QUANTUS LEARN
The Limits of Backtesting: What a Backtest Cannot Tell You
What testing on historical data is good for, and how to read it alongside overfitting, trading costs, data bias and operational constraints.
A backtest is a tool for rejecting hypotheses.
Looking good in the past is not enough to predict future performance. Where testing earns its keep is in exposing a strategy’s weak conditions and showing in which market environments its results change.
Check data bias and costs separately.
Testing only the securities that survived to today, or using information that was not yet available at the time, can inflate results. Trading costs, whether orders could realistically have been filled, missing data and the treatment of delisted securities should all be disclosed alongside the results.
Ask whether the rule can actually be run in practice.
If you have not decided what the strategy does when data arrives late, or when order size and liquidity become binding, a gap opens between the test results and live operation. Before deployment, review exceptions and stop conditions on their own.
CAVEAT
Limits to keep in mind
- A backtest is a simulation based on historical data and does not guarantee future returns.
EVIDENCE
Sources and further reading
PRODUCTS
See it in the product
STRATEGIES
