Trends

Greetings

I am doing some research at the moment that I think the results of which could be very useful for the community; to find a reliable predictor for a near term ‘good’ period (above average Edge opportunity) to trend trade using a technical approach.

So I first need a robust way be classify the quality of a trend mathematically in-order to compare the ‘Edge Potential’ of trend trading in one period to another, before looking for predictors.

I am working with the premise that the market is always (constantly) trending; but ranges from periods of zero edge potential to very high quality edge potential in any given time frame. The two extremes are usually easy to observe in PA on the chart, but I need describers that can be coded.

This is what I have decided to look at so far:

[B]Price Correlations[/B]

The correlation in price movement from one period to the next over a large sample size is normally distributed with mean zero, but if you take a small subset of say the 5 previous candles, the price correlations tend to form somewhere in the range -1 and 1. So one characteristic describer of a trending period might be when the price correlation from one period to the next is closer to 1 and or the fight path of the correlations between the extremes.

[B]Directed[/B]

There might be some overlap with above but, if you take the highest of:

[ul]
[li]Previous High - Previous Open [B]or[/B]
[/li][li]Previous Open - Previous Low
[/li][/ul]
and divide it by by the Previous True Range, you should get a score between 0.5 and 1 to let you know how much of True Range in the previous time period was in one direction versus the other.

A graduation from an undecided 0.50 to a decided 1 in PA might give and indication of the aggressiveness of a trend if one is forming.

[I][B]Interesting note[/B]: on the day candle and even the smaller time frames, from any given open price the Distinctiveness score is

[ul]
[li]between 0.75 and 1.0 65% of the time [B]&[/B]
[/li][li]between 0.50 and 0.75 35 % of the time
[/li][/ul][/I]

[B]ATR/Volume [/B]

Low ATR High Volume might suggest a period of ranging or near offerer taker price equilibrium, in which case the onset of a trend might be a break away from those extreme values

[B]Crossing Times[/B]

Using a standard value for MA on PA, you might look at how long it takes compared to an average for the price to cross (Open to Close) the MA. Higher crossing time might signify a better quality trending period, also the True Range from one cross to another might be a describer too.

Those are some I was thinking about tell me your thoughts on the subject and scribble your ideas here.

Some interesting thoughts there, a bit deep for me.
I generally trade with price action, but sometimes use a 60 sma to hunt out trends in different time frames a-la 3 Ducks System.
Im sure I’ve read that markets are only trending about 20% of the time, if you look at EurUsd for example its been ranging for a good 18 months or so.

Hey if you do PA the correlations might actually be something of interest to you. I think that if the price movement (Open to Close) in one period is not correlated with the price movement in the next then that should mean that there is no predictive application or quality of PA at that time.

That is the reason why I am skeptical about PA because overall - over a large set of data the correlations between previous and future price movement is zero (average) and normally distributed between -1 and 1.

But in smaller time intervals (smaller averaging periods / sample size) the correlations are noisy, they cluster and also move more predictably between the two boundaries, there will be periods where PA is very reliable. If at any given time or correlation value, the probability of moving up or down by a fixed amount is the same and the smaller interval correlations must average out to a larger sample interval mean of zero, then smaller interval correlation values nearer -1, have no where else to go but up.

With an interval of say the last 14 candles being near -1 and having to move up, that means the incoming set of candles must push the correlations up, say the correlation value moves from -1 to -0.6 in the next 4 candles, then that increase would be attributable to the those incoming 4 candles, meaning that they must be very positively correlated.

And the correlations at 14 candles will steadily but more randomly move towards the other barrier of 1. So I think that the most predictably reliable period for PA would be in the ascent, as the correlations in 14 trend from -1 to 1. In that case whatever the price action is doing, it is more likely to do consistently than at any other time, if its going, up, will continue,if going down will continue. That might be a good time to get in on developing trends or scalp. Indicators like the RSI and Stochastic Oscillator or other indicators that are derivatives of PA should also follow suit with reliability.

You should be able to write and indicator that tracks the correlations of PA. The graph should look something like the ATR indicator, ranging from high to low.

But I might be over generalizing the implication of the zero correlation on a large sample size, not sure.