What an options backtest can tell you (and what it can't)
Xavi ·
A backtest is a replay with rules.
You give it a strategy, historical market data, and a set of assumptions about execution. It steps through the past and shows what would have happened inside that model.
That can teach you a lot. It can also give you false confidence with impressive precision.
The distinction matters. A result such as “$18,462 net profit” looks like a fact. It is really the output of dozens of choices: which quotes were available, how contracts were selected, when signals were evaluated, where orders were assumed to fill, what commissions were charged, and how positions were handled near expiration.
A useful backtest helps you investigate those choices. It does not certify the strategy.
The questions a backtest can answer
Start with behavior, not profit.
Did the strategy enter when its rules said it should? Did it choose the intended expiration and strikes? Did the stop behave correctly? How often did no suitable contract exist? What happened on fast or volatile days?
Those questions sound basic. They catch a surprising number of problems.
A trader may describe a setup as “sell a 15-delta put spread at 10:00.” The test reveals that the nearest contract often had 11 or 19 delta, or that the requested expiration was unavailable on certain dates. That is useful information even if the P/L is disappointing.
Once the mechanics look right, historical testing can help you examine:
- trade frequency and holding time
- the size and timing of wins and losses
- drawdowns and recovery periods
- performance in different date ranges
- sensitivity to slippage and commissions
- whether a few unusual trades produced most of the result
This is evidence about the tested period. It is not a forecast.
What the backtest cannot recreate
Historical options quotes do not contain the full experience of placing a live order.
The simulator cannot put you back in the actual queue. It cannot know whether your limit order would have filled before the market moved. Multi-leg orders add another problem: a displayed net midpoint does not prove that the complete spread was executable there at your size.
Minute-level data also compresses what happened inside each minute. That is far more useful than daily bars for many intraday strategies, but it still is not a tick-by-tick recording of every event.
Then there is the trader. A backtest never gets nervous after four losses. It does not cancel tomorrow’s entry, double the size to recover, or shut down the system halfway through a drawdown. The rules keep running because the simulation has no rent to pay.
A good result can still be a bad test
Suppose a strategy looks excellent after you try 200 combinations of entry time, delta, wing width, profit target, and stop loss.
You may have discovered a stable relationship. More often than traders care to admit, the winning combination is simply the one that matched the accidents in that sample.
The backtest did its arithmetic correctly. The research process was the problem.
This is why reproducibility matters. Save the exact strategy version, dates, starting assumptions, fill model, and commission settings. If you cannot rerun the same experiment, you cannot tell whether a later improvement came from the strategy or from a quiet change in the setup.
Treat the result as a claim to challenge
After the first run, do not ask, “Is this profitable?”
Ask what would have to be true for the result to deserve attention.
Try worse fills. Add realistic commissions. Look at the largest losses one by one. Split the history and leave a later period untouched while you build the rules. Check whether nearby parameter values behave reasonably or fall apart.
The aim is not to make the equity curve survive every imaginable attack. It is to find out how much of the result is sturdy and how much depends on a friendly assumption.
Where StratVerra fits
StratVerra lets you describe an options strategy in plain English or build it without coding, then backtest it on hosted minute-level historical options data. You can model slippage and commissions and preserve the strategy for later paper or live runs through supported brokers.
That shortens the mechanical work. It does not remove the research judgment. You still decide what the rules mean, which assumptions are fair, and whether the evidence is strong enough to keep investigating.
In the next article, we will turn a loose trading idea into rules a simulator can actually test.