How backtesting works

What happens when you run a backtest?

A backtest runs in four steps: your rules go in, Fincanva replays them through history, the replay produces an equity curve, and the curve is distilled into metrics.

1. Rules. The strategy you saved — its universe, its allocation, its risk conditions, its exit rules. A run only reads them, so running a strategy never changes it. See Backtest.
2. Historical replay. Fincanva steps forward through real market history and applies those rules date by date, resetting holdings to their targets on the cadence you set. See Rebalance.
3. Equity curve. The strategy's value on every simulated date, plotted as one line — the whole path it would have taken, not only where it ended. See Equity curve.
4. Metrics. The curve distilled into numbers: CAGR, Sharpe ratio, Max drawdown and the rest. See What every number on the metrics page means.

What does a backtest actually produce?

A backtest produces two things you read directly: an equity curve, and a record of how the holdings moved between rebalance dates.

The equity curve is drawn against its benchmark, so the shape of the result is visible and not only its final figure. Falls below a previous peak show up as dips in the line — the same falls the Max drawdown figure reports as a single number.

The second output is what happens to the holdings between one rebalance and the next. Between rebalances the actual weights pull away from the target weights — weight drift. On each scheduled rebalance date they snap back to target. Nothing corrects drift in between, so a mid-period portfolio is a drifted one, not the target mix.

What biases does a backtest remove?

A backtest removes one of the classic research biases outright, reduces several more, and leaves the rest to the person building the strategy. The nine biases Fincanva helps you avoid is the canonical account of all nine; the four below are the ones that bear most directly on a single run.

• Survivorship bias — removed. Delisted instruments are retained and index membership is resolved as of the simulated date, so a run is measured against the market as it stood on each historical date. The choice of universe and period stays yours — see How does Fincanva handle survivorship bias?.
• Overfitting — still yours. It comes from how you build the strategy, not from the computation, so no engine removes it. See How does Fincanva reduce overfitting?.
• Look-ahead bias — see How does Fincanva reduce look-ahead bias?.
• Cost-ignoring bias — see How does a Fincanva backtest account for trading costs and taxes? and Simulation assumptions.

Limits and edge cases

• A backtest covers history up to the latest available market close. New market days keep arriving, so a finished run ages. A strategy you have marked Live is re-run for you as new data lands; one you have only saved is not — its run status turns Out of date after a data refresh and stays there until you run it again.
• A short history shortens the run. A backtest can only cover the period for which the chosen instruments have data, so one instrument with a short history limits how far back the whole run can go.
• Rebalance dates are discrete. Weights are reset only on the cadence you set, never continuously, so drift between two rebalance dates is part of the result rather than a flaw in it.
• A run can end early. If a strategy's value collapses far enough, the run stops trading — see Bankruptcy rules.
• Evidence is not a promise. Removing a bias makes the past more honest; it does not make the future more certain. Read every result as "what these rules would have done", never as "what these rules will do".

Backtests show what would have happened — not what will. Fincanva provides no financial advice — see Is this financial advice?.