Backtesting Options Strategies: Overfitting, Hidden Biases and Costs

Reviewed by Leon Moyer - Published October 9, 2026 - Last reviewed October 9, 2026

Backtesting is applying a strategy's written rules to past market data to see how the trades would have turned out. It is the cheapest test a strategy can get, and also the easiest to fool yourself with, because every number in it is hypothetical.

The SEC's investor bulletin on performance claims warns that back-tested results are hypothetical rather than a record of real trading. This lesson is about reading your own backtest with the same caution.

What can a backtest tell you about an options strategy?

A backtest can answer narrow questions well:

• How bad did it get? The largest loss on one trade, the worst run of losses in a row and the deepest drop from a peak are often more useful than the average result.
• How often did the rules trade? A strategy that triggers twice a year gives you little evidence and a long wait.

A backtest cannot tell you the strategy will work from here. Markets change, and the history you tested was only one path prices could have taken.

Why are options harder to backtest than stocks?

A stock backtest needs one price per day. An options backtest needs a quote for every strike and expiration you might trade:

• Bid and ask, not one price. You buy near the ask and sell near the bid. A test that fills at the midpoint assumes a better price than the market was showing.
• Implied volatility. It rises and falls on its own. A test that models option prices from a fixed volatility misses the swings that decide many option trades.
• Assignment and dividends. A short option can be assigned early, especially around a dividend. Many simple backtests never model it.
• Expiration dates. Weekly expirations, and on some products daily ones, were added over the years, so a strategy built on short-dated expirations may not have been tradable for the whole test period.

Whatever tool or data source you use, check which of these it models before trusting its output.

What is look-ahead bias?

Look-ahead bias is using information in a backtest that you would not have had at the moment of the trade. It creeps in by accident:

• Deciding to enter "today" from today's closing price, then filling the trade at today's open, hours before that close existed.
• Picking strikes from implied volatility data that was published after the close.

The fix is a time stamp on every rule: what you knew, and when you knew it. If the entry signal uses a closing price, the earliest fill is the next session.

What is survivorship bias?

Survivorship bias is testing only on what survived to today. If you test a strategy on the stocks in an index right now, you leave out every company that fell out of it, was acquired or went bankrupt during the test period. Those are often exactly the cases where a bullish options strategy lost the most.

The fix is a list of what was tradable at each date in the test, including the names that later disappeared.

What is overfitting in an options backtest?

Overfitting is tuning rules until they fit the past so closely that they describe noise instead of anything that repeats. Options backtests are especially exposed because every trade has so many dials:

• Too many parameters. Entry signal, strike or delta, days to expiration, profit target, loss exit and a volatility filter. Five settings for each of just five of those dials is 3,125 combinations, and the best of 3,125 can look excellent on any stretch of history by chance alone.
• Curve-fit thresholds. A rule that works at a 47-day average or a 0.23 delta, but not at 40 days or 0.20, has probably found a quirk of the test period.
• Data snooping. Every idea you test on the same history, and every tweak made after seeing results, uses that history up. The hundredth idea tested on one data set proves less than the first.
• Small samples. A rule that fired 15 times can show a great average by luck. Course 3's How Many Trades Do You Need? lesson shows how wide that uncertainty is.

How do you guard against overfitting?

1. Write the rules first, and log every try. Fix them before the test, as How to Write Testable Trading Rules You Can Backtest suggests. A winner picked from 200 variants needs a far higher bar than the only rule tested.
2. Hold back data. Build the rules on one period, then check them, unchanged, on a later period they were not built on. This is out-of-sample testing. Change the rules after looking at that period and it is no longer out of sample.
3. Walk forward. Repeat the split in steps: build on years 1 to 3 and test on year 4, then build on years 2 to 4 and test on year 5. Walk-forward testing shows whether the rules keep working as conditions change, not only on one lucky split.
4. Check the neighbors. A rule that works with a 50-day average and still roughly works at 40 or 60 days is more believable than one that only works at 47.
5. Demand enough trades. Treat a result from a few dozen trades as a hint, not evidence.

How much do trading costs change a backtest?

Costs are where most hypothetical results shrink first, and multi-leg options strategies pay them on every leg, in and out.

How should you use a backtest result?

As a filter, not a forecast. A careful backtest that loses money has saved you real money. One that makes money earns the rules the next test: later data, then paper, then live at the smallest size. Course 3's module Finding Your Edge from Your Journal covers judging results from your own live trades.