Sell only when both conditions are true
- Options look rich: current model ATM 30-day IV is at least 20% above trailing 20-day realized volatility.
- The stock has already fallen: the 63-day price return is −15% or worse.
When either condition fails, the strategy simply holds the stock and sells no call. The threshold was selected on a discovery seed, frozen, and then tested on fresh Monte Carlo paths.
Why compare with an exposure-matched control?
A short 0.25-delta call reduces upside exposure. So beating 100 shares can sometimes be nothing more than de-risking. The primary control therefore holds 100 shares when the gate is off and 75 shares plus cash while a call is active.
This is a static initial-delta control, not dynamic option replication. Its purpose is to ask whether option premium adds anything beyond simply owning less stock.
What the frozen holdout says
10,000 paired paths per fixed regime; +10% true variance risk premium; $25,000 account with 100 initial shares; fees, spread friction, dividends, and cash yield included.
Interpretation
The conditional strategy is not a universal buy-and-hold replacement. It still gives up a small amount in persistent strong bull markets and is effectively neutral in calm bulls. The more interesting result is that, when options are genuinely rich, the gated call adds return relative to the static exposure-matched control.
The emerging hypothesis: the edge is not “covered calls.” The edge is selectively selling expensive optionality when the compensation is high enough.
Primary holdout: +10% variance risk premium
10,000 paired one-year paths per regime. Confidence intervals are paired-path 95% intervals.
Does the edge survive when options are fairly priced?
Choose the true variance-risk-premium assumption used to generate option prices. The gate itself does not know this value.
Three-year persistence check
5,000 paired three-year paths. No gated path required negative cash with the $25,000 account buffer.
Historical-underlying bootstrap check
The underlying return sequences below are bootstrapped from bundled SPY and QQQ adjusted-close histories. Option prices are still generated by the model, so these are not archived-option backtests.
Is the rule a knife-edge result?
We froze a 3 × 3 neighborhood around the candidate rather than searching for a new optimum: IV richness thresholds of 15%, 20%, and 25%, crossed with 63-day declines of 10%, 15%, and 20%. Every combination was tested on fresh seeds.
The “average edge” is an equal-regime diagnostic, not a forecast of portfolio return. It is included only to summarize the threshold neighborhood. The original 20% / −15% rule remains frozen; we are not replacing it with the best-looking neighbor.
Frozen rule is now being tested forward in time
Simulation and robustness testing are complete enough to justify a harder test: stop optimizing and record what the already-frozen rule says using only information available at each future decision date.
The thresholds remain locked at IV/RV20 ≥ 1.20 and 63-day return ≤ −15%. Any future change requires a new experiment ID.
Reproducibility
Download the frozen protocol and the validation data used on this page.
Can the Wheel have an edge if we wait for the right conditions?
Instead of selling puts continuously, the candidate waits in cash until two observable conditions agree: 30-day ATM IV is at least 30% above trailing 20-day realized volatility, and the underlying has not risen over the prior 63 days. Only then does it sell a 30-DTE, approximately 0.25-delta cash-secured put. If assigned, it completes the Wheel with 30-DTE, approximately 0.25-delta covered calls until the shares are called away.
What is the control?
Because the Wheel changes equity exposure by state, buy-and-hold is not the primary alpha test. The control holds 0 shares when the strategy waits in cash, 25 shares while a 0.25-delta put is active, 100 shares after assignment before a call is sold, and 75 shares while a 0.25-delta covered call is active. Residual capital earns the same cash rate.
This is a static phase-delta control, not exact dynamic replication.
Why cash-secured really matters
The research engine now blocks a new put if current cash is insufficient to cover 100 shares at the selected strike. That safeguard eliminated negative-cash paths in the three-year validation. Some long sequences correctly stop opening new puts after losses reduce available collateral.
CSP or full Wheel?
The same frozen entry gate was tested as CSP-only: assigned shares were immediately liquidated and the strategy returned to cash. CSP-only had a positive exposure-controlled edge at +10% option richness, but the full Wheel was larger in every fixed regime in the one-year holdout.
Frozen Wheel holdout: 10,000 paired paths
+10% model option-richness setting. Fees, spread friction, dividends and 4.5% cash yield included.
Option-richness falsification
The gate never sees the hidden model premium. This sweep asks whether the exposure-controlled edge disappears when options are fairly priced.
Three-year persistence
5,000 fresh paths per fixed regime with a strict cash-secured collateral check.
Threshold-neighborhood test
A fresh 3 × 3 test crossed IV-richness gates of 20%, 30%, and 40% with 63-day return ceilings of −10%, 0%, and +10%. At +10% option richness, all nine variants were positive versus the exposure control in all six fixed regimes. At fair pricing, the broad positive pattern disappeared.
Historical-underlying bootstrap
Five-year SPY and QQQ return-block bootstraps; option prices remain model-generated, so this is not an archived option-chain backtest.