I found afterwards that the backtest doesn’t match reality if you use the server bars rather than tick data. I have no idea how that makes sense because it was calculating at bar close, so the bar and tick should have the exact same information at the point, but hey ho, that’s what happened… Testing with tick data and the performance wasn’t great… Back to the drawing board.
I’ve been watching @AmericanTrader thread for a while and was impressed by an indicator he used that is I believe doing a slow Hodrick Prescott filter onto a fast Hodrick Prescott filter and creating a channel. I’ve had success in creating channels out of moving averages, but they’re not very smoothed and can end up being too wide if you use a long period.
I managed to find the maths behind the indicator and replicated it into a Ctrader indicator. The problem with automating and this indicator is that I can’t see what the channel looks like at the time a trade was taken because it redraws every candle. So I changed the channel to only draw the HP filter value for the last candle and then I can see where price was relative to that at the close of every candle. This helps visualise the market and allows me to try and find patterns.
