Glossary · Backtesting & Analytics
Forward Testing
Last updated:
Forward testing is the practice of running a trading strategy on new, unseen market data — after the backtest is done — to check whether its performance holds up outside the period it was built on. It is the out-of-sample exam that a strategy has not studied for. The most common form is paper trading the strategy live for weeks; a stricter cousin, walk-forward optimization, automates the same idea across history.
The distinction from backtesting is direction: a backtest looks backward over data the strategy may have been tuned on; a forward test only sees data that arrived after the rules were frozen. That freeze is the whole point — no tweaking parameters mid-test.
Example
A trader tunes a supertrend strategy on NIFTY data from 2021–2024 and gets a backtest showing 28% annual returns. She freezes the rules and forward tests through early 2025. Three months later the forward equity curve is flat while the backtest period would have shown +6%. Nothing “broke” — the parameters were simply fitted to the past. A robust strategy will not match its backtest exactly, but a forward test that looks nothing like the backtest is a strong signal of overfitting. The figures are hypothetical.
Why it matters
Forward testing is the cheapest honest test in systematic trading. Backtests can lie through curve-fitting and lookahead bias; forward results cannot, because the data did not exist when the rules were written. A few weeks of forward testing routinely saves months of live losses.
In INDfolio AI, strategies move from backtest to forward test with the rules locked, so out-of-sample results stay honest — see Forward testing.