Get over yourself. I doubt that most people care about who's a genius and who's not. The biggest issue is about the "wanna-be system traders" preaching his posts without running any substantial work to talk about his old threads and posts. As a matter of fact, I want to ask you what your obsession towards someone being a genius is.... Do you have an inferior-complex towards being considered a "smart person" or something?
ja. You prolly have the same whitepapers and websites regarding it. I simply Googled. ... of course, don't bother asking me for some code. Do your own work. (That's if you or others thought about getting some freebie)
I don't find it useless. His old posts serves as an insight of how he processed and approached issues regarding systematic trading. The specific tools, criterias and systems don't matter so much. I doubt that people believe in "All-in-One" measurement that can tell you whether a system is ultimately good/bad, robust / curve-fitted. (If you do, you're a moronic newbie.) Seriously, it's like you're measuring a short-term arb model using a trend-following measurement. Who's stupid enough to do that?
I don't need anyone's code, I write all of it by myself It was September or October of 2007 when I first started to seriously work on the ideas. I had built data mining framework from scratch. The idea was to find statistically significant sets of identifiers that filter out the days on which particular entry provides edge (based on acrary's edge test). This has not provided any useful results, as found models fell apart in out-of-sample dataset. Next, suggested by Alan, I moved to classification of periods (e.g. days). Started from building a framework which takes price data input, volatility-normalizes it and removes noise. I.e. simplifies data to the point where I can easily write a simple algorithm to classify data. It did happen and I got some nice results, but when I tried to implement models for the most common price behaviors found I failed to come up with anything useful. They just don't beat the random (edge test). Now I'm trying to develop edge-based models from different perspective. From "reverse-engineering" point of view: what makes a model beat the random entries (edge test)? I'm not gonna disclose here now what I have come up with, and it's still work in progress, that I will finish sometime this week. The paper about AIS I have is about Celada and Seiden model. Since I have other stuff to do, I will get back to immune systems later. I'm not looking for free stuff or code or anything like that, just a few guidelines to put me on the right track, because this work is very time consuming and often leads down the wrong path.
I said - I believed he is over hyped and he did make some good points but he certainly isn't / wasn't the best trader that graced the pages of ET. There are a few more that frequent this forum that I know are outstanding.
some content: a portfolio of no correlated systems is better than a portfolio of correlated systems. instead of using normal correlation it is better to use stresscorrelation. instead of using correlation it is better to test portfolios of systems as a whole and not bother even calculating correlation. all due respect: that is common know how in the professional money management arena. all else is ... well, ...
i am talking about a professional quantitative trading operation. you are talking about something different. and sorry to say this, you seem to have not the slightest idea what this discussion is about. which does not mean that you do not make money. but that would be an additional discussion. as useless as this one. take care.
by the way, since maestro is wrong AND i am wrong ... ah, what is left? he answers the question: is corr important by no, my answer is: yes. yours is? don't bother answering, i guess you're coming with some traders don't need this kind of thing, math in general and so forth.