Methodology
This page explains where the lab's data comes from, how Quick Backtest calculates results and what it leaves out.
Where the data comes from
Kalibre does not publish market data. Each series in the lab comes from one of these sources, and the source is shown on the series:
| Source | Label | Notes |
|---|---|---|
| Sample series | EXAMPLE | Generated by a seeded random process, for getting to know the lab. |
| Your file | YOUR DATA | A file you exported from MT5 or elsewhere. Your browser reads it locally. |
| Kalibre Bridge | MT5 and broker name | Price history from your MT5 account, end-to-end encrypted (security). |
| Alpaca, Tradier | Broker name | Price history requested by your browser with your own API key. |
Candle times
Candle times from MT5 are your broker's server time, the time you see on the MT5 chart. We do not convert them to UTC: converting past candles with today's offset would shift them by an hour around daylight-saving changes. The lab shows the server time's offset from UTC separately.
How Quick Backtest works
Quick Backtest runs your rule in your browser with the same rules as Kalibre Engine in MetaTrader 5:
- Rules are evaluated at the open of a new candle, using the last closed candle. Crossover rules use the last two candles. A future candle is never used.
- Exit rules are evaluated before entry rules. A signal in the opposite direction closes the open position.
- If both directions signal on the same candle, no trade is opened.
- Lot size comes from balance, risk per trade, the volatility scale and the stop distance.
Approximations:
- The order of prices inside a candle is unknown. If both stop and target are touched in the same candle, the stop counts; this is a cautious assumption.
- The trailing stop is updated at candle close.
- Spread, commission and slippage are optional and zero by default. Swap is not included. There is no news filter.
Volatility models
The volatility lab offers three measures. EWMA weights recent returns exponentially (the RiskMetrics approach). Yang-Zhang uses open, close and intraday range together. GARCH models volatility clustering and its return to a long-run level; its parameters are estimated on your data by maximum likelihood.
How to read results
- Read return together with max drawdown.
- A result from a small number of trades may be luck.
- Testing many settings and keeping the best one fits the rule to the past. Test the chosen setting on a period you did not use.
- Past results do not predict future results.