What if it was making 5 consecutive Higher High and Higher Lows in the previous bars and the newest bar was Higher High but Lower Lows .... though Close being equalled to High. Or... Higher Highs and Higher Lows but Open = High and Close = Low??? Or... Also how about the market rising 100 days without a retracement larger than 10%? Or... 10(20, 50, 100, 200)-day moving average being up... ... ... I'm sure there are others...
you crack me up, gann. funny. how about a simple study to show that 2 consecutive higher highs will lead (or do not lead to) atleast a 3rd higher high if not additional higher highs. this will certainly quash the trendies fire, and twist their mellon a bit. LOL !
Ahhh... that'll just mix things up and it's pointless.... I'll just give the answer... Single tick, Technical Bars (1-min., 3-min., 1-day+... etc. etc.), directional swings, % retracements... have no significance in terms of serial correlation, "by itself". Larger the data of the market... the clearer it becomes... "By itself"
larger, meaning longer time frame ? with no serial correlations, how can it be said that trend exists? you are twisting their melons, man !
As usual, good question, TsG. Permit me to stay in my narrowly focussed pseudo fantasy world a little longer. Assume for the moment that I am a market player with no knowledge of the underlying market (which in my fantasy world is pseudo randomly generated as per a simple algorithm posted earlier). At this stage, I cannot define trend a priori, nor a posteriori for that matter. What I do know, however, is that I can consistently make money in this particular market by applying a technique (say KAMA) which seems to be able to exploit the very thing I can't define. Perhaps I can get closer to a definition if I eliminate KAMA. Assume that somewhere within this thing called trend, the probability of continuation is higher than the probability of reversal. (Remember I only deal with 2 states.) And I base this on logic rather than knowledge of the underlying generating algorithm which, as I stated earlier, I do not possess. Assume also that we can define a basic trade with as few parameters as possible, having entry and exit rules that are totally independent of KAMA i.e. with target(s), (trailing) stop(s) etc. Then maybe the following could be a winning strategy : (a) Enter trade(n) in any direction. (b) If trade(n) is a winning trade, trade(n+1) is to be opened in the same direction.** (c) If trade(n) is a losing trade, trade(n+1) is to be opened in the reverse direction. (d) Continue applying the rules in (b) and (c). ** Slippage and commissions don't exist in my world. However, one could achieve the same by not closing out and just resetting target and stops. The assumption here is that by the application of Bayesian probabilities, the outcome of our last trade is the best indicator of what that thing called trend is. Will I be able to eliminate KAMA using this embarrassingly simple strategy?
No... KAMA is simply a adaptive parameter basing it's figures from noise vs. "volatility"... Simply, it's a noise-filtered indicator. Your ABCD and KAMA systems are developed based on 2 different aspects of the dataset. It would miss the significance of trading systems.
I have no idea if this is what I meant. Serial correlation of the derivative of a time series is hard to conceptualize, after all. I just thought about something to test, and like before am thinking out loud. I'm not certain what the outcome will be; if I were, I wouldn't be posting.