Does selling Cash-Secured Puts beat buy-and-hold? 19 years of S&P 500 data

Selling cash-secured puts is sold as steady income - so does it actually beat just holding the index? We answered it with the real numbers, not a model: CBOE publishes total-return index series for systematic option-selling, so we compared them head-to-head with buy-and-hold over the 19 years they all share (2007–2026). The short answer: every flavor of option-selling trailed buy-and-hold on return - while cutting the worst drawdown by roughly 15%-33%. It is an income-and-stability trade, not an index-beater. Then we used our own model to settle the question the indices can't: which strike?

The real numbers (2007–2026)

These are the actual published CBOE total-return indices - real fills, dividends and premiums reinvested, no option modeling - measured against buy-and-hold SPY over the identical window:

Growth of $100k, 2007–2026

The same story as one picture: buy-and-hold (slate) pulled away on return, while the option-income indices - cash-secured puts (gold) and 30-delta covered calls (blue) - compounded more slowly but stayed flatter through the 2008–09 crash and the 2022 bear.

What the real data says

Buy-and-hold won on return, and it was not close: $100k grew to $723k holding the index, versus $361k for the cash-secured put index. In a 19-year stretch dominated by a bull market, selling options caps your upside, and that cap is expensive.

But look at the drawdowns. The cash-secured put index lost 37% at its worst against buy-and-hold's 55% - it sits in cash and is only exposed below the strike. Covered calls were different: the BXM and BXMD indices fell 40–47%, deeper than the puts, because a covered call holds the stock straight through every crash. And notice the 30-delta covered call (8.12%) beat the at-the-money one (5.64%) - the first hint that the strike you choose matters as much as the strategy.

Which strike should you sell? (real fills)

The real indices only come at-the-money (or, for BXMD, a single 30-delta). To see how far the strike moves the result, we ran the same monthly cash-secured put on real OptionsDX end-of-day SPY option chains, 2010–2023 (163 monthly cycles). At each roll we pick the put whose actual delta is closest to the target, sell it at the bid/ask mid, and hold to expiration - the full account backing one put at a time (100% cash-secured, no leverage, collateral sitting idle). No option pricing model anywhere; these are traded prices.

^† "Win rate" = the share of cycles that ended in profit - the premium covered any assignment loss. That is not the same as expiring out-of-the-money: near the money a put can be assigned and still pay, so the out-of-the-money share is lower. Either way it is a count, not a size: the 10-delta won 95% of its months yet compounded to just 1.31% a year.
^* Fills are at the bid/ask mid; selling at the worse bid instead costs about 0.1% a year (5.93% vs 6.03% at the money) - SPY is liquid enough that slippage isn't the story. Over 2010-01-04 to 2023-12-01 - the nearest points in our monthly index series to this ladder's window, each being a month's first session - the real CBOE PUT index compounded at about 7.2% a year against this ladder's 6.03% at the money. One difference behind that gap: PUT invests its collateral in Treasury bills, and this ladder leaves it idle. We do not compute a T-bill return here, so we do not attribute a share of the gap to it.

The shape is the lesson, and it is monotonic: the closer to the money you sell, the more you make. The far-out-of-the-money "safe" 10-delta strike - the one most retail sellers reach for - made just 1.31% a year - barely above cash, before the interest this backtest leaves idle - and a fraction of the 6.03% the at-the-money strike earned. Its thin premiums simply never added up. Same strategy, same window: the gap between the best and worst outcome was almost entirely strike selection.

Worth spelling out, because the 10-delta result is so counterintuitive: it won 95% of its months, ending almost every one in profit - yet it turned $100k into only about $119k over 14 years, because the rare months it was assigned gave back most of those small premiums - a single one cost 25.1%. That's the whole "win rate isn't profit" point in one line: many small wins, a few big losses, a thin net. Win rate measures how often you win, not how much.

The full year-by-year record: a defensive trade, not an outperformer

Here is every complete calendar year - not a hand-picked few. The column that matters is the last one, the difference. Put-selling beat the index in only 5 of 18 years: 2008, 2009, 2015, 2018 and 2022. Most wins came in flat or down markets, but not all - it also beat SPY in 2009 (SPY +22.7%) and 2015 (SPY +1.3%). In the other 13 up years it lagged, often badly. That is the signature of a defensive income trade, not a rule that every up year loses and every down year wins. (This longer 30-delta series is modeled with Black-Scholes + VIX to reach back through 2008; the real OptionsDX fills above match it to within about a point a year.)

The defensive case is clearest in the down-market wins: 2008, 2018 and 2022. In 2008, the Put Seller still lost -21.9% against SPY's -36.2% - materially less, but still a brutal loss. You pay for that cushion in most bull years. The honest read isn't "Put selling wins" - it's "Put selling trades away much of the good years to soften some of the bad ones." Whether that trade is worth it, and which strike actually pays, is the actionable part - covered below.

The myth this kills: far-OTM Puts are not safe income

The single worst environment for option-selling is not a slow bear - it is a sharp crash followed by a fast recovery, and 2020 is the textbook case. Here is the counterintuitive full-year result, straight from the real fills: the further out-of-the-money you sold, the worse the strategy finished that calendar year. This is not a comparison of each strike's worst month. (We unpack this trap on its own page: why selling far-OTM puts for "safe" income is a myth.)

That still breaks the "sell far-OTM Puts for safe income" pitch. The 10-delta Seller collected a few cents of premium, then watched SPY fall roughly a third straight through the strike - the thin premium cushioned little, so the strategy finished the year at -19.8% versus -6.1% at-the-money. But do not turn that calendar result into the wrong crash claim: the 10-delta's worst month was -25.1%, while the at-the-money Put's was deeper at -29.6%. The honest conclusion is narrower and still sharp: far-OTM produced the least long-run income and still took a severe crash hit. A high win rate did not make it safe.

And a put-seller doesn't get the rebound either - after the crash you just collect premium and watch the V-recovery you're not part of. That is the structural cost of being short options, invisible until a 2020 happens.

Covered calls are the same trade

The real indices above already include covered calls, and they make a point worth spelling out: a covered call and a cash-secured put at the same strike are the same trade. By put-call parity their payoffs and risk are essentially identical - both cap your upside, both leave you exposed below the strike, both pay you a premium today. The only difference is your starting inventory: you sell a cash-secured put when you hold cash and want in; you sell a covered call when you already hold shares and want income. The Wheel just alternates the two. That's why the put-write and buy-write indices land in the same place - lower return than the index, smaller drawdown (deeper for the share-holding buy-write). It's also why the popular covered-call income ETFs trail their index: see what QYLD's 14% yield actually pays.

What this study is actually for (and who should ignore it)

Fair question to ask of any backtest: what do I actually do with it? Straight answer - for most people, anyone young, accumulating, and able to leave the money alone through a crash, this study says buy-and-hold, full stop. The real PutWrite index made 6.87% a year to the index's 10.79% - $361k versus $723k on $100k - and it even loses slightly on a risk-adjusted basis. If that's you, stop here and buy the index; the drawdown you'd be paying to avoid is just noise to you, and pretending otherwise would be a sales pitch.

Here is the one finding that earns the page - the number you cannot get from "buy-and-hold wins": the strike you pick decides everything, and the one most people default to is the loser. Same strategy, same 14 years of real SPY fills - the far-OTM 10-delta "safe" put returned just 1.31% a year, turning $100k into $119k - barely above cash before the idle-collateral adjustment, while the at-the-money strike made 6.03% ($222k). It "won" 95% of its months, but a high win rate isn't profit: the thin premium never paid for the rare deep hit. In calendar 2020, the 10-delta returned -19.8% versus -6.1% at-the-money; across their worst months, the order reversed at -25.1% versus -29.6%. The long-run CAGR gap - 1.31% versus 6.03% - was driven by nothing but where you set the strike. If you sell 5-10 delta Puts to feel safe, that is the exact mistake this study tells you to stop making.

The second, narrower use: this is a decumulator's tool, not an accumulator's. If you're drawing 4-5% a year, what decides whether you run out isn't 20-year CAGR - it's how deep the hole gets while you're pulling cash out. Put-selling cut the worst drawdown from -55% to -37%, and the premium is cash in hand in flat and down years when a buy-and-holder may have to sell shares at a loss to fund spending. That's sequence-of-returns insurance, worth a point of return - but only if a drawdown would actually cost you. The honest catch: a sharp crash-and-rebound breaks it. In 2020, the Put Seller finished -13.8% while the index finished +17.2% - it ate the drop and missed the bounce.

If you're going to sell puts anyway, the one rule that matters: sell the income strike (around 30-delta) on a stock you'd be happy to own, when IV is high - never the far-OTM "safe" strike, the one that barely beat cash before the idle-collateral adjustment, still took a severe crash loss, and finished the crash year worst. Check whether premium is genuinely rich with the IV rank calculator, and size the strike to a real dollar yield with the cash-secured put calculator: if the annualized return on capital is low single digits with a huge cushion, you're in the 10-delta trap.

Caveats - read these

• Two real datasets, two windows. The headline CBOE PUT/BXM/BXMD figures are published total-return indices (2007–2026). The strike ladder is real OptionsDX end-of-day SPY fills (2010–2023) - also real, just a shorter window the free option data covers (it includes the 2020 crash but not 2008). The year-by-year 30-delta table above extends to 2026 with a Black-Scholes + VIX model, which the real fills confirm to within about a point a year.
• Idle collateral. The ladder leaves the cash collateral earning nothing. Parking it in T-bills (near zero in the early 2010s, ~5% by 2023) would lift every row a little - but nowhere near enough to close the gap to buy-and-hold's 12.66%.
• One window, dominated by a bull market. 2007–2026 starts just before the 2008 crash and is mostly a historic bull run, which maximally penalizes any capped-upside strategy. A different 19 years would land differently.
• Indices aren't free to run. The published series exclude the fund fees, commissions, slippage and taxes a real seller pays - so a real-world result would be a touch lower than the index.
• No active management. These are mechanical, hold-to-expiration rules - no rolling, no closing at a profit target, no dodging earnings. Active management changes the picture.
• Past performance is not predictive. Educational, not advice.

Sources: CBOE PUT / BXM / BXMD total-return indices and SPY dividend-adjusted prices (2007-02-16 to 2026-05-15) for the headline; real OptionsDX end-of-day SPY option chains (2010-01-15 to 2023-12-15) for the strike ladder, sold at the bid/ask mid with strikes chosen by actual delta. Every figure regenerates from the underlying data; none are hand-entered. Every backtest behind this site, including this one, is published as plain JSON and CSV at /data/, licensed CC BY 4.0 for reuse with attribution.