I find it very interesting that a few QQQQ patterns in an articel by Michael Harris published by Active Trader in 2002 still perform very well, as a matter of fact all of them. I read the details in the following article: http://www.tradingpatterns.com/About_Us/articles/article3/article3.html I'm thinking of allocating a significant portion of my capital to a QQQQ trading system of this sort and for this reason I am asking your opinion on what you think may go wrong and these patterns stop performing well. Ron
This is some kind of a 'white box' system in that it generates some code for your platform based on proprietary short term patterns (looks a little like the Ross Hook pattern) . http://personal.fidelity.com/myfidelity/atn/archives/august2003.html You could probably program and backtest such a system in Esignal, TS etc. Correct me if I am wrong, but current price is $995 for every 2 years. It was $175 lifetime license purchased before Jan. 26, 2007. There is a thread on the trade2win forum.
It would be MORE interesting if the verification were independent. It seems like the walk forward testing was performed by the same person? In other words, I do not find it interesting unless independently monitored, Otherwise, is looks like data mining, AI/neural systems, curve fitting, etc. They often look good, but seem mostly to fail when YOU try trading them...
I wonder what Michael Harris uses for risk level and position size. I do not find risk level or position size in the article mentioned above.
I paid $1,495 for a five year lease. I think $175 is for an upgrade from earlier versions. In principle you are correct, APS is a "white box" but it does not disclose the algorithm to find the patterns. The program does not check for pre-programmed patterns, like the hook pattern that was presented in that article just as an example. Although I do not fully understand the algorithm, it appears it is a polynomial time modified permutation method to check for formations that fulfill the user performance criteria. The algorithm is fast, I'll give you an example: for the QQQQ patterns using a data file with 3,000 daily bars it takes less than 8 minutes on a 2.40 GHz dual core with 2 GB RAM to complete a search for a pair of target/stop values. Ron
Hi Ron, thanks to you I purchased APS a while back and have learned a lot from it. I do not think it uses a modified permutation method but a data mining algorithm instead. Regardless, it does exactly what it's supposed to do. Thanks again. John
John, Data mining is a very powerful method for discovering patterns of all kinds. I think even retailers use programs based on such methods to discover hidden patterns in sales data. I would be surprised if APS is using such technique. It is not based on neural nets for sure. What systems have you developed with APS? Ron
I found more interesting the article about the Kelly formula. I have read another article in the past somewhere and thought Kelly% was the amount to invest but Michael Harris is saying that is the amount to risk. Huge difference. Check it out: http://www.tradingpatterns.com/About_Us/articles/articles.html