So does one have parameters magically set themselves when the market changes? How is that any different than that same person eyeballing parameters if it's done in a predictable fashion?
Just to say you have not optimised and sure not many parameters but if the parameters are based on your market experience you have done what I call tacit fitting. Its extremely likely your backtest results are optimistic. GAT
I agree there is definitely an element of the parameters matching the market, but I am not looking for an auto quant bot. I am looking for something that replicates what I would do manually so I don't have to do it. Modifying parameters (or tacitly fitting) while live trading is part of it. If my backtest results are optimistic, then you would expect that I should not have the same results out-of-sample right? So pre-2010?
Pre-2010 without modifying parameters. Clearly there are periods where it's flat. Also this algorithm is purely "short."
Modifying algorithm to be both long and short basically inverting the same parameters. Was hoping that it would magically fix the flat areas, but didn't. Not a surprise - I think this is a regime change (is that the lingo???) These are from two separate sources btw, first one only has 2003-2014, second has 2001-2018
(Btw, I have no illusions that this is not going to work - I've had the "smooth equity curve" issue before)
What frequency is the trading? Intraday, daily, weekly? What prices are you using to enter trades in your backtest? Bid/ask, mid, last trade? Are you taking into account overnight interest if you are trading FX spot (if you have overnight positions)? Depending on the answers to these questions, your backtest could be plausible or not realistic at all.