Optimized Trend Trading | Page 55 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Optimized Risk vs Reward Equation 335 replies

Loss Less Optimized Strategy 63 replies

Optimized Parameters in Metatrader Testing? 3 replies

"Optimized Parameter" feature in EA? 0 replies

Strategy with Optimized parameter 1 reply

Attachments: Optimized Trend Trading
Exit Attachments

Optimized Trend Trading

  • #1,081
  • Apr 3, 2011 5:50am Apr 3, 2011 5:50am
hi,

god. this is crazy long, not to mention the drama.

i been trying to read this whole thread for the past week. stop at page 60 or so. really tiring.

anyway, i want to ask, so far has anyone found that DSP alone can give a trading system that is reasonable? in that you get a decent sharpe ratio, with acceptable drawdown? i know someone said he uses band pass filter and relative vigor index to decide if he is in a no trade range, but i just want to double check with people here.

not looking for a machine to print money, but some thing to help guide trading. i have personally took the pain to backtest all ehler strategy in one of his book. as a lot of you might already know the result sucks, then i went on to try other ideas, and only recently revisited the concept after seeing this thread.

so i hope someone can point me in the right direction, for example, what is promising currently so i can go in and study it in depth instead of jumping from indicator to indicator.

thanks.
  • #1,082
  • Apr 4, 2011 12:27am Apr 4, 2011 12:27am
Quoting assqwert
hi,

god. this is crazy long, not to mention the drama.

i been trying to read this whole thread for the past week. stop at page 60 or so. really tiring.

anyway, i want to ask, so far has anyone found that DSP alone can give a trading system that is reasonable? in that you get a decent sharpe ratio, with acceptable drawdown? i know someone said he uses band pass filter and relative vigor index to decide if he is in a no trade range, but i just want to double check with people here.

not looking for a machine to print money, but some thing to...
Check this out
http://www.forex-tsd.com/manual-trad...strend-30.html
  • #1,083
  • Apr 4, 2011 7:47am Apr 4, 2011 7:47am
I just want to share an article. It is about modern technical analysis and high frequency finance.

This is the best so far.

http://www.scribd.com/doc/35885528/A...quency-Trading

Please consider this practical guide:
http://www.csie.ntu.edu.tw/~cjlin/pa...uide/guide.pdf

Well as a retail trader we do not have access to high frequency technologies.
However there are some very interesting ideas about ICA (independent component analysis) and SVM (support vector machines) combination.

Until now I have considered another possible combinations of technology:

Wavelets + SVM
SSA + SVM

I consider end - of day models because they seem to me more realistic, because of the problem with data integration.

As a software my choise is Rapidminer. We have both ICA and SVM.

I just want to add a bibliography about SVM models. Well you can build some of those models with Rapidminer.

http://www.svms.org/finance/
  • #1,084
  • Apr 4, 2011 9:59pm Apr 4, 2011 9:59pm
Quoting assqwert
hi,

god. this is crazy long, not to mention the drama.

i been trying to read this whole thread for the past week. stop at page 60 or so. really tiring.

anyway, i want to ask, so far has anyone found that DSP alone can give a trading system that is reasonable? in that you get a decent sharpe ratio, with acceptable drawdown? i know someone said he uses band pass filter and relative vigor index to decide if he is in a no trade range, but i just want to double check with people here.

thanks.
Market condition is one of those really important, and difficult to answer problems. And you don't just want to know when to stay out of the market - you want to know the market condition in order to consaider switch trading strategies, or to adapt the one you're using.

A number of solutions I've seen for filtering on market condition
- using FDGI and IVar
- using the angle of Ema(34) to determine trending vs consolidation

but you also need to define exactly what market condition means - does it mean trending vs ranging? At what timeframe - ranging on a daily or weekly timeframe could well be trending on a 1hr or 30min - at least trending enough to be tradable if your strategy is trending following one. So this probably not the best definition. Its probably better to look at stability and volatility of the market on the timeframe you're on. John Last posted a pretty good definition lifted from the turtle traders here

http://www.forex-tsd.com/385636-post108.html

Question for you - how exactly did your buddy use a band pass filter and the RVI to determine trend vs range?

cheers
f451
  • #1,085
  • Apr 6, 2011 11:54pm Apr 6, 2011 11:54pm
hi,

thanks for pointing out the range of market condition, i was aware that it is different time frame, but i never thought of looking at it that way.

as for the band pass filter and vigor index, it is from a post a few page back, so i am not sure how he is using it.

i took a look at ema 34, looks interesting, seems to follow the trend quite well.

as for FDGI and IVar, not sure where to get the code for this. i googled, and saw a tread call ASCTrend? am also not sure which is the newest version. but from what i saw from john last post, it does not seem to do a good job, it seems to able to get it only some time, and often when a trend is already well underway.
  • #1,086
  • Apr 7, 2011 2:03am Apr 7, 2011 2:03am
Make a google search there is only one version of iVAR and FGDI.

I think that you should consider that the trend may have different meanings depending on your method of defining it, especially in technical analysis.
The trend may be just a visual perception of a price movement over or below a MA. The trend may be just the output of some classification algorithm.

Or,

The trend may be just higher, highs and lower lows, or the trend definition may require a break - out over previous resistance in order to be established.


Some price action school of thought may have another definition of the trend.
For example make a google search: The Law of the Charts by Joe Ross and you will see a nice example. I recommend this free e-book a lot. The difference of this school of price action is that it is focused on objective price action recognition methods.


The fractal dimension is related to other characteristics, that are the fractal geometry of the price - time series. As long as I observe it, that appears to me as an independent characteristics that should be examined together with the volatility. Together they give you some information. The direction (up, down, sideways) of the price time series is another independent characteristics.

There are some patterns of the fractal dimension change, like its ciclicity and break - outs. It is possible to use it in a fuzzy way to make a market classification together with the volatility. You can use it to measure the probability of the next movement, persistent or antipersitent.

The different fractal structure underlines different characteristics of the noise you may expect. Really a lot of things the bad thing is that it is not really organised as ideas as it would be if it was in a book.

This is a different way to look at things that does not replace the technical analysis. Not at all. It is independent and from there comes its strength.

I introduced the entropy indicator too that add another level to the analysis.

The Lyapunov exponent gives another layer of analysis.

The idea is to build a complex picture of the market state.

However I recommend to read the entire threads before making any assumptions for what works or not works.

In the other thread I just do not recommend the combination (especially intra-day) between a lot of money, high leverage, little experience and a lot of greed.

This is a job, some of the best human minds and the most powerfull computers are working on it everyday.

You are not in Kansas anymore.
  • #1,087
  • Edited 3:21pm Apr 13, 2011 3:02pm | Edited 3:21pm
Well

After searching a lot in deep, in mathematical concepts and documentations, I can provide you a simplified Lyapunov indicator, helping you to check out if the market is predictable or chaotic. This indicator is sensitive.

The calculation mode is based on a squared matrix from my own research, the indicator can detect any strange movements or sudden variations, helping traders *not* taking decisions, during a short time, due to heavy risk of trend inversion.

Generally, if the result is negative, the market is predictable, o/w chaotic.
I put some algorythm in case of sudden variations.

I would remind you to install "JJMA" for using the Jurik mode. In this case, the results are more precise.

All the best
jaguar1637
Attached Image (click to enlarge)
Click to Enlarge Name: Check1.gif Size: 38 KB
Attached File(s)
File Type: mq4 Lyapunov.mq4   6 KB | 1,075 downloads
  • #1,088
  • Apr 14, 2011 1:34am Apr 14, 2011 1:34am
Quoting jaguar1637
Well

After searching a lot in deep, in mathematical concepts and documentations, I can provide you a simplified Lyapunov indicator, helping you to check out if the market is predictable or chaotic. This indicator is sensitive.

The calculation mode is based on a squared matrix from my own research, the indicator can...

Thanks Jaguar, very impressing work !

It's possible to get it work with renko charts, because on mine nothing displays, but on regular timeframes it works great !!

Thanks !
  • #1,089
  • Apr 17, 2011 4:00pm Apr 17, 2011 4:00pm
Thanks John Last for all of the time and effort you have put into this thread in order to help open "the doors of perception". I particularly find your version of caterpillar bands 4m interesting as per your previous post

http://www.forexfactory.com/showpost...&postcount=919

I'm wondering if it can be made to work with the non repainting version of SSA as presented by mladen and referenced in post #992 on this thread. If so, would such a modification help to ease up on some of the computing "crashing and bogging" besides overcoming any repainting? Is my logic sound or am I off base on this?
Stay calm, be brave, wait for the signs - Jasper Friendly Bear
  • #1,090
  • Edited 5:51am Apr 25, 2011 5:45am | Edited 5:51am
Well,

As you know, this thread is based on some good assumptions and wonderful intuitions, telling us, a bunch of well-known indicators are not anymore enough efficient to work with.

The market has changed!

My main basis was this week-end to check out w/ logics and mathematics, the purpose of this thread

So, what I did : I used RapidMiner , wrote 3 indicators to retrieve historical datas for checking old indicators

=> I choosed the Vector Linear Regression as algorythm and the EURUSD in 30MN as chart

And the result is astonishing, I can confirm which indicators are obsolete now, and which one are becoming not good enough for working w/

So, this is the list (and you can definivtly remove all those obsolete indicators for ever)

LIST OF OBSOLETE INDICATORS FOR FOREX MARKET TODAY
*******************************************

iAD
iADX
iBWMFI
iCCI
iGator
iMACD
iMFI
iOBV
iRSI
iRVI
iStdDev
iStochastic
iWPR


LIST OF USELESS INDICATORS
*******************************************

iBands
iBullsPower
iDeMarker
iForce
iFractals
iMomentum
iRVI
iStdDev


Yep, Stochastic, RSI, BB, MACD, OBV ... gave wrong or fake values today. Now, this is the proof

That's all folks !

Jaguar1637
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Result_Test_indicators_RapidMiner_VectorLinearRegression.gif Size: 33 KB
Click to Enlarge Name: Test_indicators_RapidMiner_VectorLinearRegression.gif Size: 96 KB
Attached File(s)
File Type: mq4 exoticwavein_test_ind1.mq4   5 KB | 671 downloads
File Type: mq4 exoticwavein_test_ind2.mq4   5 KB | 624 downloads
File Type: mq4 exoticwavein_test_ind3.mq4   5 KB | 682 downloads
File Type: doc Calculation of the CLOSE.doc   31 KB | 768 downloads
  • #1,091
  • Apr 25, 2011 6:01am Apr 25, 2011 6:01am
Hi

I did some other works w/ Polymonial regression (kernel calculation mode) on RapidMiner and the results are quite similar

The market has changed => Happy Eastern !
john
  • #1,092
  • Apr 25, 2011 6:16am Apr 25, 2011 6:16am
Quoting jaguar1637
Hi

I did some other works w/ Polymonial regression (kernel calculation mode) on RapidMiner and the results are quite similar

The market has changed => Happy Eastern !
john
just to ask: what are the results you get, what result is bad(obsolete) and what should be good, why 30 min timeframe, why Linear Regression or Polymonial regression?
  • #1,093
  • Edited 9:24am Apr 25, 2011 7:33am | Edited 9:24am
Quoting Rikers
just to ask: what are the results you get, what result is bad(obsolete) and what should be good, why 30 min timeframe, why Linear Regression or Polymonial regression?
Well,

If you read the codes and watched the images, what I did, I just collected some historical datas from indicators
functions (iCCI, iBands and so on)

The exoctic_Wavein indicator stored all the datas for a selected chart (EURUSD in 30 mn bcoz, I wanted to fetch big waves and why EURUSD ? bcoz the volume is huge there) inside a file .csv

The datas from the Excel file (in .csv) can be analyzed properly by RapidMiner.
[why Vectorial Linear Regression or Polymonial regression?] there are well-knowns (of course polymonial was the oldest algorythms w/ Gaussian calculation) and they are quite fast as algorythms (I tried Relevance Vector Machine, and also, I tried AUTOMLP, but they need too much time calculation. In fact, after 10 hours processing w/o any result, I cancelled the process .)

So, be sure now, you can get rid of RSI , CCI , even Stochastic... and so on

If there were perfect as they should be, the algorythms would have used them and assigned them w/ proper weights values.

Rgds
  • #1,094
  • Apr 25, 2011 8:39am Apr 25, 2011 8:39am
Quoting jaguar1637
....
To my concern, I've noticed that linear regression doesn't work well for non stationary process so the optimal fit won't work. It will work for short time and then brake up.

So the methodology to prove that something doesn't work with something that shouldn't work at the first place....

The second question is that in the place where majority of the "stuff" doesn't work (since there are more loosers then winners) there is a strong bias to get results that show that it doesn't work. I can take a lot of indicators and show that they don't work but there will be only one way they will work (if they work in the first place).

Kind regards
  • #1,095
  • Apr 25, 2011 9:43am Apr 25, 2011 9:43am
Quoting Rikers
To my concern, I've noticed that linear regression doesn't work well for non stationary process so the optimal fit won't work. It will work for short time and then brake up.]

Hi Rikers

Did you ever try the Burg algorythm ? IMHO, it's a linear regression too and it works.

Try the one I wrote, w/ a...
Regards
Attached File(s)
File Type: mq4 Extrapolator_Method_4_jyt_LWMA_Weigthed_37_V19.mq4   18 KB | 1,281 downloads
File Type: mq4 Lyapunov.mq4   7 KB | 1,229 downloads
  • #1,096
  • Apr 25, 2011 9:54am Apr 25, 2011 9:54am
[quote=Rikers;4568680]To my concern, I've noticed that linear regression doesn't work well for non stationary process so the optimal fit won't work. It will work for short time and then brake up.

So the methodology to prove that something doesn't work with something that shouldn't work at the first place....]

Another example of a linera regression indicator,based on polymonial calculation.
This indicator works. It gives you the center of gravity and also, the real bands and the futur channel (if set)

Rgds
Attached File(s)
File Type: mq4 i-Regr.mq4   5 KB | 933 downloads
  • #1,097
  • Edited 10:58am Apr 25, 2011 9:58am | Edited 10:58am
[quote=Rikers;4568680]To my concern, I've noticed that linear regression doesn't work well for non stationary process so the optimal fit won't work. It will work for short time and then brake up.

So the methodology to prove that something doesn't work with something that shouldn't work at the first place....]

Sometimes, you are right... When they are break out or huge reverse w/o notice, the polynomial calculation algorythm as mean squared root, failed . That's why, some traders go ahead in another direction, the digital filters on fractals, the purpose of this thread

If, here everybody thinks I am wrong w/o checking out w/ RapideMiner, I will go somewhere else, in an another forum

john
  • #1,098
  • Apr 25, 2011 4:05pm Apr 25, 2011 4:05pm
[quote=jaguar1637;4568879]
Quoting Rikers
To my concern, I've noticed that linear regression doesn't work well for non stationary process so the optimal fit won't work. It will work for short time and then brake up.

So the methodology to prove that something doesn't work with something that shouldn't work at the first place....]

Sometimes, you are right... When they are break out or huge reverse w/o notice, the polynomial calculation algorythm as mean squared root, failed . That's why, some traders go ahead in another direction, the digital filters on fractals, the...

J-P Poton, Fractals, Technical Analysis and other things..., has a fairly fresh entry in his blog, "The possibility of cognition", and I believe what he writes in paragraphs II and III are very true in terms of what we have been looking for/ought to look for, being the "Fractal aficionado" that I am. What do you think guys?
  • #1,099
  • Edited 6:30pm Apr 25, 2011 6:24pm | Edited 6:30pm
Hi all

I just performed again the Vector Linear Regression w/ RapidMiner on the list of the operators given before, by removing HIGH, LOW, and OPEN

Quite direct !

the result are:

TEST1 :
**********
iBearsPower(Symbol(),Period(), 13,PRICE_CLOSE,shift) is the pivot one

CONFIRMATION WITHOUT OPEN / HIGH / LOW
att5 = - 0.001169 * att6 - 0.000551 * att7 + 0.001390 * att8 - 0.000061 * att9 + 0.000055 * att10 - 0.000053 * att11 + 0.004902 * att12 + 0.924177 * att13 + 0.104506

TEST2 :
***********
Att13 = iMACD(Symbol(),Period(),12,26,9,PRICE_CLOSE,MODE_MAIN,shift), that I found OBSOLETE, is the pivot

CONFIRMATION WITHOUT OPEN / HIGH / LOW
att5 = - 0.001169 * att6 - 0.000551 * att7 + 0.001390 * att8 - 0.000061 * att9 + 0.000055 * att10 - 0.000053 * att11 + 0.004902 * att12 + 0.924177 * att13 + 0.104506

TEST3:
*******
Att13= iStochastic(Symbol(),Period(),5,3,3,MODE_SMA,0,MODE_MAIN,shift), that I found, useless and OBSOLETE, is the pivot one

CONFIRMATION WITHOUT OPEN / HIGH / LOW
att5 = - 0.001169 * att6 - 0.000551 * att7 + 0.001390 * att8 - 0.000061 * att9 + 0.000055 * att10 - 0.000053 * att11 + 0.004902 * att12 + 0.924177 * att13 + 0.104506

________________________________________________________
Also, I did another test w/
iCustom(Symbol(),Period(),"iVAR",n,nBars,0,shift),
iCustom(Symbol(),Period(),"LaguerreFast",gamma,CountBars,0,shift),
iCustom(Symbol(),Period(),"Hurst_Difference_Total_jytv3",f_period,type_data,0,shift),
iCustom(Symbol(),Period(),"rsx new",Length,Price,0,shift),
iCustom(Symbol(),Period(),"PVT",ExtPVTAppliedPrice,0,shift),
iCustom(Symbol(),Period(),"Goertzel_Cycle_V1_1",SquaredAmp,MaxPer,MinPer,bar2update,1,shift),
iCustom(Symbol(),Period(),"Percent_Bollinger_Bands",MA_Period,Deviation,0,shift),
iCustom(Symbol(),Period(),"FFatl" ,Shift,Input_Price_Customs,0,shift)

Result :
***********
for bet iCustom(Symbol(),Period(),"FFatl" ,Shift,Input_Price_Customs,0,shift) is the pivot one
But I choose for FFatl the Input_Price_Customs = 10 => 0.5*TRENDFOLLOW,

(extern int Input_Price_Customs = 0;//
//(0-CLOSE, 1-OPEN, 2-HIGH, 3-LOW, 4-MEDIAN, 5-TYPICAL, 6-WEIGHTED, 7-Heiken Ashi Close, 8-SIMPL, 9-TRENDFOLLOW, 10-0.5*TRENDFOLLOW,
//11-Heiken Ashi High, 12-Heiken Ashi Low, 13-Heiken Ashi Open, 14-Heiken Ashi Close.)

PVT is OBSOLETE as "RSX NEW" !

VectorRegression
att5 = - 0.001169 * att6 - 0.000551 * att7 + 0.001390 * att8 - 0.000061 * att9 + 0.000055 * att10 - 0.000053 * att11 + 0.004902 * att12 + 0.924177 * att13 + 0.104506

That's all folks!
Attached File(s)
File Type: doc Calculation of the CLOSE_WO_HIGH_OPEN_LOW.doc   37 KB | 728 downloads
  • #1,100
  • Apr 26, 2011 8:05am Apr 26, 2011 8:05am
First post

http://forum.mql4.com/ru/40165/page10

( You have to translate it from russian forum )

Really thanks jaguar1637, for your contributions


1 trader viewing now