Can someone explain to me how "rigorous scientific research" aka "backtesting" predicts the future? If something works 7 out of 10 times the past 100 occasions, why could it not work 3 out of 10 times in the next 100? Isn't that exactly how setups go from profitable to outdated? More importantly, how long would it take someone trading such a setup to realize that things have changed, so that the setup is no longer profitable? The more data you have to "support" your setup, the longer it will take for you to decide things have changed, and the more money you will lose before you figure it out.
That's all nice if we're talking about coin flips. But what if the coin was not equally weighted? Futhermore, the one flipping the coin was also the one who fabricated a set of coins and could switch between various tendencies for heads/tails as he pleased? Now assume another person gets to call it in the air, heads or tails, and he is betting $100,000 for each toss with the tosser. Throw in the fact that 100 observers are also betting on the toss in a parimutuel system, 100K per bet, and that the coin-maker and coin-tosser can also ad bets anonymously. Now tell me, how reliable would backtests be for this scenario?
That's quite a claim, but there's a great deal of evidence to the contrary. In fact, quantified pattern recognition is in many applications far superior to the human eye. Now granted, your basic "off the street" trader, armed with EasyLanguage is not going to be able to do anything but the most naive, the most extremely rudimentary pattern quantification, so they might be led to this conclusion that it can't be done. But to make the leap to calling it "impossible" - well, I disagree. I've seen quantification at work in many applications that is very, very impressive as compared to human judgment "by eye." Fletch
It cannot predict future. Reason is that true predictive methods are not price and numerically based so they cannot be computer back tested.
Just curious. Before you started this thread announcing and proclaiming that this book is the best TA book ever published (with two exclamation points, no less), would you be so kind as to advise whether you actually read it from cover to cover before reaching and publicly declaring your lofty conclusion? In case you're wondering why I ask: http://www.elitetrader.com/vb/showthread.php?s=&postid=1044216#post1044216 http://www.elitetrader.com/vb/showthread.php?s=&postid=1044221#post1044221
<i>"But to make the leap to calling it "impossible" - well, I disagree. I've seen quantification at work in many applications that is very, very impressive as compared to human judgment "by eye." Fletch"</i> Agreed completely. It is possible to code most parameters of a pattern setup. Assuming it includes two or three chart studies to complete the pattern setup, one or more persons would have to quantify where exactly the boundaries are between valid pattern and invalid pattern. The "human eye" can discern a patterned long signal that doesn't quite measure up to perfect parameters and still be a great trade... when the overall trend is straight up in the air. Likewise, the exact-same long signal pattern parameters would be passed up in a defined downtrend. All of those many nuances would need be settled and written into code. It could result in a very workable format, admittedly different than visual - discretionary trading. * The pure statement that what can be rigidly quantified is valid, what has degree of varience is invalid across the board is in error. There is no doubt that what can be quantified is good, but also much of what cannot be quantified is equally good.
Walther is correct. Only the dead chicken I wave over my keyboard has this capability. You can read all about it in my upcoming book "True Quantifiable Technical Analysis - The Holy Grail".