Forward testing
Forward testing: prove your strategy on data it has never seen
Last updated: · Reviewed by the INDfolio AI research desk
In short
Forward testing on INDfolio AI runs a strategy on new, unseen NSE market data as it arrives — with virtual capital, the same execution engine as live trading and the full Indian cost model — so a backtest that was tuned to history either keeps performing or gets found out. It is the final validation stage after backtesting, walk-forward analysis and Monte Carlo simulation, and the strategy moves to live deployment unchanged.
A backtest can be flattered — tuned, consciously or not, to the very data it is scored on. Forward testing removes that flattery: the strategy runs on new, unseen market data as it arrives, and either keeps performing or gets found out.
On INDfolio AI, forward testing is built into the workflow: freeze the strategy after backtesting, let it run forward on live NSE data with virtual capital, and compare the forward results against the backtest before any real money moves.
Out-of-sample by construction
Forward data did not exist when you built the strategy — the one test that cannot be curve-fit.
Side-by-side with the backtest
Judge whether forward performance matches the backtest’s character: win rate, drawdown, expectancy.
Zero capital at risk
Forward tests execute with virtual money on live data — the cost of finding out is time, not losses.
Cloud-run and monitored
Tests run server-side with alerts on entries, exits and risk events. No terminal to babysit.
Specification
Forward testing on INDfolio AI — specification at a glance
| Dimension | Forward testing on INDfolio AI |
|---|---|
| What it is | Running the strategy on new, unseen market data as it arrives, with virtual capital |
| Data | Live NSE prices; out-of-sample by construction |
| Engine | Same engine as backtest, paper trading and live deployment |
| Costs | Full Indian cost model and slippage applied to every simulated fill |
| Comparison | Forward results shown side by side with the backtest over the same dates |
| Recommended duration | Several weeks across different market conditions (see the guide below) |
| Availability | All plans (within paper-strategy limits) |
| Coding required | None |
Backtest, forward test, walk-forward: what each one proves
A backtest proves the strategy would have worked on the past — necessary, never sufficient. Walk-forward analysis re-optimises on rolling windows to check the edge is stable across regimes rather than tuned to one period. Forward testing is the final, simplest filter: fresh data, frozen rules, no second chances.
Strategies that pass all three have earned real capital. Strategies that ace the backtest and fail forward are the curve-fits the process exists to catch — and catching them on paper is the cheapest lesson in trading.
How long to forward test
The answer is in trade count, not calendar time. An intraday system producing several trades a day can prove itself in weeks; a positional system trading twice a month needs longer or must lean more heavily on walk-forward evidence. The threshold that matters: enough forward trades that the comparison to the backtest is statistics, not anecdotes.