Closing at 21 DTE, replayed: it earned about what T-bills paid

The conclusion first

We sold a 25-delta cash-secured put on QQQ or SPY once a week from April 2022 to October 2026, 30 to 45 days out, 574 trades in all. Entries filled at the real closing bid and every early exit at the real closing ask. Then we scored the same trades three ways: held to expiry, closed early at half the credit, or closed early at 21 days to expiry.

This is about when to close, not when to open. Opening a trade at 21 days and holding it to expiry is a different trade, and it is not tested here.

Held to expiry, they earned +5.6% a year on the cash tied up. Closed at half the credit, +8.4%. Closed at 21 days to expiry, +4.4%. Over the same years a 13-week Treasury bill averaged about 4.15%, with no risk at all. The 21-day rule kept the work and the tail of selling puts and paid roughly the T-bill rate for it.

The numbers

QQQ and SPY, 25-delta puts, 574 trades over 55 expiry months. Held to expiry: won 89% of the time, +0.59% of collateral a trade (95% interval +0.26% to +0.88%), 39 days in the trade on average, +5.6% a year, worst month -3.8%. Closed at half the credit: won 95%, +0.37% a trade (+0.11% to +0.57%), 15.5 days, +8.4% a year, worst month -3.6%. Closed at 21 days to expiry: won 76%, +0.22% a trade (-0.01% to +0.42%), 18 days, +4.4% a year, worst month -2.8%.

Split by fund it reads the same. SPY: +5.1% held, +6.8% at half the credit, +4.1% at 21 days. QQQ: +6.2%, +10.7% and +4.8%. On the leveraged ETFs we also tested (TQQQ, UPRO and others) the 21-day exit lost money over the window, -2.6% a year, while holding earned +13.0% with far deeper bad months.

The per-year figures credit early exits for freeing cash: each is the profit per dollar per day the cash was tied up, times 365, so a trade that earns half as much in a third of the days scores higher. That assumes the freed cash goes straight into the next trade, and it leaves out any interest your broker pays on the cash.

Why it costs so much

The 21-day exit closes every trade at the same point, the winners included. At 21 days a put sold at 37 days still holds most of its time value, so the rule buys back premium you would mostly have kept, and pays the ask to do it. Closing at half the credit also frees cash early, but only on trades that have already decayed; the ones still worth a lot are left to finish.

The usual case for the rule is that the last three weeks carry the sharpest risk. It did trim the worst month, from -3.8% to -2.8% of the cash set aside. Over these years that protection cost about 4 points a year against closing at half the credit, and it pushed the win rate down, not up: 76% against 95%.

Spreads tell the same story

Our earlier study of 25-delta put credit spreads, net of fees, found the same pattern per trade: on the largest QQQ set (396 trades) closing at 21 days to expiry kept about a quarter of what holding earned, and on both TQQQ sets it lost money while holding made money. That study did not adjust for faster capital turnover. See the lesson on stop-losses for credit spreads.

Against just owning the index

None of these rules beat buying SPY over the same window, which returned about 13% a year with a drawdown of 21% on the way. A put seller's cash sits mostly idle; the fair comparison for the strategy is T-bills on the same cash, and the fair comparison for the account is the index. Over 2022 to 2026, selling puts and closing at half the credit beat the first and lost to the second.

What it does not show

One window, from the start of our quote history in 2022: one bear market, two sharp dips and a long recovery. Exits fill on daily closing quotes, not intraday. Entries were 30 to 45 days out, nearest 37; the classic version enters at 45, which leaves more time for decay before day 21, and that variant is not tested here. Commissions are left out, which flatters both early-exit rules slightly.

If you run the 21-day rule, the test is simple: score your own closed trades both ways, against the T-bill rate on the same cash. Paper results must use your real exit rule explains why the exit you assume changes the answer, and the strike check shows how often a put at your distance finished in the money since 2016.