Planned: not available yet

Backtest audit

Are the data, assumptions, and calculations in my trading test correct?

A backtest is a test that applies trading rules to past price data. It calculates what the rules would have done. An error in the data or in the code can make a weak rule look strong.

This audit will check one backtest. A person will examine the price data, the assumptions about fills and fees, and the calculations. The written report will list each check, its result, and the evidence for it.

The audit will find errors and unsupported assumptions. It does not show that a strategy will make money. A correct test of past data does not predict future results.

Who it is for

Traders who built a backtest with an AI tool: They have a result but cannot check the code, the data, or the assumptions themselves.
Developers who want a second check before they trust a result: They need an independent examination of the data and calculations, with a written record of each check.
Traders who bought data and wrote their own test: They need to know whether the data has gaps, duplicates, or time errors that change the result.

What you will get

A written report on one backtest, with a list of each check and its result.
Data checks: time order, duplicate rows, invalid prices, time zones, and gaps in the records.
A check of the assumptions about fill prices, fees, and the timing of orders.
A check that the test includes expired contracts where the rules need them.
A check of the calculations for returns, position sizes, and costs.
A list of the problems found, with the evidence for each problem.

What it will need from you

The trading rules, in words or in code.
The price data that the test used, or a description of its source.
The test results that you want to check.
The assumptions about fees, fills, and order timing.
The dates and contracts that the test covers.

How it will work

Describe the backtest and the result that you want to check.
Agree on the scope: one strategy, one data period, and one report.
Send the rules, the data or its source, and the results.
The data, assumptions, and calculations will be checked.
Read the written report and its list of findings.

Limits

An audit does not show that a strategy will make money. It checks the test, not the future.
Each audit covers one defined strategy, one data period, and one report.
Minute bars cannot show the order of trades inside a minute or a price that you could have traded at. The audit will mark results that depend on such details.
Separate option legs can trade at different times. Adding their recorded prices does not give a spread price that you could trade at.
The audit can compare your data with the archive only where the archive has records. Coverage will be checked for each audit.
The report is a record of checks and arithmetic. It is not investment advice.

Common questions

Will the audit tell me if my strategy is good?

No. The audit tells you whether the test has errors in its data, assumptions, or calculations. A test without errors can still describe a strategy that loses money.

What kind of errors can a backtest have?

Common examples are gaps or duplicates in the price data, wrong time zones, and missing expired contracts. Others are fills at prices that the records do not support, and missing fees.

Can you check a backtest that an AI tool wrote?

Yes. The audit checks the rules, data, and calculations, whatever wrote the code. You must be able to supply the code or a clear description of the rules.

Do I need to use TradeRecount data for my backtest?

No. You can send your own data or name its source. Where the archive has records for the same contracts, the audit can also compare the two.