What a backtest is, and what it cannot tell you
Caliente Signals · Updated
A backtest runs your strategy over candles that already happened and records every trade it would have taken. You get an equity curve, a full set of performance figures, and the trade-by-trade table behind them.
Everything the strategy carries goes in: both rules with all their conditions, the match-all or match-any choice on each, your risk controls, and your cost assumptions. Windows are 90, 180 or 365 days on your chosen pair and timeframe, with the 365-day window on Pro. Backtests themselves are unlimited on both plans.
What makes it honest
Decisions come from closed candles, and fills happen at the next candle's open, which is a price that genuinely existed after the decision. Nothing that happens at candle 400 changes what the engine did at candle 300. Your fee and slippage assumptions come off every trade. A position still open at the end is marked to the last close rather than quietly dropped.
And it is the same engine that sends your live alerts. There is no separate backtest mode with friendlier rules, which is the whole no-repaint claim in one sentence.
Where a good-looking result comes from
A window of candles is one sample of one market over one stretch of time. That limit is obvious and people account for it.
The subtler one shows up while you tune. If you try forty variations of a rule and keep the one with the best curve, some of what you are looking at is the rule and some of it is the luck of that particular window. The two are not separable from inside a single result.
Here is an example, using a rule anyone could build in five minutes.
Worked example
LTC/USDC on 1h, over the 365 days to 30 August 2026. Entry RSI(14) crosses above 30, exit RSI(14) crosses below 70, at the default 0.1% fee and no slippage. Only the period changes between the rows.
| Setting | Return | Trades | Win rate | Drawdown |
|---|---|---|---|---|
| RSI(12) | +1.73% | 41 | 58.54% | 36.50% |
| RSI(14) | +8.52% | 31 | 70.97% | 39.81% |
| RSI(16) | −10.29% | 21 | 61.90% | 42.89% |
An example run, not a strategy we suggest. Buying and holding over the same window returned −56.79%.
One notch either side of the middle setting, and the result hands most of itself back in one direction and changes sign in the other. Nothing about the idea changed. Only the period did.
So prefer rules that survive being slightly wrong. Nudge the period, nudge the threshold, run the shorter window as well as the longer one. A rule whose result collapses when 14 becomes 16 was fitting noise, not describing a market.
Not every rule behaves this way. The same three-settings check on a price-versus-SMA crossover, on a different pair and timeframe, produced three results within about a percentage point of each other. Some rules are stable across neighbouring settings and some are not, and running the check is the only way to know which one you have.
A backtest is a record of what your rules did over a past window. It is not a forecast, and nothing in it is a prediction. Reading the numbers it gives you is its own short article.
Read next
Write the rule. We'll watch for it.
Unlimited backtests on the free plan, and no card to start.
Caliente Signals is an alert tool, not an investment advisor. Nothing on this page is financial advice, and past performance is not indicative of future results.