Old Dog with New Tricks | Page 51 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Old Dog's Taming of the Beast? 34 replies

Old Dog Fib Trading - EA Development 18 replies

Old Dog Fib Trading - EURUSD 399 replies

5 Tricks of Trend Trading, by Rob Booker 25 replies

How To Work Around A Market Maker's Tricks 0 replies

Attachments: Old Dog with New Tricks
Exit Attachments

Old Dog with New Tricks

  • #1,001
  • Apr 30, 2011 3:57pm Apr 30, 2011 3:57pm
This is an amazing piece of code because it works with statistics. If you know anything about quantum mechanical math, statistics rules probability. But in my honest opinion, even after I broke down the optimization to file the stats into block (small time-frame divisible by bars in one day), day of week and month of year. There was still more to the market that makes it difficult to use straight logic. So I now work with mostly neural networks and AIs so I can gauge more parameters at once. I still use a lot of this code, just not so much the HiLow strategy of it.

I hope that answers your question
  • #1,002
  • Apr 30, 2011 10:12pm Apr 30, 2011 10:12pm
thanks, I appreciate your answer. I asked you because you gave it a lot more thought into it than most of us.
  • #1,003
  • Apr 30, 2011 10:29pm Apr 30, 2011 10:29pm
Hey all I'm new and like the other new guy above me, I also can't hardly think to read this whole thread. However, I did write a self-optimizing EA in the past. It didn't really work very well. Probably because the underlying strategy wasn't too good. But it backtested every day, on the first tick of the day, kind of like running an "optimization" on MT4. It would then switch to whatever settings it found to have worked best in that prior day. But, again, it didn't work because I guess the underlying strategy wasn't that great. And, perhaps, because the market doesn't do the same thing for two days in a row lol. Long term strategies don't need constant reoptimization in my humble opinion, because of the market doesn't change behavior so often I suppose, at longer time frames. And now I trade mostly longer time frames. I assume that this code here in this thread was more for shorter time frames? Anyway good luck to all...
  • #1,004
  • May 1, 2011 1:15am May 1, 2011 1:15am
Quoting rfking2
This is an amazing piece of code because it works with statistics. If you know anything about quantum mechanical math, statistics rules probability. But in my honest opinion, even after I broke down the optimization to file the stats into block (small time-frame divisible by bars in one day), day of week and month of year. There was still more to the market that makes it difficult to use straight logic. So I now work with mostly neural networks and AIs so I can gauge more parameters at once. I still use a lot of this code, just not so much the HiLow...
This EA is profitable though right? I recall you saying you hadn't had a losing week with it at one point. In your opinion is it worth the time and effort and computer resources to run and find consistently profitable settings?

Can you explain what modifications you have made with the EAs above?

Thanks
  • #1,005
  • May 1, 2011 1:42am May 1, 2011 1:42am
Yes it can be profitable and absolutely worth your time. The ability to self optimize is golden and should be included with any EA you code. RR is brilliant for creating it. As for my mods, I encourage you to reverse engineer the code. That is the only way you will learn to code yourself. I worked on it for so long that its hard for me to remember it all. If you have any specific questions, I will be more than happy to help.
  • #1,006
  • May 1, 2011 6:57am May 1, 2011 6:57am
I put the EA in my MT4 but when i run a backtest i get no errors and no results.
  • #1,007
  • May 1, 2011 11:45am May 1, 2011 11:45am
Quoting sissi
I put the EA in my MT4 but when i run a backtest i get no errors and no results.
Please read this thread and inform yourself. Don't be lazy.
  • #1,008
  • May 2, 2011 2:41am May 2, 2011 2:41am
Quoting iulli
Please read this thread and inform yourself. Don't be lazy.
I am not lazy, but english is not my mothertounge.
  • #1,009
  • May 2, 2011 4:08am May 2, 2011 4:08am
Quoting rfking2
Yes it can be profitable and absolutely worth your time. The ability to self optimize is golden and should be included with any EA you code. RR is brilliant for creating it. As for my mods, I encourage you to reverse engineer the code. That is the only way you will learn to code yourself. I worked on it for so long that its hard for me to remember it all. If you have any specific questions, I will be more than happy to help.
Thanks a lot for this EA. Could you please post the set file of the settings that you use for it? This would give a good place to start with demoing it.

Many thanks
The best trade I ever made was my SIN for Jesus' RIGHTEOUSNESS.
  • #1,010
  • Edited 3:01am May 3, 2011 2:11am | Edited 3:01am
In order to backtest you must first optimize then copy the "optimized settings" from mt4-experts-files to mt4-tester-files and turn "reopt" off when testing.

If you read the optimized settings file you will notice it files by month of year, day of week and block of day. So reopting more than once a week is pointless.

Im sorry, I cant give you a set file because I don't know what you want to trade or how you want to trade it. You must backtest and find your own settings.


EDIT: By the way, the only way you will make this work is to completely quarantine your data to your broker. That means delete all your metaquotes data and only download from your broker!!
  • #1,011
  • May 22, 2011 8:30pm May 22, 2011 8:30pm
I've been playing around with the settings for a while now, however it still has not given me a trade yet with HiLo Self Opt.

My settings are simple: everything default except for Bars to Optimize = "50000" (my pc can't handle too much bar data), Min WinRate="60%".

My demo account is with Gomarket metatrader, so I'm not sure if they're an ECN broker or not. Therefore, I ran with the default settings set to No to ECN broker.

One time, Trade style said breakout. There was even something at Trade Trigger other than 'None'. However, no trades were opened. Did it not satisfy the Min WinRate?

Would someone be able to explain to me what I'm missing? Thanks.
  • #1,012
  • May 22, 2011 8:39pm May 22, 2011 8:39pm
Here is a pic of when I ran it with fxcm demo. Default settings except for Bars to Optimize = 50000, Min WinRate = 55%, Min SuccessScore = "150".

I'm wondering why it says -100% for the Current WinRate and -100 for Current SuccessScore? Is this expected when EA doesn't think the market meets the trading criteria?
Attached Image (click to enlarge)
Click to Enlarge Name: Untitled.png Size: 26 KB
  • #1,013
  • May 23, 2011 4:11am May 23, 2011 4:11am
Quoting idash
Here is a pic of when I ran it with fxcm demo. Default settings except for Bars to Optimize = 50000, Min WinRate = 55%, Min SuccessScore = "150".

I'm wondering why it says -100% for the Current WinRate and -100 for Current SuccessScore? Is this expected when EA doesn't think the market meets the trading criteria?
I Guess it means there isn't a stratagy found which holds your criteria (you can see it at the bottom of the feedback: there is no trading style. Have been playing with the settings but haven't been able to found any profitable settings/timetrafes yet. All are giving around 50$ win rate with negative (smaller then 1) R:R ratio.

Weird because when i look back to pics here people did seem to get good R:R's
  • #1,014
  • May 23, 2011 7:17pm May 23, 2011 7:17pm
Quoting idash
Would someone be able to explain to me what I'm missing? Thanks.
You need to read the "Optimized Settings" excel file that is created after you optimize. That is where you will find how often it can trade with your settings.

Also, if you downloaded any ticks from the history center, your data is false, and it will be impossible to find good settings. The only data that matters is from the broker you are using at the time.
  • #1,015
  • May 23, 2011 8:25pm May 23, 2011 8:25pm
I am also testing HiLo and Month Opt. I am a little confused about something. for Month Opt, I optimized a pair using OOTB settings and had pretty good success score. I would then try to save those settings by re-applying the EA with ReOpt turned to false. But eventually, the optimization setting would change (i.e., turns to Current Win Rate 1100%, Risk Reward -1, Success Score of -100). I didn't change any setting aside from ReOpt. I suppose that is not the best way to save an optimization?

Anyway, I don't know if this will prove profitable but it is definitely interesting and fun. RR and RF, thanks for your contributions.
  • #1,016
  • Jun 2, 2011 10:51am Jun 2, 2011 10:51am
Quoting idash
Here is a pic of when I ran it with fxcm demo. Default settings except for Bars to Optimize = 50000, Min WinRate = 55%, Min SuccessScore = "150".

I'm wondering why it says -100% for the Current WinRate and -100 for Current SuccessScore? Is this expected when EA doesn't think the market meets the trading criteria?
idash,

Did you ever get this to work? I am having the same issues using no minimums so I would expect at least a few trades.
  • #1,017
  • Jun 3, 2011 2:52am Jun 3, 2011 2:52am
If you are using month opt, open up the optimized settings folder. Look at the first 3 numbers in each line, the first number is the month of year 0 = january, the second is the day of the week and the third is the block of day. So that means if it says 0;0;0, that line is the strategy for the first hour Sunday morning in January.

EVERY NEW LINE IS A DIFFERENT STRATEGY TO ACCOUNT FOR EVERY BLOCK OF EVERY DAY OF EVERY MONTH OF THE YEAR!!!
  • #1,018
  • Jun 3, 2011 3:16am Jun 3, 2011 3:16am
Thanks for sharing your work. Much appreciated.
  • #1,019
  • Jun 4, 2011 12:27pm Jun 4, 2011 12:27pm
  •  dyos
  • | Joined Jul 2007 | Status: Automating forex strategies daily | 48 Posts
Jed,

I have a question about the hour optimization of this strategy. It appears it just simply queries the server for the hour of the bar.

if(HourUsed == TimeHour(TimeCurrent()))

However, it doesn't appear to account for DST adjustment at all. I have read in some of your previous posts that you use ForexTester.com for your data source which does not account for any type of DST adjustment either. The only broker I'm familiar with is Oanda's data where they adjust because they are based on the east coast.

The reason I bring this up is I have built strategies in the past and seen completely different results when using non-DST adjusted data. When building a strategy that is built on using specific trading times.

Below is a short snippet of my code that does this for me:

else if(DayOfYear()>beforeDST && DayOfYear()<afterDST){
if(HourEST+DSTOffset>23){
if(Hour()==HourEST+DSTOffset-24 && Minute()==0)
set_Prices();
}
else if(Hour()==HourEST+DSTOffset-1 && Minute()==0){
set_Prices();
}
}

//+------------------------------------------------------------------+
//| Set the days of the year when DST chanes |
//+------------------------------------------------------------------+
void get_DST(){
switch(Year())
{
case 2001:
beforeDST=91;
afterDST=301;
case 2002:
beforeDST=97;
afterDST=300;
case 2003:
beforeDST=96;
afterDST=299;
case 2004:
beforeDST=94;
afterDST=304;
case 2005:
beforeDST=93;
afterDST=303;
case 2006:
beforeDST=92;
afterDST=302;
case 2007:
beforeDST=70;
afterDST=308;
case 2008:
beforeDST=68;
afterDST=306;
case 2009:
beforeDST=67;
afterDST=305;
case 2010:
beforeDST=73;
afterDST=311;
case 2011:
beforeDST=72;
afterDST=310;
}
}
//-------------------------------------------------------------------

Anyways I've been meaning to write a "self optimizing" strategy myself before and I just haven't found the time. But the work you put forth in this post is very inspiring and means I do not have to start from scratch. Mine will be based on sets of moving averages. But I like your idea of taking the highest high of a given period and the lowest low. It is definitely a crude way of analyzing but it sounds like it has worked for you.
  • #1,020
  • Jun 5, 2011 11:32pm Jun 5, 2011 11:32pm
I agree DST adjusted data is better.

Problem becomes how do I account for the two 2:00AMs and the lack of a 2:00AM twice a year?

For now, I've avoided that issue by simply using relatively small backtesting periods.

DST shifts tend to affect the markets for a week--two at most. I suspect this is driven mostly by the new sleep schedule traders need to accommodate. I would like to learn what advantages and opportunities you're seeing with DST-adjusted data.

I was hoping to use the law of large numbers to my advantage in this EA concept. The problem with this version is accounting for spikes and other abnormal market events. I've been especially fortunate to have perfectly captured three major market shifting events. (Tsunami, Reactor issue, Osama). So this year has been especially fruitful for me.

I'm currently working on an update to this release to

 

  1. Clean up the code--A lot
  2. Allow for more variables to be tested in addition to breakout/counter and hour.
  3. Provide a better UI to explain to the user exactly what the EA is doing.
  4. Move away from a file-based date transfer to array-based.

Current prototype takes about 2 minutes to process 40,000 bars and about 15k permutations of systems.

Also, you mentioned you are trying to self optimize sets of MAs, here's why my code won't work with that specific application.

During the initial pass, the EA is scanning through the bar history looking for trade events. Because trade events are based on the day's high and low, there is a fixed number of events in the bar history. This fixed set of events allows me to only have to do one pass.

When using multiple MAs, the number of trade events changes based on the values of each MA. This requires one pass per unique set of values, and can quickly become very resource intensive.

You'll need to modify the code to account for the multiple passes required if you want to write your own self optimizing EA.

0 traders viewing now