A covered call overlay that added 2.1% a year in a ten-year backtest lost about 1.5% in its first live year. The backtest filled at mid, the live market was 2 vol points wide, volatility halved and one month rallied 8%. Where did the gap come from?

The gap is mostly the volatility regime, then the fills, then the rally. At 12 vol the monthly premium drops from about 0.61% to 0.18% of the holding, which cuts the overlay's expected net from +2.1 to roughly +0.6 a year. Selling 1 vol point under mid each month costs another 0.7 points. The 8% month paid out 3%, about 1.4 points more than the backtest's average payout. Together they land at -1.5%.

Think of a shopkeeper who sells insurance against his own stock rising in price. His income is the premium; his cost is the months when the stock does rise and he hands over the gain above the strike. The premium on a 5% out-of-the-money one-month call is set almost entirely by implied volatility, so the overlay's income is a bet on the volatility regime, not on the index. Priced with Black-Scholes at a 7% rate and no dividends, the call is worth about 0.61% of spot at 18 vol and 0.18% at 12 vol. Twelve months of selling at 18 vol collects about 7.3% a year; at 12 vol it collects about 2.2%.

Now read the backtest through that lens. Collecting 7.3% and netting 2.1% means the calls paid out about 5.2% a year in rally months, roughly 71% of the premium. That ratio is the market's price for the upside the fund gave away. If the same ratio holds at 12 vol, the overlay collects 2.2%, pays back 1.6% and nets only about 0.6%, before costs. The premium level bar in the waterfall is that fall from 2.1 to 0.6, about -1.5 points.

A market quoted 11 bid at 13 offered in volatility terms means the desk sells 1 vol point below the mid every month. The cost per month is the option's vega, about 0.055% of spot per vol point for this call at 12 vol, so the year's fills cost about 0.7% of the holding. The backtest assumed that cost away by filling at mid. For a strategy whose entire gross income is 2.2% a year, a 0.7% execution cost is not a rounding error; it is a third of the income.

The index rose 8%, the calls were struck 5% up, so the fund handed over 3% of the holding in that one month. Payouts of 1.6% were already expected at 12 vol; the 8% month added about 1.4 points of loss above that expectation. Months like it were inside the backtest decade too, which is why the backtest showed only 2.1% net and not 7.3%. The honest reading is that the rally is the smallest of the three causes and the one most likely to reverse.

Three things, each of which was available before going live. Fill every sale at the bid, or at mid less half the historical spread, and show the result net of that. Split the decade by volatility regime and report the overlay's net in the low-vol years on their own, because a strategy that only pays in high-vol years is a different product from one that pays every year. And show the distribution of monthly outcomes, not the average, so the committee sees that the overlay earns small amounts most months and gives back large amounts in a few. A backtest that fills at mid and trains on a higher-volatility decade overstates what a premium-selling overlay can earn, and the overstatement is largest exactly when volatility is low. Say the limitation out loud: the attribution here holds the backtest's payout ratio fixed, and that ratio is itself an estimate from ten years of data.