Yes, I have background in statistic, fuzzy logic, neural network and machine learning. However, I am not a profficient programmer. Therefore I can program in Matlab or SAS or such a high-level languages only.
This is a *major* concept, and most traders don't give it proper weight. A favorite quote... "All you have to do for the market to knock you off is keep doing what has been working".
Personally, I prefer the odds offered by evaluating 1,000,000 random strategies compared to the handful that could be created manually in the same time.
the best soft ware i have found for real time data mining of patterns is www.trade-ideas.com these guys actually use dr. andrew lo's quantified definitions of common TA patterns. its an easy program to use also. enjoy !
Lo is the anti Brownian Motion (Random Walk) guy. Whether he makes better money with this than the other school, I dunno. Either one didn't help me very much. You have to find the tricks that work by yourself. The trade-ideas' commercial gimmick being "based on dr. andrew lo's quantified definitions" looks very questionable to me. It certainly doesn't seem to make much sense in view of Lo's insistence on SIMPLICITY. To me it looks more like a supermarket outfit allowing you to pick your thing from their shelves (and pay for it - no money return garantee). nononsense
Simplicity is a goal to continually strive for and it's something that explains the "chinese menu" of choosing alerts and filters to model strategies in Trade-Ideas. This is in contrast to other tools that are formula-driven and require the use of a handbook or manual to create. Alot of inspiration comes from the valuable work of Edward Tufte, Yale professor, and his work on the quality of information design and the ability of people to reduce mountains of data into clear, insightful knowledge. âFor information displays, design reasoning must correspond to scientific reasoning. Clear and precise seeing becomes as one with clear and precise thinking.â -- Edward R. Tufte, <i><u>Visual and Statistical Thinking</u></i> (Graphics Press, 1997) http://www.edwardtufte.com/tufte/ He is a mentor and a great place to start when learning about how we process information.
My opinion is that "we do what we can." Backtesting will always be somewhat problematic but what is the alternative? Knowledge is simply knowledge. We can apply it sensibly or not. In terms of data mining I believe it is important that your results show strong performance amongst similar patterns or strategies. If they do not then the odds of them being junk increases. However if you begin to see a pattern amongst the patterns, ie similar styles producing similar results, even the most crude amongst the patterns, then you have higher odds of capturing something inherent to the markets.