After making some adjustments to the Advance/Decline indicator, I discovered some astonishing results. I posted the equity curves of those results here, but noted that the results seemed to too good to be true. I had some time this weekend to go through the trade-by-trades and I found an error which is allowing the the system to have more than one open position, even though there is only enough cash for one position. I believe that this error is what juiced the results.
Too often, equity curves that look to good to be true, usually are.
Very true. Thinkorswims Prodigio did the same thing, I was excited to work towards results of +220%/yr only once I checked the trades I noticed it was trading $600k Instead of the allotted 100k.
Exactly. I’m going to send the code in to AmiB support. In the default mode, this should not happen. The portfolio should not be holding more than one position, when in fact it may be holding 2 (and I guess at times it could hold more). I’m sure it is something that I’m doing. I’m not blaming AmiB. But to my knowledge, none of my code is telling it to use another backtest mode besides the default. Of course I can add 2 extra lines and keep this from happening, but my point is that it should not be happening if I am in default mode.
Wood, Which data provider do you use for AmiBroker? could that have something to do with it?
As you well know, this has happened dozens of times to anyone that has been building mechanical systems long enough WS, no sweat. Regards, JP
Were you using the ExRem() function to remove excessive signals? Not sure if that is an issue; just a thought.
Hi Chris, yeah, I am now, but not initially because in the default mode, AmiB is not supposed to allow redundant/multiple signals. I am not using any custom backtest procedures- only the default backtester and really no complex code except for some looping to create the composite. All my settings are for 1 position only. Even if I switch to a boilerplate, set to allow one position only, I still get the multiple entries. Strange.