I may be missing the point, but I remember an Active Trader or Stocks and Commodities article a few months ago that showed the probabilites of an up day after one, two or three down days in the market. And the results definitely seemed very non-random. Doesn't this show some kind of simple correlation and TA? Also: I've done a lot of backtesting and I've definitely seen what you guys are talking about: it is difficult to find a way to make oscillators beat break even. But to me the key is to create hybrid indicators in combination with volume, price patterns, etc. and then test across many markets and many years. If your "system" can survive that, then I can't imagine how it you could be just an "outlier".
The stochastic oscillator - a favorite of many - what does it tell you? What market participant behavior was the originator trying to quantify? Since all indicators of any nature are merely quantified behavior, what behavior had he noticed empirically that he wanted to put in a usable form that might demonstrate that which we all know.... human behavior repeats itself? So does human behavior repeat itself exactly the same way each and every time? Do all humans behave the same way at the same time? Does the stochastic oscillator do what it was designed to do? Is this information useful in the quest for profits? Show me any indicator that always works - and I'll make you an offer you cannot refuse!
An oscillator only has to give you an edge some of the time. Also, try looking at different approaches to using an oscillator other than the "canned ones". Try looking at a detrended oscillator and see what it tells you above/below 2 or 3 stdevs out. Is it just an outlier event or is there any edge there? Try looking at intermarket oscillators. Isn't seasonal hedging and stat arb just an oscillator of price between two different underlyings? Every looked at stat arb'ing oscillators of two different underlyings? How about an oscillator of volatility? Think outside the box. There are edges to be found. But once you've found an edge, you still need to build a tradeable method around that which is suitable to your trading psychology. Not all edges are tradeable.
I think I see your point: no mathematical tool can promise to always predict the complex psychology of a market or even individual stock under all economic conditiions. But, that said, is there any reason to believe it cannot predict the psychology 55% of the time? or maybe 65%? Isn't that all you need to win the majority of the time with proper risk management?
Not sure about oscillators, but i always fade any stock that is up six days in a row provided bar 6 is two points higher or more than bar 1, these are generally good for a 5-8% pullback, before they may or may not continue uptrend. No homeruns, but a lot of singles.
Even simpler systems can be profitable... I don't mean to say that yours is complex, but 6 rising bars in a row is a relatively rare occurence.
ShoeShineBoy you don't even need that high of a percentage wins if the profits v losses are significant. I believe that is the theory behind Turtles, to name a famous one, but even that is lame with information access and commissions so cheap these days. I am not a big P/L analysis type, but I advocate neither type. However, it seems that the lower the P/L ratio is per methodology, the higher the win rate must be to pay the rent.
Mike, I completely agree...the premise of oscillators is faulty. Also...I don't think it can be proven that oscillators are an effective predictive tool. I personally use them as "confirmation" tools. Dottom said it perfectly.... "Think outside the box. There are edges to be found". Thus, I rarely use indicators the way they are intended (default) to be used. NihabaAshi
Wally, you are right about that, there are not a lot to choose from, but taking the ones that come along is high percentage. Also i find that most normal breakouts will fade after day three, generally for 1/2 a point or a little more. The trouble comes with the ones that do not.