Trend following .......

[wombat953]

And now a simple question. Has anyone tried to identify the current trend?

Clearly there are many ways to do this, but I am wondering what people have found (and tested) to be the most successful, with a good balance between providing late information, and getting whipsawed by being too early.

For example, one could use a MA (Of which there are 5 types, I believe). So now what criteria do we apply? If the current value of the MA is greater than the prior, we have an uptrend and vice-versa. Or might we need three (3) successive values, such that the current is greater than the prior, and the prior is also greater than the one before that (and again vice-versa for a dowtrend).
Is it necessary to close above(below) the MA? Should we be more cautious and seek two(2) such closes? Should we seek a close by a certain distance(percentage) above(below) the MA line? What about the angle of the line, assuming this is measurable (by EFS)? Dr. A. Elder is ambivalent about the application of the jazzier filters to the simpler MA approaches(p.126, TFAL).

Or is a MACD Histogram a better bet? (This is Elder's first choice in Triple Screen to identify the major trend). And if so, naturally similar questions apply.

Of course, one can use combinations of trend-following indicators too.

Clearly there are many ways to do this, but I cannot be the only one to have thought about this. People must have backtested this kind of thing.

I am assuming a 13 bar MA, typically exponential, on, say, a 60 minute chart to identify a trend, and then zoom in on a shorter term chart (5-10 minutes) to find overbought/oversold conditions (using an oscillator - and so another bagful of similar questions) to enter a trade, but only in the direction of the identified trend.
Thanks
George

[JJordan]

Trend

Hi George,

The way I determine trend is use a simple moving average. Interval doesn't matter much....your preference.

The easy way for me is to watch the price crossing the
9ma and the 40ma. I generally accept the 40ma as the trend.

If you put a bollinger band on that , the centerline is usually
the 20ma.

When the cross is to steep you can expect the support resistance
to be broken. Beyond simple method.

Good luck, try it trade with the 40ma. Here's an example
The blue line
~j

http://share.esignal.com/groupcontents.jsp?folder=Moving%20Averages&picture=0624ndx_2min_Envelope.png&groupid=225

[wombat953]

Linus
I am a short-term trader of the eMini index furures. I am looking for something that can be coded into EFS, and that will ultimately return an unambiguous result: Up, Down, or neither(presumably transitioning between the two; but it would also be desirable to identify a state that is truly trendless, as opposed to transitioning between the two states). Right now, I am using the 13EMA > prev_EMA, and the prev_MA > Previous_previous_MA. And, that the price must close above the EMA. So, three (3) conditions.

Reverse these to identify a downtrend.

I realize there are many ways to do this. I am wondering what set of parameters/indicators, etc. people have found to be "optimal" after back-testing.
Thanks
George

[fabrizio]

Wombat, JJ

Here is -MHO on your discussion, if I am allowed to.

Jolly good Bravo to Linus: very wisely he spotted :
* the 50 SMA
*The RTC ( adjusted as per STD)
* What kind of issue?

And Very nicely Linus plotted ellipses to carefully understand where that trendy price movement could retrace without changing is original direction.

RTC: No matter if you work on 3 minutes or weekly: RTC gives you the direction for:" Until a price stays inside the +2 -2 basic STD trend channel there is no reason that it would change its direction" If it break its well that is another story ( a) break up in up trend is one think /(b ) break down in a up trend is a different meaning....

Simple is simple really : Once you make of the RTC your basic starting analisys point you will not be able to work without it.
(carefully always plot the Pearson correlation to see how viable is the RTC)
Of course the XREF it is of the outmost importance...

So said ( and the 50 SMA -or even EMA little more dancing....- is a great tool too I always have it on)

Butt lets play an IBSEN comedy: why there should be a defined trend at the present moment?

Ok 80% of the time the market is trendy, but if you made this question is because you too are baffled by the behave of the market, right?.

IMHO there is quarterly DOWN TREND, (ELLIOT, Demark, confirm it) but IMHO the daily/ weekly is there on the border:
the market and better the COMMERCIALS have not taken full decisions on HEDGING what (direction) in their portfolio:

so if on DJ and S&P the BIG speculators and the tol traders are definitively short , the big are still long on the OI.

But we are 3/4 down.......the remnaining 1/4 is huge indeed

That means something : indecision.

Cordially

[wombat953]

I hope that this can be a continuing discussion. What range of values for Pearson's correlation (r) are considered minimally desirable to have a suitable level of confidence in a 2SD Regression Trend Channel? 0.8? 0.9?
I am developing some EFS which I am happy to share with others, Right now, it is based on overbot/oversold conditions within a prevailing trend. Since it uses two different timeframes, it cannot be tested witheSignal backtester. I am trying to set up a testing infrastructure which writes the output of a longer term EFS (the trend follower) to a file , and then process this file in a second EFS which correlates the price bars in the "shorter term" chart with the timestamps from the bars from the first EFS, and then apply the overbot/oversold indicators, depending on the direction of the trend (if any), as supplied on input.
This is almost in place.
In the end, with such an infrastructure in place, the indicator variables themselves...our trading "edge"...can then be changed at will, back-tested, and then forward-tested with a suitable sample size (i.e. at least 20 trades according to Mark Douglas, with real money).
Thanks
George

[fabrizio]

George

I consider Parson's 0.85 /0.80 acceptable , 0.9 and above optimal.

Ciao

looking forward your efs

Cordially

[wombat953]

Getting back to this idea of identifying trend (in an EFS) within a "longer-term" EFS, I am using EMA(0) > EMA(-1) >EMA(-2) > EMA(-3). And vice versa for a downtrend. Of course there are infinite variations on how to detect trend, and we discussed this a few weeks ago.

Now my question pertains to what happens if the cycle is broken, if we get an EMA value which is less than the prior value (or greater, in a downtrend).

As I have coded it currently, this breaks my sequence, the existing trend is "nullified" (there is now no trend in place), and I wait again for three successively higher (or lower) values) to constitute a trend , before my second "shorter term" EFS can use the trend to provide a trade signal, based on overbought/oversold oscillator.

I am using hourly interval for trend identification, and 10 minute interval for overbought/oversold indicator. This is basically Alexander Elder's Triple Screen (these EFS's being the first 2 screens).

But something inside tells me that this is not right, the trend may well still be in place when the sequence is broken, that this may simply be a counter-trend retracement that my shorter-term oscillator should indicate. In which case, I should not code the EFS to nullify the prior trend, but leave it as is, and not change it until another three successive values form a new sequence (which may, of course, be the same trend direction). Of course, we don't
know if a retracement is - in fact - a retracement, or a genuine, change of trend, but presumably retracements are more common, at least in a trending environment.

I am wondering what the thinking is on this interpretation?

Separately, for those who recommended Regression Trend Channels as a trend identifier, I am wondering if traders here ever
go against the primary trend, and ever trade the retracements? So, in an uptrend, when it hits the upper channel line, do you then short, or only ever trade in the direction of the primary trend?

Thanks
George

[fabrizio]

Separately, for those who recommended Regression Trend Channels as a trend identifier, I am wondering if traders here ever
go against the primary trend, and ever trade the retracements? So, in an uptrend, when it hits the upper channel line, do you then short, or only ever trade in the direction of the primary trend?

Thanks
George [/B]

George
RTC is the best and simpliest trend identifier.

1) Get back to Graham-Dodd Bell curve distribution. Plot it and "define" the STD areas : the center (corresponding to the top) of the bell is =0 than you got always on the x axis: +1, +2, +3
on the right and -1,-2,-3 on the left as STD values.

2)Than : 68% of the prices falls between STD -1 to 0 &0 to +1
29% of the prices falls between STD -1 to -2 & +1 to +2
3% of the4 prices falls between STD -2 to -3 & +2 to +3

3) Therefore FROM 95% to 97% of the prices falls within a STD -2 to 0 & 0 To +2

4) A RTC with a STD set by default on +2 / -2 is exactly the same :
Upper channel is +2 middle line is 0 lower channel is -2.
You have 95 % of the chances that until prices moves with the same direction INSIDE the RTC there IS NO REASON FOR HIM TO CHANGE DIRECTION.

[u]So said : Adjust an up RTC STD up to the highest bar volatility (in this case : up)[/b]

Now I answer to your question:
If price approach the UPPER channel I know that FOR SURE that is a CRITICAL POINT.
SOMETHING WILL HAPPEN.
RIGHT THERE IS THE PLACE WHERE "PEOPLE FIGHT"


Fore sure If he Broke the Channel and CLOSE the Bars OUT, the trend will not be as before.

Maybe istitutionals stepped in and will push it up and up in a steeper way with impluse designing a new "tail" towards a +3 STD or +4 STD and the day from NORMAL will become TREND.

Or he just reached the max of the tail wher the "auction" is no more feasible " and he will steeply reverse its direction towards new minimum.

Or He does not break the channel completely and bounce down again going towards its natural "fair value " which is close to the median line of 0 STD at the center of the channel.

CERTAINLY: PRICES AWAY FROM FAIR VALUE = OPPORTUNITY

[color= red] One more thing: the difference from a RTC a MA is that RTC HAS NO LAG: READ THE MARKET IN REAL TIME[/color]

Hope I have been clear and helpful.

[wombat953]

fabrizio
You have given me a lot to look at this weekend!
Let me ask, can we code for this in an EFS? Is the RTC the same as the Basic Studies "Linear Regression" Study?
vLinReg = new LinearRegressionStudy(Source, Length, StdDev);
which has the following members:
study.getValue(LinearRegressionStudy.BASIS);
study.getValue(LinearRegressionStudy.LOWER);
study.getValue(LinearRegressionStudy.UPPER);
Thks
George

[fabrizio]

>You have given me a lot to look at this weekend!
If ever I helped you I'm sincerely glad

>Let me ask, can we code for this in an EFS? Is the RTC the same as the Basic Studies "Linear Regression" Study?
Yes , the Basic vs the Aget RTC are the same, Thus the basic need a little more config.


have good W.E. George

Fabrizio

[daytradingjunky]

just when I thought I knew everything, there's a bunch of useful info. thanks!

[fabrizio]

Originally posted by daytradingjunky
just when I thought I knew everything, there's a bunch of useful info. thanks!

daytradingjunky

I'm glad - as well as many others in this BB - to support others or share with them knowledge & experiences.

Thanks to you for dropping your lines!

[Jharvey407]

There is a lot of useful information here. I see that everyone has a different way of defining the trend.

I use a SMA, that I test for each currency that I trade, until I find the one that works.



I have written a EFS that returns a value of Up, Down, or Sideways. I do not trade against the trend. My backtesting and experience have shown it to be only marginally profitable.

James

[junkyforjc]

Trend Indicator

I was wondering if anyone had an efs that they were willing to share. I am looking for something that, while trading a 15 min chart can indicate the trend on a 1, 2 or 4 hour chart. I have read this post and see lots of great discussion about this subject but do not know how to code it in an efs.

Thank you in advance.

jerry

[ACM]

jerry
You may want to check the efs JasonK posted in this thread.
Alex