Hi,
If I select a previous period (e.g. Jan-1 2017 to Jan-1 2019) when the graph is in historical mode, it will show this period on the graph, so you think all is OK. But there is a bug: it keeps using the more recent data (Jan-1 2019 to current) as well.
If you start optimizing it will include this data without you knowing it (so the results shown are incorrect for this period!). The result shown on the ‘detailed report’ also includes this more recent data and is also incorrect.
How to recognize it: if you follow the equity line displayed on the graph, it will show a spike or drop at the end. This is the change from the end of the graph to the end of the data used in calculating (e.g. jump from Jan-1 2019 to June-22 2021). The simulated future orders will also show up (combined) on the last bar with a future date on them.
When it happens: it does not happen all the time. I have the impression it occurs after you first run a backtest on recent data and then change the graph to older data. It is independend of the coding or which strategy I use. The first time I noticed this was in Q3 last year. I mainly work with 1 minute strategies on NQXXXX with a premium account. You should be able to replicate this with any kind of strategy.
How I solve it (quick fix): I use ‘date > datamax’ as exit condition in the Strategy code. However, as it still uses the more recent data the optimisation calulation time is considerably longer than it should be.
It would be nice if there is another way to solve this, or to include this in a version update.
Best regards
.
