Wheel Strategy Backtesting: What Actually Works (and What Lies to You)

A convincing wheel backtest is one of the most seductive things in options trading. Someone shows you a chart: "here’s the wheel on SPY over the last 10 years — 18% annualized." You want to believe it. It’s a beautiful, clean number.

Most of those charts are wrong. Not because the person is lying — the wheel really can produce that kind of number — but because the way most backtests are constructed makes them significantly optimistic. Understanding why, and how to run a backtest that isn’t optimistic, is the difference between a strategy that works on paper and one that works in your account.

1. Why backtesting the wheel is genuinely hard

Backtesting a stock strategy is easy — you have decades of clean daily OHLC data, adjusted for splits and dividends, freely available. Backtesting an options strategy is orders of magnitude harder because:

• Historical options data is expensive. Bid/ask spreads, implied volatility, and full option chains at every timestamp — actual clean data costs $1,000+/year from serious providers (ORATS, OptionMetrics, LiveVol).
• Fill assumptions matter enormously. Was your fill at the midpoint? The bid? Somewhere in between? A 5-cent difference on a 45-cent option is 11% of the trade.
• Assignment paths are non-deterministic. The same "put on SPY at 0.25 delta" trade at the same time can have different assignment outcomes depending on when you would have chosen to close or roll. Backtests have to make assumptions.
• Corporate actions break historical option chains. Splits, mergers, spinoffs, dividend changes — all mess up option chain continuity in ways that require careful data cleaning.

2. What data you actually need

A defensible wheel backtest needs, at minimum:

1. End-of-day options chains for the underlyings you're testing, for the full backtest period. Bid, ask, mark, IV, delta, and open interest per strike.
2. Underlying daily OHLC (widely available for free from Yahoo, Alpha Vantage, etc.).
3. A dividend history for the underlying — dividends affect assignment probability and cost basis.
4. A corporate actions history — splits, mergers, symbol changes.
5. Realistic transaction cost assumptions — $0.65/contract commission + $0.05–0.15 slippage per share is a reasonable modern-broker cost model.

3. The tools that actually work

Option 1: OptionAlpha (free tier, paid tier for serious backtesting)

The most polished wheel-oriented backtester. Point-and-click, decent data quality, reasonable fill assumptions. Free tier is limited to a handful of pre-built strategies; the paid tier ($100+/month) unlocks custom rules. Good for someone who wants a working backtest without writing code.

Option 2: Tastytrade backtesting (built into their platform, free)

Reasonable option backtest engine bundled with the broker. Good for spot-checking simple strategies (short puts, short strangles). Fill assumptions are optimistic (midpoint), and the strategy universe is limited. Best used for gut-check on ideas, not for definitive numbers.

Option 3: Custom Python + paid data (~$500 to $2,000 setup)

For the technical: buy end-of-day options data from a reputable provider (ORATS, IVolatility, LiveVol), write your own backtester in Python or R. Total control, honest fill modeling, complete customization. Requires ~40 hours of engineering to get right. This is what serious option shops do.

Option 4: CBOE historical index rules

CBOE publishes rule-based indices that approximate wheel-adjacent strategies: BXM (buy-write on S&P 500), PUT (short put on S&P 500), WPUT (weekly short put on S&P 500). These aren’t exactly the wheel, but they’re professionally maintained and give you a solid baseline for what a mechanical premium-selling strategy on SPX has produced historically. Not a substitute for real backtesting, but useful sanity check.

4. The four ways backtests lie

Lie #1: Survivorship bias

"Wheel on quality names over 10 years" backtests almost always look great because the "quality names" chosen today are the ones that survived. Companies that went bankrupt, got acquired at fire-sale prices, or were delisted are usually excluded from historical backtests. Result: the "wheel worked great!" chart quietly ignores what would have happened if you'd wheeled Bed Bath & Beyond, Peloton at $170, or First Republic Bank.

Lie #2: Midpoint fills

Backtests typically assume you fill at the midpoint of the bid/ask. In reality, you fill at somewhere between mid and slightly-worse-than-mid, especially on lower-liquidity names. On a wheel that trades 12–20 times per year per position, the fill slippage compounds. Realistic slippage assumption: 3–8% off midpoint per trade. That’s enough to knock 100–300 bps off annualized returns.

Lie #3: Ignoring the "assignment tax"

When you get assigned in real life and then sell covered calls, the shares don’t magically match your desired call strike — there’s a gap between what you want to sell at (your cost basis) and what the market makes available. This produces small drag every cycle. Many backtests assume you can always sell calls at exactly your ideal strike, which overstates results.

Lie #4: Ignoring taxes

Every backtest shows gross pre-tax returns. In a taxable account, wheel returns are short-term ordinary income — a 30% blended tax rate is realistic for a middle-earner. A "20% annualized" backtest in a taxable account is really more like "14% net." Not a lie in the backtest itself, but a lie by omission when someone shows you a return number without the tax context.

5. Metrics that actually matter

Backtest reports throw a lot of numbers at you. Only a few actually matter for the wheel:

1. CAGR (compound annual growth rate). The single most important number. Look for something in the 8–18% range for quality names — anything higher and you should be suspicious.
2. Max drawdown. The worst peak-to-trough decline the strategy experienced. Look at this in every market crash covered (2020, 2022, ideally 2008). If max drawdown is under 10%, the backtest is probably ignoring assignment risk.
3. Sharpe ratio. Return per unit of volatility. A Sharpe of 0.8–1.2 is reasonable for the wheel; over 1.5 is suspicious.
4. Win rate. Percentage of trades that closed at a profit. Wheel win rates are typically 70–85% (which sounds great, but the losses on the 15–30% of losing trades can be much larger than the wins).
5. Average trade P&L in $ terms. Sanity check — does the arithmetic actually add up? If you have 100 trades/year averaging $150 profit, that’s $15k. On a $100k account, that's 15% — plausible. On a $30k account, that's 50% — implausible without heroic assumptions.

6. What real backtests look like on quality names

For calibration, here’s roughly what defensible backtests of mechanical wheel strategies on quality names have produced over multi-year periods (net of realistic slippage but pre-tax):

If a backtest claims 25%+ CAGR on quality names with only −10% drawdown, it’s almost certainly wrong. The wheel produces good but not miraculous returns; the tradeoff for the premium capture is capped upside during rallies and paper drawdowns during selloffs. Both are baked in.

7. The limitations of any wheel backtest

Even a perfectly-constructed wheel backtest tells you less than you'd think about how the wheel will actually perform for you:

• Your decisions won’t match the backtest. A backtest assumes mechanical rule-following. You will deviate — closing early on a scare, holding through a losing put you should have taken assignment on, missing a roll because you were on vacation. Real returns are almost always lower than mechanical backtests.
• Market regimes change. The 2013–2019 low-IV environment behaves very differently from the high-IV 2020–2022 period. A 10-year backtest that spans mostly one regime tells you about that regime, not the next.
• Correlation surprises. A backtest of "wheel on 5 quality names" typically assumes those names move independently. In real crashes (2020, 2022), correlations spike toward 1 and all positions drop together. Backtests usually miss this.
• The wheel is discretionary at the edges. Stock selection, strike selection under uncertainty, decisions to close or roll early — all discretionary. Backtests can’t capture the value of good discretion, but they can’t capture the cost of bad discretion either.

8. Next steps

The practical path:

1. Don't make a strategy decision based on a single backtest. Cross-check with the CBOE PUT and BXM indices to sanity-check what a mechanical premium-selling strategy on the S&P should look like.
2. Paper trade the wheel for 3–6 months on 2–3 real positions. This teaches you the parts of your process that no backtest can capture.
3. Start small with real money once your paper process is boring. Journal every trade. Your first 6–12 months of real data are more informative than any backtest.
4. Compare your actual results to a backtest baseline every quarter. If you’re significantly underperforming the CBOE PUT index net of fees, your discretion is destroying value — figure out where and fix it.

If you want to skip the backtest step and just watch real wheel trades in real time with the reasoning behind each decision, the Omega Membership is where I share the weekly trade plan on real tickers. Or grab the free Starter Kit and the free Wheel Return Calculator to model returns on your specific setup.