If I was backtesting an es system which used limits, I would never count a trade as filled until the price trades through. Man those lines are getting long.
Just curious - wouldn't the lack of follow through seem to cut both ways in terms of projecting winners vs. losers?
For example, I've been continuing to track Quah's SVS system, and in my record keeping, it seems that by following a strict "if it hits the limit, I count it" approach, the system is taking both losses and wins that may be questionable in real life. I did consider using a .25 follow through buffer, but I figured that would negate the times in real life when you would have been filled or stopped out precisely at the limit. So, what I'm trying to say - rather clumsily - is that I think both approaches (follow through/no follow through) are not going to mirror real life, and I'd think that it would be more fair to say, if it did actually hit that price, I will count it. Otherwise, it seems more arbitrary, at least in my way of thinking.
When I've been testing, I've used worst case- that is, filled on the stop if it trades there at all (if using hard stops with IB this is the case 99% of the time), and not filled on the limit entry or limit exit unless it trades through. It makes a huge difference in scalping systems. One system I was testing goes from about $2,000/contract in three months to $18,000/contract over the same time period, depending on how you count the fills. I'd rather be pleasantly surprised with real world testing then horribly dissapointed. -JPB
That makes a lot of sense - and I think if I were preparing to actually trade a certain system, I would do the same. Right now I'm just trying to "pre-screen" this particular system to see if it's even worth looking at in the kind of detail you describe. But I think you're right overall.
You got it JPB, That's exactly right. Hey, you want to sell that scalping system? That's all I do is scalp, and I know darn well there are computers out there doing exactly what I and about a half a million other traders are doing. Just curious, what's the max drawdown? You have it broken down into say max profit max drawdown per 100 r/ts? Oh by the way, just to make matters worse, there's also the chance that if you were in there with any size it wouldn't have traded through!
Okay- This isn't the same exact system I was thinking of, I pulled those numbers from memory. This is one I was looking at that shows the difference pretty well. Scalp1 is the results from counting the fill only if it trades through, Scalp2 is the results from counting the fill if it hits the target. Quite a difference, eh? The time frame was July 22 - Oct. 11. And sure I want to sell this system, just be sure to let me know how it works real-time! -JPB