These are the questions traders ask most often once they start backtesting seriously. The answers are deliberately plain and conservative. None of them promise that a passing test will make money — backtests do not guarantee future results — but they do explain how to run a test you can actually trust.
How much history do I need?
Enough to cover several different market conditions — trending, ranging, calm, and volatile — not just the stretch that happens to flatter your idea. A strategy tested only through one kind of market has not been tested; it has been auditioned. More important than raw length is that the history spans regimes the strategy will eventually have to survive.
How many walk-forward folds is enough?
Enough that one lucky window cannot carry the result. A single out-of-sample fold is one sample; a handful of rolling folds starts to show whether the edge is real or whether you simply found a good window. The point is not a magic number but consistency: an edge that holds across many unseen windows is far more believable than one that shines in a single run.
Are fees and slippage included?
They must be, or the test measures a market that does not exist. The realistic cost model charges fees, the spread you cross, and a slippage buffer inside the test, so a strategy has to beat its full friction to score positively. A backtest run without costs is a flattering fiction, not evidence.
My backtest looks perfect — is something wrong?
Probably. A flawless equity curve is usually a warning, not a reward. It often signals overfitting, look-ahead bias, or missing costs rather than a genuine edge. Real strategies have rough patches. A curve with no pain in it should make you more suspicious, not less.
Does a passing backtest mean I can go live?
No. A pass is permission to paper trade, never permission to go live. Paper trading against real prices is the next gate, and only a sustained, cost-clearing paper run earns a look at the paper-to-live decision. The sequence exists precisely because backtests miss things the live market does not.
Which metric should I trust most?
No single number tells the whole story. Read reward and risk together — Sharpe alongside the worst drawdown — and confirm both hold across out-of-sample folds rather than on one tuned window. A high headline return earned through a brutal drawdown can be untradeable in practice.
For the methods behind these answers, read walk-forward testing explained and backtesting myths that cost traders money. To read results honestly, see how to read Sharpe and drawdown together. Nothing here is financial advice.