How accurate are the backtest metrics

Most of Jesse's metrics module uses the popular Quantstats library behind the scenes and is pretty accurate but not always 100% accurate.

Notice that to calculate metrics, we need daily returns of the portfolio. For most of the metrics that is perfectly fine, but for a few such as the max drawdown you have to be careful.

Here's an example of when the max drawdown is not 100% what you would expect:

Let's say you buy an asset at $100 at the start of the day. During that day, the price drops to $50 which is a 50% loss. Then the price goes back to $90 (-10% loss) before the end of the same day. The max drawdown only knows about closing prices of the day, which means the max drawdown will only record the -10% loss. So can you now say that your strategy will never lose more than 10%? No.