The Wheel strategy, backtested: 21 years on the S&P 500

The Wheel is the internet's favorite options-income strategy: sell a cash-secured put, and if you get assigned the shares, sell covered calls against them until they get called away - then start over. The pitch is "get paid while you wait, and keep getting paid." So we ran it, mechanically, for 21 years on the S&P 500, and put the equity curve next to simply buying and holding.

The honest result: the Wheel did not beat buy-and-hold. It lost by about 3 points of return a year - and that is exactly what a fair backtest should show. But it lost more gently, and that is the real story. This is a risk-reduction trade, not a market-beater, and anyone selling you "12-25% a year, triples the market" is showing you return-on-capital in a cherry-picked window, not a 21-year equity curve.

The scorecard

$100,000 run on the Wheel from 2005-02-18 to 2026-05-15, against the same $100,000 sitting in SPY with dividends reinvested:

The Wheel lost the race on return - but its worst drawdown was -35.9% against buy-and-hold's -55.2% (about two-thirds as deep), at roughly two-thirds the volatility. You felt 2008 and 2020 far less. Read the two lines as "less return, much less pain", not "it failed."

Real fills change the verdict - and not in the Wheel's favor

The scorecard above prices every option with a Black-Scholes + VIX model. We since bought real OptionsDX end-of-day SPY chains and re-ran the exact same machine on traded bid/ask prices, 2010–2023 (163 cycles), picking each strike by its real delta. The risk shape held - but the return came in lower, exposing a flaw that runs the opposite way from what we'd assumed.

On real premiums the Wheel earned 6.45% a year against buy-and-hold's 12.66% - a wider gap than the model implied. The machine itself behaved identically: 22 put-assignments and 22 call-aways, 25% of the time in shares, and the worst cycle still the March-2020 assignment (-28.5%). The defensive profile is intact; it's the income that was overstated.

The culprit is volatility skew on the call leg. The model prices a 30-delta covered call off at-the-money VIX, but real index calls trade at a markedly lower implied volatility - in this data a 30-delta call's IV averaged 14% against 18% for the matching put. The Wheel sells those calls whenever it holds shares, so the modeled covered-call income was partly phantom. The put leg is the opposite and roughly a wash (real out-of-the-money puts are slightly richer than VIX implies). Net: contrary to the "conservative floor" caveat we first wrote, the VIX model overstated the Wheel, because the call leg dominates whenever shares are held.

Growth of $100,000, 2005–2026

The chart tells the truth on its own. The Wheel (gold) tracks the index for years, then the bull runs of 2013, 2019 and 2021 - and the post-COVID V-recovery - pull buy-and-hold (slate) away, and it never catches up.

How the machine actually runs

The Wheel is a two-state machine, and our backtest follows it mechanically every monthly expiration:

• When flat (in cash): sell a 30-delta cash-secured put, fully collateralized, no leverage. Collect the premium and wait for the third-Friday expiration.
• If the put expires worthless: keep the premium, stay in cash, sell another put. This is about 70% of all periods - the Wheel is mostly a cash-and-premium machine.
• If the put is assigned: you now own the shares at the strike. Switch legs and sell a 30-delta covered call against them, marking the shares to market and accruing the dividend.
• If the call is assigned (called away): the shares are sold at the call strike, you are back in cash, and it restarts.

Over 256 monthly cycles there were 30 put-assignments and 30 call-aways - it always ended flat, perfectly balanced. Because the index drifted up over 21 years, the realized assignment rate (~17% per put sold) ran below the 30-delta nominal: in an uptrending market fewer puts finish in the money than the delta implies.

Assignment, and the 2020 V-crash cycle that hurt most

Assignment is a designed outcome, not a failure - but it is also the exact moment the Wheel turns a put-seller into a share-holder, right as the market is falling. The worst single period in 21 years was the cycle ending 2020-03-20: the put was assigned at strike 326 while SPY closed the expiration at $228.8 - a -28.8% hit for that month, and the premium barely dented it.

That is the mechanism critics warn about: as price falls your delta ratchets from ~30 toward 100, so you "double down" on the stock exactly when you would want to de-risk. Then the V-recovery punished it twice - after assignment you sell covered calls, and the snap-back rally got the shares called away near the lows, capping the upside on the way back up. That is why 2020 ended -4.6% for the Wheel while SPY ended +17.2%. The premium lowers your cost basis; it does not eliminate the drawdown.

Where it shines: the flat and down years

The Wheel wins in flat and down years - the documented signature of buy-write strategies (Wheel vs SPY total return):

2022 is the best case: a grinding, choppy bear where premium outpaces flat-to-down price action and you are often in cash. But say the mechanism precisely - the edge comes mostly from the cash phases and collected premium, not from the covered call shielding assigned shares. Once you are assigned, you hold the full stock drawdown; the Wheel cushions by roughly the premium collected, then rides the rest down. It does not protect in a bear, it softens.

The bull-market drag: the price you pay

The flip side, and where the 3-point gap comes from - strong bull years, where the Wheel leaves more than half the gain on the table:

Two leaks: when flat you only earn premium while SPY runs without you, and when long your covered call caps the gain at the strike. This is a structural cost, not a tuning error - you are always short the right tail. A higher-delta or further-dated Wheel changes the numbers but not the shape.

What the model assumes - and which way each leans

This is a documented model, not a promise, and honesty means stating which direction each simplification pushes the result. The real-fill re-run above settles the biggest one - and it turned out to overstate the Wheel, not understate it (the call leg trades cheaper than VIX implies). The rest:

• Premiums priced from VIX overstate the Wheel, via the call leg. VIX is an at-the-money volatility. Real out-of-the-money index puts trade slightly richer than that (a small tailwind), but real out-of-the-money calls trade markedly cheaper - and the covered call is the leg that runs whenever the Wheel holds shares. On real fills that pulled the CAGR down to 6.45% (above), so the modeled figure is an optimistic ceiling, not a floor.
• Idle cash earns 0% here - unrealistic, since the Wheel is in cash ~70% of the time. A real trader parks the collateral in T-bills or money-market; crediting even ~2% would add roughly 1.3-1.4% a year, of the same order as the entire gap to buy-and-hold. This assumption penalizes the Wheel.
• No commissions, spreads or slippage - premiums are theoretical mid-prices a seller never quite gets. This one leans the other way, flattering the Wheel slightly.
• Held to expiration, zero management - no rolling, no early close, no defensive adjustment. The -28.8% March-2020 print is the unmanaged robot result; active management would change the path in both directions.
• One path, one underlying, one window. A single 21-year run on SPY over a bull-dominated period - no Monte Carlo, no alternate underlying. On a sideways decade or a single volatile stock (concentrated assignment risk), the ranking could flip. Past path does not predict future path.

Who this actually suits

Our result lands squarely in the credible middle of the evidence. The rigorous, long-run anchor is the CBOE BXM buy-write index - about 8.5% a year versus the S&P's ~11% since 1986, at two-thirds the volatility - and spintwig's SPY Wheel study, where none of ten variants beat buy-and-hold on total return. We are in sober company; the "beats the market" crowd is quoting return-on-capital, not equity curves.

• Genuinely suits someone who wants equity-like exposure with materially less drawdown and will accept giving up a chunk of bull-market upside to get it. That is the real trade.
• Wrong fit for anyone chasing maximum compounding - buy-and-hold won this 21-year race and the math isn't close ($501,695 vs $907,574).
• Only wheel an underlying you would be content to hold through a 30%+ drawdown, because assignment will hand it to you at the worst time. A broad index like SPY qualifies; a single hot stock does not.

Method: real daily SPY and VIX history, 2005-02-18 to 2026-05-15 (256 monthly third-Friday cycles). 30-delta puts when flat, 30-delta covered calls when assigned, held to expiration, cash-secured (no leverage). Premiums priced with the same Black-Scholes engine used across this site's calculators; held shares marked to market with a 1.7%/yr dividend; collected premium is reinvested into the next position (continuous compounding), and idle cash earns nothing. Buy-and-hold uses dividend-adjusted SPY over the identical window. The numbers were independently reproduced and the share-P&L accounting audited. The "real fills" section re-runs the identical machine on real OptionsDX end-of-day SPY chains (2010–2023), pricing every leg at the bid/ask mid and choosing strikes by actual delta - four independent backtests agree on the 6.45% real result. Every figure regenerates from the underlying data; none are hand-entered. Educational, not advice.