sell-window Backtest: Every Recommended Trade on SPY, QQQ and IWM, 2018 to 2026

More than 5 million backtests led to the sell-window endpoint. For each listed expiry it tells you whether today is a day to sell options or a day to stand aside, and which trades to place when it says sell. This backtest follows that advice on every session from January 2018 to October 2026, on SPY, QQQ and IWM: 20,883 trades in seven strategies.

The builder lets you test the advice the way you would trade it. Pick one strategy, such as the short put or the short straddle, or all of them. Trade a fixed number of contracts per signal or size by volatility, add a multiplier, and switch between fixed size and compounding.

What following every trade earned

Against the same trades sold every day, following it cut the deepest drawdown from $958,924 to $255,105, cut the worst month from −$307,241 to −$170,829 and raised the Sharpe ratio from 0.61 to 0.76. At that size it earned less in total, $945,541 against $2,321,150, because it placed about a third of the trades (20,883 against 56,975).

A higher Sharpe ratio leaves room to size up. At 3.76 times the contracts, both books have the same deepest drawdown, $958,924, and following it made $3,554,225 against $2,321,150, with a worst day of −$258,132 against −$384,980. Its worst month was deeper at that size, −$642,135 against −$307,241. On a $2.7 million account, the most margin selling every day ever needed, a 1x margin limit cut the sized-up book on 145 of 2,213 sessions and the total was $3.53 million.

With one contract set per signal, halfway fills and a $1,000,000 account at fixed size and a 1x margin cap, all seven strategies together made $949k over 2018 to 2026. That is 7.9% a year with a daily Sharpe of 0.77 and a maximum drawdown of 23.9%. It lost money in two years: 2018 (−5.7%) and 2022 (−13.1%). Compounding raised the total to $1.31 million, with a drawdown of 23.5%.

The trades it lists were chosen from these same years, so read these tables as an upper bound. How the rule held up on data it was not chosen on is in why you can trust this backtest.

The builder

The builder opens on all seven strategies, one contract set per signal, at halfway fills on a $1,000,000 account with a 1x margin cap. Each strategy combines every sell-window recommendation of that type on SPY, QQQ and IWM. To change it:

• Pick strategies. Tick a row in the table to add or remove it, or click a strategy's name to overlay its own curve.
• Sizing. Contracts trades the multiplier in contract sets on every signal: 2 means two of each trade. Target vol weights each strategy by its volatility and scales the book to the yearly volatility you set; the multiplier then scales that book.
• Position size. Fixed trades the same size every day. Compounding resizes every day to the account at the previous close: at $1,500,000 on a $1,000,000 start, each signal trades 1.5 times the size.
• Margin cap. When the estimated margin passes the cap times the account at a close, the book is cut back to the cap from the next session on. Set it to 0 to turn it off.

Settings and your selection are kept in this browser.

Strategy by strategy

One contract set per signal, halfway fills, fixed size on a $1,000,000 account with the 1x margin cap. Sharpe ratios are daily.

What compounding and the multiplier change

All seven strategies, $1,000,000 account, 1x margin cap, halfway fills unless marked.

Year by year

All seven strategies, one contract set per signal, halfway fills, fixed size on $1,000,000 with the 1x cap. The deepest drawdown is measured from the highest account value reached so far, so a fall that starts in December counts in the next year too.

• The short put carries the book. It earned 37% of the P&L. The short strangle has the highest Sharpe since 2022, and the short call adds little in dollars at one contract.
• A multiplier raises returns faster than Sharpe can hold them. Two contracts per signal at fixed size nearly doubles the P&L, but the drawdown grows from 23.9% to 43.2%, and the 1x cap has to cut the book back on 624 days.
• Compounding with leverage needs room. At two contracts with compounding, the cap binds on 1,012 days and the Sharpe drops to 0.66. With the cap off, the same book makes $3.6 million but needs up to 4.06 times the account in margin and falls 42.9% from its peak.
• Fills matter. Mid fills add 0.11 to the Sharpe over halfway fills, and bid/ask fills take 0.07 away.

Why you can trust this backtest

We tried to break this result before publishing it. Here is what we did.

• More than 5 million backtests. We replayed strategies, deltas, expiries, exits and entry rules on real quotes from 2018 to 2026. Most rules did not survive, including ones that looked good at first.
• We tested it against luck. Trading less cuts drawdown on its own, so we compared the rule with 1,000 random schedules that skip the same number of days. On standard trades across SPY, QQQ and IWM from 2018 to 2026, its Sharpe ratio beat 90% of them and its drawdown beat 83%. On SPY from 2022 to 2026, and on SPX weekly options over the same years, its drawdown beat more than 97%.
• We set the bar first. For each test we wrote down what would count as a pass before running it. Rules built on the VIX curve and on price trend failed those tests, so we threw them out.
• We report what it did not win. In 2017, a calm year with no selloff to avoid, its Sharpe ratio was 1.55 against 1.70 for selling every day.
• No hindsight in the signal. Each answer uses only the previous session's close. Trades enter at 09:35 ET at that day's quotes, fill halfway to the bid or ask, pay $1 a contract and are valued every day.
• Check it yourself. On the Alpha plan, replay any session since 2018 with the historical API and get the answer it gives, strikes included.
• Watch it live. A paper bot follows the endpoint every session and posts its results on our Discord, so the forward test is in public.

Ready to follow it? See plans, or watch the paper bot on Discord first.

Backtest a single recommendation yourself

This page pools each strategy type. To test one recommendation on its own, such as a single expiry or symbol, replay the sell-window endpoint for any past session with the FlashAlpha Historical API and trade the candidates it returns. Each answer uses only data available before that session's open, so a replay is a fair backtest.

Method

On every session where the sell-window endpoint said sell, each recommended trade opens one contract set at 09:35 ET and is held to the exit the endpoint specifies. The signal reads only the previous session's close, so no trade uses information from after its entry. Every trade comes from the engine behind our endpoint study, replayed day by day. Each trade needs only its own contracts: a put trade opens whenever its puts meet the rules (open interest of at least 250, a spread of at most 20% of mid, and a delta within 0.05 of the target), whether or not a call at the same delta does. Positions are valued at each session's 15:55 ET quote; a missing mark carries the last usable one forward. Overlapping positions add up. Commissions are $1 per contract on entry and exit. Mid fills trade at the midpoint, Halfway halfway to the bid or ask, and Bid/ask at the bid or ask; both non-mid fills add $0.01 per share.

Sizing. Contracts: each strategy trades the multiplier in contract sets on every signal. Target vol: each strategy is divided by the standard deviation of its daily P&L, the book is their sum scaled to the target yearly volatility on the account, and the multiplier scales the result. Compounding scales each day's positions by the account at the previous close over the starting account; an account that reaches zero stays there.

Margin is estimated in the Reg-T style: 20% of the short strike times 100 for a naked short option (the larger strike for a strangle or straddle), and the most the position can lose for spreads and butterflies. The cap never cuts the size you set in advance; it cuts back at a close where margin passes the cap, from the next session on. Brokers' requirements differ, and portfolio margin is usually lower for index options. Early assignment and margin interest are not modeled. Drawdowns are measured on the account in the builder, not on the margin the book uses.

Measures. Daily Sharpe is mean over standard deviation of daily P&L, every session, times the square root of 252; with compounding it is measured on daily returns. Max drawdown is the largest fall from a running peak of account equity. Worst day is the largest one-day loss as a percentage of the previous close. Peak margin is the book's largest estimated margin as a multiple of the previous close.

Checks. Every strategy's daily P&L adds up to the engine's own realized P&L within $1, and the total matches the day-by-day replay of the 13 listed trades to the dollar.

This is research, not investment advice. Results are hypothetical and before taxes; past results do not predict future ones. The skip call applies to new trades, so it cannot protect positions already open when a sudden shock hits.