don't forget, in _street smarts_, they talk about the 'turtle soup' setup, which is definitely fading a certain system (in this case, the 20 period breakout) with some parameters. but they have stops in as well...preventing the bad tail.
Amigasearch, It seems that you are bent on trying to fade a losing system. If that is indeed the case, then by all means go for it or, better yet, test it. That will be time far better spent than accumulating validation or support for the idea here. Just know going in that finding a suitable system to fade will be as difficult as finding a good system to trade, for the reasons I indicated earlier. Good luck.
hypo, i must have missed a beat to be honest i didn't even look at posters names, i just skimmed the thread and noticed nobody had mentioned transaction costs, so i thought i'd drop it in in case the original poster hadn't encountered this yet (even though it's a pretty obvious answer). misrepectfully yours v
Vanilla2, yes, my equity curve factors in the commisions ( i am using backtesting software I designed which allows it - not sure what other products out there allow). Regardless of my results, they are in there. All of this talk has opened my eyes and ears to this concept of trading something I dont know a damn thing about. Seems the only thing that makes sense is what works (what is profitable over time) rather then what I think should work. Hypo, again, I am not going to trade this thing - just testing it out. If i do end up trading it, it wont be for another 6 months (test it real time) and then I will let you know. One other thing and off topic - Hypo, I ordered Babcocks book on your reccomendation. Is there anything about Z scores in there, and if not, can you direct me to a source? I have statistic books that dont cover this....
...I do not anwer questions, I only pontificate unsubstantiatedly. Z-score is the criterion my SW (E-Signal) uses as a suggestion of statistical significance. I think it is Student's T test converted to standard deviation space (someone who knows will correct me if that's wrong). For example, if I remember from one of my system tests, a Z-score of 1.48 sigmas implies a 14% probability of being a random result. That's typical of what I get. Do a Yahoo! search on "Z score" and you'll find it, that's what I did. I know next to nothing of improbability and sadistics, which is probably a good thing considering how bizarre my systems are. I only know I BELIEVE, which is the best attitude for trading a system.
good systems do not predict price. they simply place you in a position that is the most efficient for the next anticipated move. your question indicates that you do not understand this yet.
Transaction costs are not to be underestimated. Included in that is also the bid/ask spread for systems that execute at market or that test on tick data where bid/ask is not available. Vanilla makes a good point that is particularly applicable to intraday trading systems. I can't say how many times transaction costs have been the difference between a very profitable system and a mediocre system. I've tested systems that lost money when tested normally AND when testing the reverse, so reversing the signals isn't a sure bet... But to answer the question, yes it is a viable method of coming up with a system. Sometimes the counterintuitive is what ends up working. And my editorial for amigas, how can you say you don't understand the opposite of your system? If you understand your system then surely you can understand the opposite, and even WHY the opposite of your original system "works". But I have to say that in system development, "why" is the least important to understand of, "what, when, and how much" because "why" is often an after the fact kind of thing.
This may be true for a thinly traded instrument but for something as "viscous" as ES (or NQ), the absence or presence of bid/ask has no bearing on the transaction costs when testing on tick data - they are the same in both cases. I say this because it is a simple exercise to determine the bid/ask in ES from the tick data.