Accuracy

It's amazing to be able to backtest a strategy on historical data. It shows whether your strategy will be worth running or not. But just how accurate are those backtest results?

Accuracy

When starting a backtest, you're able to select an accuracy. The lower the accuracy, the faster the backtest is. For a very rough idea of your strategy's perfomance, we recommend between 1 and 3 accuracy.

If your strategy has no real time actions, like real-time trailing or selling on a fixed %, you can always use 2 accuracy, as it will just reference the open and close of each candle.

When you select 4 accuracy and above, each candle gets split up in parts (more parts if the leverage used is higher). For every part of the candle, all values are recalculated and all paths in your strategy are determined. We have designed the system in such a way that the backtest always goes to the unfavorable side first: if you're in a short, the backtester will first do the parts the go towards the high of the candle, then go back down to the low. This is done to get the view always from the worst case scenario, which is what you want to be prepared for.

Accuracy 4-6 are the same in functionality, but the higher you go the more parts the candles are split up in. This ultimately results in more accurate results, but longer backtesting times.

  1. If possible, start backtesting on 2 accuracy until you have a solid base strategy set up. Stay on 2 accuracy as long as possible to quickly test more variations

  2. Move to higher accuracy to verify your backtest more precisely

  3. Do one final render on 6 accuracy to see the closest to reality view on your strategy

Disclaimer

Backtests can never 100% represent what happened: we don't store the real-time data, but use the candlestick data from the exchange. There can be differences in your advantage, or disadvantage.

Last updated