing at 50% of maximum profit. Held to expiry the figure is higher, but that is a different set of rules measured over 402 non-overlapping windows, and quoting the higher number next to close-at-50% settings would describe a strategy nobody ran.

The survivable range is 25 to 40 percent, not 40 to 50

What changed: charging margin on the shares you already hold.

An earlier version of this research put the survivable range at 40 to 50 percent of notional. That was published before a correction to how margin was calculated, and it was optimistic.

The backtest charged margin on the short put and treated the long book as free. It is not. A typical broker requires 25% maintenance on the shares you hold, and that book is usually the larger position by some distance. Including it moved the range down to 25 to 40 percent, and turned 100% notional from something that survived into something that got a margin call on 2008-10-10.

This matters more than the numbers suggest, because a margin call is not a deeper drawdown. It is a position closed at the bottom by somebody else, which converts a loss you would have recovered from into one you keep.

Every worst moment in that table falls on 2008-10-27, at every size. A 33-year sample containing one 2008 is not a large sample of crises.

The betas this tool uses

Measured from three years of daily returns, published here because they do the arithmetic on your book.

Every cushion figure in the calculator is computed on a blended beta: each holding weighted by its share of the book, then the market fall that would exhaust your margin divided by the result. A book that falls faster than the index runs out of room sooner, at the same dollar exposure. So the beta applied to your holdings is not a footnote, it is an input to the number that says whether a position survives.

These are the figures the tool applies, measured 2026-08-28 as cov(r, r_SPY) / var(r_SPY) over three years of daily returns:

They are measured rather than looked up because published betas disagree. The figure depends on the window, the return frequency and the benchmark, and a vendor rarely states which it used. A number quoted to two decimals with no method behind it is not more precise than one you compute yourself, it is only quieter about its assumptions.

The gap between a rounded beta and a measured one is not academic. On a portfolio that is 58% QQQM, rounding the NASDAQ holdings to 1.15 gives a blended 0.963, while the measured figures give 1.017. That is the difference between a book that falls slower than the index and one that falls slightly faster, and it moves the market fall that would liquidate you by three points.

Why they go stale

Beta is not stable, and it does not drift in a convenient direction. It rises in exactly the selloffs the cushion exists to survive, as correlations across equities converge and the things that usually diversify a book stop doing so. A table measured in a calm year therefore understates the risk it is used to compute, at the moment that understatement costs the most.

The build warns when these figures are more than six months old, which is a reminder rather than a fix. Treat every cushion here as computed on a beta that was true in an ordinary market, and assume the real one is higher when it matters.

What the blend leaves out

Cash and money market funds are held at beta 0 and stay in the blend. That is not a gap, it is the correct answer: cash does not fall with the index, so a book holding it genuinely falls less, and the blended figure should say so.

Option contracts are excluded from the blend entirely. A contract is a liability at market value rather than market exposure, and this tool models no option delta, so it has no beta to contribute. Carrying one at zero would not be neutral. It would pull the blended figure toward 1.00 and make a concentrated book read as closer to the index than it is, which is an error in the direction that flatters the account.

Contracts still count toward equity, because equity is what the account is worth and that is what the contract count is sized from. They are absent from the beta blend and from nothing else. The consequence worth understanding: the cushions treat a contract as a fixed value rather than modelling how it behaves in a fall, so a book whose contracts are large relative to its shares is described less well by every figure here.