APS- Automatic Pattern Search

Quote from forextrades:

Has anyone ever used it?

Opinions?
More...

Excellent program but only if you know what you are doing. I like the company, support is excellent. I also think Michael Harris is a true market wizard and I have learned a lot from his papers, books and software.

APS can find tons of historically profitable patterns but you must find a way of determining which of those patterns have a high probability of staying profitable in the future. This is the key issue here. You can do that by forward testing but I think it is not enough.

I use APS to develop systems that trade probabilities rather than signals from individual patterns. Without going into much detail here, I look for patterns of pattern formations and I calculate probabilities.

My advice to all trading software buyers: there is no software out there that can make you easy money unless you put hard work into developing your own edge. At least this is what my experience shows.

I highly recommend the new book by Michael Harris as a starting point for understanding system development, price patterns, risk management and automated system discovery:

http://www.wiley.com/WileyCDA/WileyTitle/productCd-047022908X.html

Ron
 
I find very interesting the possibility that some patterns can stay profitable for a period of time as long as 6 years or so as it is suggested in the following analysis:

http://www.tradingpatterns.com/About_Us/articles/article3/article3.html

Programs like APS are just tool for finding patterns with good performance quickly but I need a method for determining which ones will remain profitable in the future and I have not found a way of accomplishing that.

Any ideas?
 
Quote from frankl:

Have a look on this tool:

http://www.vectorbull.com/eng

Frank :cool:
More...

Looks like a black box of some kind. APS is a white box. It generates code you can use immediately in 4 different platforms or you can use the exact pattern descriptions provided by the program to develop code for other platforms.
 
For me, APS is too static. You can extract pattern, add them to your trading system, hoping, that they will work in the future.
But how often will you have to do this process ?

Yearly ? Monthly ? Daily ?

Is seems, that the real time version of VectorBull does this on a continously approach.

I think, their integrated real time trading systems speak for themselves.

Frank :cool:
 
Quote from hektor:

There is a free Tradestation Code from Mike Bryant that is better than APS and easier to use for intraday data.

http://www.breakoutfutures.com/Newsletters/Newsletter0907.htm
More...

I took a closer look and it is just a random code generator based on OHLC. Then, his exits are based on NBarExit rather than on a profit target and stop-loss like it is done in APS.

APS is a lot different from that. I have been studying its algorithm with the hope of coming up with something close to it myself. Basically, it uses a set of generic pattern groups - up to 88 in number - and the seaach is based on those. Each generic group is like a superset that contains maybe thousands of pattern formations. The groups were selected after studying their relevance to price action. For example, APS will not generate a pattern like H(8) > L (2) AND C(6) > O(2) or something like that which is fundamentally random and although it may have a good performance it is probably due to selection bias rather than relevance to price action itself.

I think the following paper about the statistical significance of the APS patterns and their performance endurance is an indication that there is no random number generation involved:

http://www.tradingpatterns.com/About_Us/articles/article3/article3.html

Basically, one can determine using just the Demo version of APS and QQQQ data that the patterns identified by APS in 2002 have remained mostly profitable. I find that very interesting and it seems there is more into APS rather than just naive random pattern generationas the author of that website seems to think.
 
You must log in or register to reply here.
Back
Top