Hey surf...while you're at it....could you also ask the homeless bum on your local corner the best way to make a million dollars?
LOL, citing sources without an idea of what the source is actually saying? When you try to optimize a losing system, all possible outcomes (ie. trying with 10MA, 20MA, 30MA, etc,etc increasing in 5p or 10p increments) will lose.
No, what he is saying is: Academics consider a MA crossover with 200 different parameter sets to be 200 rules, whereas actual traders consider that to be one rule with 200 different parameter sets. They didn't test 5000 actually distinctive rules. Not that it matters. We already know that public-domain TA is crap. Any of it that ever worked historically has long been "mined out".
ok, thanks. So, these folks that are claiming high win rates and years of profits following al brook style Price Action and or traditional chart reading are simply making things up? surf
The paper is online: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1181367 See page 8 ff. 'Trading Rule Specifications': 'We apply 5,806 of the technical trading rules suggested by Sullivan, Timmermann, and White (1999). Sullivan, Timmermann, and White (1999) test in excess of 7,000 rules, but one of their five rule families requires volume data which are not available for the MSCI indices we examine. The four rule families we test are Filter Rules, Moving Average Rules, Support and Resistance Rules, and Channel Break-outs. Sullivan, Timmermann, and White (1999) provide an excellent description of each rule in the appendix of their paper, which we recommend to the interested reader.' The referenced paper by Sullivan, Timmermann and White is not online for free, however it is clear from the context that there a really only 4 different 'systems', this is what they call 'rule families'. Everything else are just different parameters sets. As I already said, this is the way it is done in many other academic studies also. No surprise here. I have yet to see a paper which actually tests thousands of different trading systems, let alone something 'original'...
Which invariable leads back to the idea that you cannot test something you do not know, aka a proprietary combo of inputs and trading rules.
OK now you're pretending like you don't know the difference between discretionary trading and mechanical trading. I can't comment on discretionary trading. If it works for somebody, more power to them. And academics typically don't try to test discretionary trading; I doubt they did so in the paper you cited.
surf, I hope the above faulty logic study is not the entire foundation of your argument against TA...not to mention that you're being highly delusional if you believe that your Price Drivers are not TA