- Selling a call on your own shares every month and buying the dips cuts SPY's drawdown from −55.4% to −37.1% over twenty-six years, while the average annual return rises from 9.26% to 11.90%. QQQ and IWM behave the same way.
- In 2008 the loss is cut in half: −36.9% → −19.0%. And it isn't a one-off: the same thing happens on three different indexes, across five different down years.
- The cost is symmetrical and has to be stated: in years above +25% the strategy gives up between 3 and 8 points versus the index alone. You buy peace of mind on the way down and pay for it with part of the best rallies.
- The part almost no backtest shows: 89% of the premiums collected go back out as intrinsic value at expiration. On QQQ the options leg actually ends up losing money.
- The capital you need is almost twice the starting amount, and the premiums here aren't a made-up percentage: they are calibrated to actual historical option prices. The backtester is free and open source, so every number in this article can be reproduced.
Prefer to watch? The same analysis on video (in Italian), with the backtester used live — 43 minutes. If you just want the numbers, they're all below.
There's a situation every investor knows well and few people talk about: I want to stay exposed to a given market, but not the way it is. Not because it doesn't return enough — because every so often it drops 40% and then produces nothing for years while you wait for it to come back. In 2008 the S&P 500 lost 37%. From 2000 to 2002 it lost three years in a row. If you were invested, there was nothing to do but watch, and nothing in your hands but the shares.
That's where the question behind this article comes from: can you stay invested and cushion the declines at the same time? Not eliminate them — you can't — but collect something precisely while the price is falling, so that the loss on your account is smaller than the market's. And, ideally, bring home a harvest every year instead of waiting twenty.
The short answer is yes, with a specific price to pay. The long answer is twenty-six years of data on three indexes, which you'll find below — including the numbers that cut the other way, which are the most interesting part.
The idea: two engines that work at opposite times
The first engine is a covered call. You own the shares and every month you sell a call on them: the buyer gets the right to take them off your hands at a set price, and in exchange pays a premium up front that the seller keeps no matter what happens. It's a trade-off: you give up part of the upside in exchange for certain, immediate income. The strike is at 0.50 delta, essentially at the current price: it's the call that collects the most time value of all. On SPY it's worth on average 1.84% per month of the value of the covered shares.
The second engine is Buy The Dip. Every time the previous month closes negative, more of the underlying is bought at the next month's open, in proportion to the depth of the decline plus a fixed amount:
These extra shares are not covered by the call, so their upside is uncapped: that's why the structure is asymmetric, and it's also why I call it boosted. The two mechanisms push in the same direction but at opposite times: the call collects when the market isn't running, which is almost always; the purchases buy when the market has fallen, which is when shares are cheaper. Neither predicts anything — both react to what has already happened.
In January you buy the underlying with the set capital. Those shares are the covered base and stay unchanged all year. In December you liquidate everything and withdraw everything: shares and cash. The following year only the starting capital goes back in, always the same amount; anything above that is the harvest.
It looks like giving up compounding, and it's a deliberate choice: a strategy that resets every year is a strategy you can judge every year, and one you can walk away from without having to unwind anything. Above all, it produces cash instead of paper gains.
What happened from 2000 to today
Fixed capital of €50,000 every January, boost at 1.5%, full withdrawal every December. The comparison isn't with a buy and hold that never liquidates — that would be unfair — but with the underlying alone following the exact same cycle: same capital in January, everything liquidated in December, no options and no dip buying. It's the only honest comparison.
| SPY · QQQ · IWM, 2000–2026 | SPY | QQQ | IWM |
|---|---|---|---|
| Average return — underlying alone | 9.26% | 11.95% | 9.37% |
| Average return — strategy | 11.90% | 12.66% | 12.64% |
| Drawdown — underlying alone | −55.4% | −83.3% | −59.6% |
| Drawdown — strategy | −37.1% | −68.7% | −42.2% |
| Drawdown reduction | 33% | 17% | 29% |
| Return divided by volatility | 0.53 → 0.91 | 0.42 → 0.65 | 0.52 → 0.91 |
| Profitable years | 21 / 27 | 21 / 27 | 23 / 27 |
| Median annual harvest | €9,739 | €9,272 | €8,479 |
How far the account falls at the worst moment in twenty-six years
How to read it. Maximum drawdown measured every trading day, not at month-end. The extra return the strategy produces is modest — between half a point and three points a year. What really moves is the decline.
The years the market lost
This is where you find out whether the cushion actually works. In the five worst years of the period, the comparison looks like this — year by year, the return of the underlying alone versus the strategy:
SPY: index-only return vs strategy return
In 2008 the loss is roughly halved, and in 2000 and 2001 the year closes close to breakeven while the market loses 10 and 12 percent.
| Year | SPY | QQQ | IWM |
|---|---|---|---|
| 2000 | −10.6% → −0.8% | −39.3% → −24.3% | +6.0% → +3.5% |
| 2001 | −12.3% → −3.4% | −33.6% → −12.3% | +1.8% → +2.4% |
| 2002 | −22.1% → −8.1% | −38.4% → −15.4% | −21.0% → −8.5% |
| 2008 | −36.9% → −19.0% | −41.8% → −25.7% | −34.4% → −17.3% |
| 2022 | −18.4% → −10.9% | −32.8% → −19.5% | −20.9% → −6.3% |
And when the market rips?
Here comes the bill, and it's exactly the same mechanism seen in reverse. On average, versus the underlying alone:
| Average return difference | SPY | QQQ | IWM |
|---|---|---|---|
| Years the market loses | +9.74% | +14.44% | +10.28% |
| Years above +25% | −3.52% | −8.33% | −4.36% |
The uncomfortable part: 89% of the premium goes back
This is the piece almost every covered call backtest gets wrong, and getting it wrong changes everything. The call isn't bought back in the market: it's held to expiration. And if the price has gone above the strike, the cost is the intrinsic value — paid in cash on indexes, or by taking assignment, delivering the shares at the strike and buying them back at the market price. Both routes cost exactly the same.
Premiums collected, intrinsic value paid, and what's really left
On SPY, over twenty-six years: €300,810 collected and €268,894 paid back at settlement. Net: +€31,916.
The call is not a return machine
It's a transfer of return between different years: it takes part of the best rallies and moves it into flat and negative years. The value isn't in the total, it's in when it arrives.
And on QQQ it's even starker: €407,634 in premiums, €409,597 in intrinsic value paid, net −€1,963. On the Nasdaq the options leg loses, and the strategy's entire edge comes from buying the dips. That's with premiums calibrated to real option prices: it's not a flaw in the model, it's the Nasdaq having gone up too much.
How much capital you really need
This is the easiest mistake to make, and it's the one that blows up homemade covered calls. The starting capital isn't all the capital you need: buying the dips requires fresh cash during the year, and it has to be paid for.
| Capital deployed in a year (base: €50,000) | Annual median | Maximum | Relative to base |
|---|---|---|---|
| SPY | €58,374 | €83,152 | 1.66× |
| IWM | €62,994 | €85,425 | 1.71× |
| QQQ | €61,938 | €103,191 | 2.06× |
Monthly or weekly?
The same strategy exists on a weekly schedule: 52 calls sold a year instead of 12. Each premium is worth about half — time value grows with the square root of time, not with time — but the gross income for the year doubles. Measured on SPY over the same period, though, the weekly version tells a very specific story:
SPY: what changes when you switch to weekly
Weekly harvests more and protects far less. Average return 13.49% vs 11.90%, median harvest €10,930 vs €9,739 — but drawdown −49.6% vs −37.1%.
The reason lies in the dip buying: 608 purchases vs 115. Buying after every week that closes in the red means adding exposure throughout a decline, not once a month, and the weekly premium — which is half as large — isn't enough to cover the difference. In 2008 the weekly version made 32 purchases and closed at −21.7%; the monthly one made 9 and closed at −19.0%.
Why these premiums aren't made up
The call premium is the most delicate quantity in the whole setup: get it wrong and every result shifts. In most backtests you find online it's a fixed percentage — and that's why those backtests are worthless. Here the chain is different: you measure the volatility the underlying actually realized on daily data, multiply it by a volatility risk premium that depends on the volatility level itself, and use that to price the 0.50-delta call with Black-Scholes.
The general coefficient is calibrated on 1,666 real at-the-money call sales. For SPY, QQQ and IWM it was derived directly from the historical option chains, one by one.
The result is instructive: on SPY the general model overestimated premiums by 5%. Sounds small. Over twenty-six years it was worth €16,000 on the options result — a third of everything they produced, and almost a point of annual return. On an uncalibrated underlying, read any backtest with that margin in mind.
How these numbers were measured
Three measurement choices that change the results quite a bit, and that are worth knowing in order to judge them.
The account is marked every day
Decisions are made monthly, but the account value is calculated every trading day, with the call marked to market until expiration. Looking only at month-end closes, a crash that recovers before the 31st disappears: in a purpose-built test case the measured drawdown was zero, the real one −35%.
Simple return, on capital deployed
Not a compound rate on the account, which would be diluted by idle cash, but the year's result divided by the capital actually deployed in that cycle: January capital plus the dip purchases.
The withdrawal is real
Every December everything comes out. It sounds like bookkeeping, but it matters: if you let cash pile up, it earns interest year after year on money someone who withdraws would never have had. In one SPY test that interest was worth 24% of the reported profit. There's none of that here.
What this strategy doesn't do
- It doesn't protect against crashes. It cushions them. A −37% is still a −37%: the premium collected is a cushion, not a parachute.
- It doesn't participate in big rallies. In years above +25% it gives up between three and eight points. That's the price of the cushion, and you always pay it.
- It doesn't compound. Every year it restarts from the same capital: a choice, not a flaw, but you need to know it.
- It requires work and available capital. Twelve trades a year plus the dip purchases and the liquidation, and a cash reserve of up to twice the starting capital.
- It doesn't work where options aren't liquid. With wide spreads or sparse expirations, the calculated premium is not the premium you actually collect.
- It realizes capital gains every year. The December withdrawal is a taxable event, and these simulations don't include taxes.
The verdict
Across three indexes and twenty-six years, the worst decline is cut by a third on SPY and by almost a third on IWM, with a return still higher than the underlying alone and a median cash harvest of about €9,000 a year on €50,000 of capital. But the extra return is between half a point and three points: if you're looking for a strategy that beats the market, you're on the wrong page. This one is for those who want to stay invested while cushioning the falls and collecting every twelve months, and who accept giving up part of the best rallies to get it.
The fact that the options leg loses money on QQQ and that 89% of the premiums go back aren't footnotes: they are the result, and anyone who sells covered calls should know them before starting.
Run the numbers yourself
Everything in this article is reproducible, free and with no sign-up. The backtester is public: you pick the underlying — indexes, single stocks, crypto, anything with reasonably liquid options — set the capital, frequency and the other knobs, and export everything to JSON and CSV. The real-price calibrations for SPY, QQQ and IWM are already built in and apply automatically.
- Open the backtester — free, no sign-up
- Download the full paper (PDF) — the document with every detail of the strategy
- Source code on GitHub — the calculation engine, open and inspectable
- The video with the walkthrough — the strategy explained and the tool used live (in Italian)
If you work with options, on the same blog you'll also find the numbers on systematic call selling on SPX, the numbers on put selling and the other free tool for reading options positioning.
Methodology. EODHD daily and monthly prices on SPY, QQQ and IWM, January 2000 – September 2026, 27 annual cycles. Fixed capital of €50,000 redeployed every January, 1.5% boost on every dip purchase, full withdrawal every December. Call sold at 0.50 delta expiring at the end of the period, solved by bisection on Black-Scholes; realized volatility estimated with Yang-Zhang over 126- and 504-day windows (blended at 0.60) and multiplied by a volatility risk premium dependent on the vol level, calibrated on 1,666 real ATM call sales and, for these three ETFs, on the actual historical chains. Settlement at expiration at intrinsic value with shares unchanged — arithmetically identical to assignment plus repurchase. Account marked every trading day with the call marked to market; drawdown on time-weighted returns; simple annual return on capital actually deployed. The benchmark is the underlying alone following the same annual cycle, not buy and hold. Commissions, bid-ask spreads, execution slippage and taxes are not included; fractional quantities are assumed tradable at the stated open and close prices; premiums are still model-reconstructed, with an expected error of around 5% on uncalibrated underlyings. The code that produces these numbers is public on GitHub.
This article is also available in Italian.