Trendiness measurement

ProfLogic · Jan 30, 2008

Quote from PetaDollar:

It all comes back to the same thing: can you measure the effectiveness of your entries-- and if so-- does requiring a trend (defined however you wish) increase the effectiveness of your entries?

That being said, I have never experimented with magnitude of trend, only existence of trend. My definition is a series of higher highs and higher lows on a particular timeframe (reverse for trend down). I identify it by eye. I have found it to enhance the effectiveness of entries dramatically.

If anyone is going to try this for the first time, please note that measuring the P/L of a series of trades is not the same thing as measuring the effectiveness of entries.

Well stated.
Trend is something one doesn't trade, it is something one uses to determine the strength or lack-there-of in the effectiveness of entries &, to add, the potential longevity of each trade.
When one can accurately read the trend of the chart they are trading then executing specific trades at oscillations that put that trade into the direction of the confirmed strength of that chart is the goal!
They shoot . . . they score. Probability increases exponentially!
Thanks for the mental trigger Peta.

bearmountain · Jan 31, 2008

Thanks for a great discussion of ideas. My original intent was to somehow classify stocks/markets that trend more than others.

So for example if I make a list of 300 actively traded stocks, somehow determine the top 10 which exhibit the greatest 'trendiness' intraday. The hope there is that these top 10 would be better cadidates for a trend following system than those which display poor 'trendiness'.

ProfLogic · Jan 31, 2008

Quote from bearmountain:

Thanks for a great discussion of ideas. My original intent was to somehow classify stocks/markets that trend more than others.

So for example if I make a list of 300 actively traded stocks, somehow determine the top 10 which exhibit the greatest 'trendiness' intraday. The hope there is that these top 10 would be better cadidates for a trend following system than those which display poor 'trendiness'.

What's better is make a list of the top stocks with a volume of 5 million shares a day or more. This will give you stocks that offer enough volitility and liquidity each day to "BE TRENDY".

Then monitor that list each day for the stocks that are confirming the "Trend" of chart (Intraday or Swing) of each stock you are monitoring.

Common sense really is applicable to the price movement of the markets IF you give yourself the opportunity to learn it. It's up to you though not someone else to give you the information on a platter.

Hook N. Sinker · Jan 31, 2008

Quote from bearmountain:

Thanks for a great discussion of ideas. My original intent was to somehow classify stocks/markets that trend more than others.

So for example if I make a list of 300 actively traded stocks, somehow determine the top 10 which exhibit the greatest 'trendiness' intraday. The hope there is that these top 10 would be better cadidates for a trend following system than those which display poor 'trendiness'.

If you can't define something then you likely can not measure it. So it is important that you define trend. Define trendiness. Define terms in your personal way.

You might be thinking backwards regarding the trend following system idea. Final profit of a trend following system applied to historical price data of a particular security might be one measurement of trendiness.

A trend following system might be a measuring tool to identify which price histories show the greatest trendiness.

ProfLogic · Jan 31, 2008

Quote from Hook N. Sinker:

If you can't define something then you likely can not measure it. So it is important that you define trend. Define trendiness. Define terms in your personal way.

You might be thinking backwards regarding the trend following system idea. Final profit of a trend following system applied to historical price data of a particular security might be one measurement of trendiness.

A trend following system might be a measuring tool to identify which price histories show the greatest trendiness.

YUP!!

Jossan · Apr 22, 2008

The Hurst Exponent is an estimate of the predictability of a stock and can be used to find stocks where the price is not that efficient.

That makes it an indirect measure of the trendiness of the equity.

ucf_student · May 3, 2008

Quote from Jossan:

The Hurst Exponent is an estimate of the predictability of a stock and can be used to find stocks where the price is not that efficient.

That makes it an indirect measure of the trendiness of the equity.

What software calculates this? I suppose you can have a different Hurst exp for the same financial instrument over different time frames? Can you use daily weekly and monthly data too? Does it make any sense to use it on intraday data?

oTzt · May 4, 2008

More precisely, Hurst exponent is a metric of the randomness of a given serie.
If the serie is a time serie, it will tell you in what proportion the detrended datas "fill" their space (see it as a square crossed by a line : 1 means minimum fill : the line is perfectly straight ; 0 means maximum fil : the line occupies whole the availlable space in the square).

It can be computed several ways. The easyest one is called "rescaled range" (Google it to find lots of answers, or Google "bearcave hurst" to find a really didactic site (I'm not the author and don't know him...)

Typically, H will range between 0 and 1, with
* 0.5 meaning "perfect randomness" (Brownian time serie),
* 1.0 meaning "strong persitency (an up day tend to follow an up day)
* 0.0 meaning strong anti-persistency (an up day is usaually followed by a down day).

However there is no clear bounds splitting each of the three possibilities (persintent serie, anti-persitent, or brownian motion) and, as usual, "what happened in the past hours, days, etc. is not necessarilly what will happen in the near future", so I'm not sure it will give you any usefull information.

Olivier.

oTzt · May 4, 2008

And to answer your question about softwares to compute it :

It is known in the tech analisys world under the name "fractal dimension", and has been the subject of article(s) in TASC (don't remeber wich issue(s).

Most usual contributors proposed code to programm it into their respective softs. You will certainely easilly find it.

Olivier.

ps : fractal dimension is worth "(number of Euclidian dimesions of your datas series +1) - H".
Hence, for financial time series df := 2- H; and ranges between 2 and 1, with 1.5 meaning "brownian motion")

Mercor · May 4, 2008

Quote from Wood474:

You don't need to be a scientist to look at a chart and see. Yes ADX is ok, but quite frankly, send me over the charts and I'll tell you in 5 seconds which has the best 'trendiness'! Granted, that would only be at one chart. Simple, look back over charts and you should be able to see at a glance. B]

Yes, Study Charts!.

Try reverse engineering, look at a chart and try to estimate where the ADX is, estimate where the moving averages are.

Once you read enough bar / candle charts you should be able to plot these indicators fairly accurately by eye.