Backtesting a SPY Put Spread: Our Paper Record

Almost every article about selling put spreads quotes a win rate. Few say where the number comes from. This one starts from the opposite end: here is what a real, short, imperfect record looks like, and the questions to ask before you trust any backtest, including ours.

How to read the FIREDesk numbers in this article. They come from our paper-trading record (14 May to 9 Oct 2026): no real capital was used, option prices are estimated by a model rather than taken from real fills, and 13 of the 102 logged signals use credits calibrated from historical data. The model has no sudden jumps in volatility, so real fills and real stops would be less favourable. On 12 October 2026 we also fixed a calculation error that had recorded stop-losses the market prices did not justify; the figures here are the corrected ones. It is a short, early record and it is not a forecast. Full detail is on the performance page.

What a backtest can and cannot tell you

A backtest replays a rule set on past data. For a put spread the rules are simple to state: sell a put near a chosen delta, buy a put 10 dollars lower, close at a profit target or a stop, otherwise hold to expiry. The hard part is the data and the assumptions.

• Option prices. Historical option quotes are expensive and patchy. Many backtests estimate prices with a pricing model, which smooths out the real bid/ask spread and the jumps that happen in fast markets.
• Fills and costs. Real fills are worse than mid prices. Commissions are small but they matter on a trade that makes a few hundred dollars.
• Overlapping trades. If you open a position every day, many of them live through the same market move. 60 trades are not 60 independent tests; they can behave like a handful.
• Open trades. A record that only counts closed trades can look better than the full picture, especially when winners and losers close at different speeds.

What FIREDesk actually has

We do not have a multi-year backtest built on real option quotes, and you should be wary of anyone who claims one without showing the data. We have two things, and they are not the same:

• A simulation of 2022 to 2024 on the performance page, based on weekly SPY options, with no slippage assumed. It reports a win rate of about 78%.
• A live paper-trading record since 14 May 2026, using longer expiries (roughly 45 to 126 days). This is the record described below.

The two use different expiries and different periods, so they should not be compared one to one. The live record is the one that follows the current rules, and it is the one we weigh more.

Why the headline win rate misleads

An 83% win rate (70 of 84) is easy to quote and easy to over-read. FIREDesk closes at 50% of the credit received, and it stops out when the spread costs 1.5 times the credit, which caps the loss at 50% of the credit instead of letting it run to the full spread width. With a one-to-one payoff on the credit, a win rate near 50% is the break-even point before costs. Here the prices are modeled without sudden jumps in volatility and many trades overlap in time, so the 83% and the profit factor (gross profit divided by gross loss, 7.0 here) are a ceiling, not an expectation. Real fills and real stops would be less favourable.

The same record, split two ways

Averages hide structure. Two splits of the same 84 closed trades:

• By month of closing: May +$186 (1 trade), June +$609 (16), July +$1,555 (17), August +$6,038 (23), September +$1,181 (7), October up to the 12th +$4,198 (20). August alone supplies about 44% of the total. Income from this kind of strategy is lumpy, not smooth.
• By expiry length: trades opened with about 60 days to expiry closed 29 winners and 14 losers (10 still open). Trades opened with about 120 days closed 41 winners and no losers so far (8 still open). The longer trades had less time to prove themselves, sit further from the price (short strike 8.8% below versus 6.8%) and were opened under different market conditions, so we do not read a conclusion into it beyond this: all 14 stops had the short strike less than 8% below the price. It is exactly the sort of pattern a single win-rate number would have hidden. We look at it in What is the best credit spread strategy?.

A short checklist for any backtest

• Does it say whether prices are real quotes or modeled?
• Does it include costs and slippage?
• How many trades are there, and how many were open at the same time?
• Does it show the worst period, not only the average?
• Does it split results by market regime or expiry, or only report one blended number?
• Does it disclose how many trades are still open?

For the rules behind the signals see the complete SPY bull put spread strategy guide, and for how the 50% take profit and the stop interact read the 50% take-profit rule and the 1.5x stop loss.

Educational content, not financial advice. Options trading involves significant risk of loss, including the loss of the whole amount at risk on a trade. Past and simulated results do not guarantee future performance.