A very large number of strategies have values (x) in the millions for:
Exit after x bars
Trailiing Stop x pips
Stop Loss to Break-Even x pips
Trailing activation x pips
This is also true when utilizing multiples of the ATR where appropriate for the Exit Types...for example Move SL to BE = x * ATR (y), where again x is in the tens of thousands or millions.
In all these cases, such values make absolutely no sense as far as I can tell. And the effect of such large values is the same as turning these Exit Types off.
Curiously the only one I haven't seen this for is Stop Loss (which is always set as required in my builds). The other exit types are always on, but not required. Just a thought. I dont have time to experiment further for you I am afraid, but safe to say the extreme values/error are there.
Hi, I already reported this in sq4_3596 like a month ago and it hasn't been fixed yet.
To me this is very risky and serious as there's no way to put any strategy in live.
Please fix asap.
I have 250 million generations to date, of these only 12 so far have passed all the rigorous robustness tests I have thrown at them. Going in and reading the Pseudo code for these 12, all of them except one suffer from the problem of extreme values on the exit types, and so are now only fit for the bin - or an alteration of the code and running the full set of robustness tests again.
250 million = quite a lot of processing time, actual time and electricity costs only to discovering this error at the end.
It needs fixing though - the fact that the ranges are not respected at all for the Exit Types, be they the default ones, or custom reduces diversity in the population if the extreme numbers are effectively turning off the options of other exit types. Leads to bloat. Even if those exit types are checked and turned on, the frequency with which they are actually used in generation is significantly reduced.