Interesting GBP/USD MA Crossover system | Page 7 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Need help with MA crossover system 15 replies

Interesting GBP System 77 replies

A "different" crossover system - 600 pips on Friday alone in EUR and GBP 48 replies

Interesting Crossover - Am I seeing things? 11 replies

Scalping with crossover system 0 replies

Attachments: Interesting GBP/USD MA Crossover system
Exit Attachments

Interesting GBP/USD MA Crossover system

  • #121
  • Edited 12:56am Mar 8, 2007 12:46am | Edited 12:56am
Quoting Nupe1911
This will be helpful so as more of us have this skill set, we all can work more collectively in Optimizing and Testing this EA.
Here maybe a good start reading:
http://www.strategybuilderfx.com/sho...t=optimization

More detail information:
http://www.strategybuilderfx.com/sho...t=optimization

You may also reading this book for overall concept:
Design, Testing, and Optimization of Trading Systems
ISBN: 0471554464

Refer to MT4 official site for step-by-step tutorial:
http://articles.mql4.com/192

Programming consideration in testing:
http://articles.mql4.com/72

If you are interesting about internal algorithm of tester:
http://articles.mql4.com/134

A sample of complete MT4 EA development cycle:
http://articles.mql4.com/163


Since there are too many things to be care when you are doing optimization, I think here are some things you must keep in mind:

1. Get historical M1 tick data as long as you can.
2. Watch out the timezone difference between your M1 tick data and your broker's server. Difference timezone may produce difference result sometime.
3. Make sure each testing produce 90% modeling quality result, and don't trust any result if MQ<90%.
4. Beside to M1 tick data, make sure you have also download all timeframe (M5, M15, H1, Hxxx.....) data from your broker. If the downloaded data cannot cover your testing period (ex: you want test 2 year, but broker only provide 6 months data), you may produce all timeframe data from your M1 historical tick via period_convert script which build-in with your MT4.
5. Make sure your computer has fastest cpu and largest memory, or you may waste you life....
6. Do not use OVER-OPTIMIZED result (someone call it CURVE-FITTED). Since over-optimized result only perform good to historical data, and will make very bad performance in live trade.
7. More longer testing period (ex: more than 3 year) will produce more reliable result.

The hardest part is: how do I know which result is over-optimized?

I have no answer too, but if your result can be used with difference pairs (or testing period) with only a little parameter shifting, that may be a good result. In other word, if you get very difference result with difference data pattern, or get very weird parameter number, especially if you get very HIGH PROFIT with only FEW TRADE times, that may be a very bad result.

By the way, someone's optimization result may not fit to your live trade style, since you may using: difference historical data, difference broker timezone, difference broker price, difference risk strategy.....I have found many times that someone published optimized parameter cannot produce same result with my backtesting.
- Ray Tracy - Novice forex trader from GMT+8
  • #122
  • Edited 7:43am Mar 8, 2007 1:43am | Edited 7:43am
New version -dsy1.2 released as following:

1. some speed improvement
2. Add Normalization to OrderClose function.
3. Add TF external variable for more flexible trial.
4. Change exist trade search from ascending to descending. Since ascending search will produce a bug if there are multiple trades generated by this EA.

Please refer to release post #114 for attachment.

By the way, I won't change any original idea or design from Aicccia, I just doing some things which let this EA worked faster and smother. It is not necessary to bind my code into the Aicccia original one. I do it only for my private usage, and I may made any unexpected mistake inside the modification, please use as your risk.
- Ray Tracy - Novice forex trader from GMT+8
  • #123
  • Mar 8, 2007 1:52am Mar 8, 2007 1:52am
I would suggest that you speak to Vegas for some ideas, seeing that he has been using the tunnel method for more than 19 years using 12,144,169 ema's

just a thought
  • #124
  • Mar 8, 2007 3:12am Mar 8, 2007 3:12am
Raytracy, thank you so much for adding these little extras to the EA. I don't have much experience in computer programming so I sometimes skip things in the EA that aren't relevant to the strategy but relevant to the EA as a little computer program.

You probably don't relize this but the only places in the whole EA where I call on Bid or Ask is when placing an order, and I only place an order at the start of a new candle. All my other variables run off the last close. This is very important in the backtesting because it means you can choose the "Open Prices Only" model when backtesting and optimizing, since my EA doesn't look at what goes on between a candle's open and close. For me to backtesting over half a year's worth of data on the "Every Tick" model(the slowest) takes about 40 seconds. To backtest on the "Open Prices only" model (the fastest), it takes less than 1 second. If you don't believe me, test the strategy over a period using both modeling types and you'll get the same exact results.

But thanks again for the additions. Hopefully the whole thing will be done by tomorrow.
  • #125
  • Mar 8, 2007 7:38am Mar 8, 2007 7:38am
Quoting aicccia
To backtest on the "Open Prices only" model (the fastest), it takes less than 1 second. If you don't believe me, test the strategy over a period using both modeling types and you'll get the same exact results.

But thanks again for the additions. Hopefully the whole thing will be done by tomorrow.
No, dear Aicccia, You have did great job than me. I have said this is my private modify, since it running very slow in my elder computer and got some trouble with my tick data. I believe that running well with you, and hope you continue to improve this EA.

I am a newbie in FX, and just like to share some programming skill with everybody. I have not try to interfere with you or your EA. I am appreciate that you have share your system with us, and just like to make some help if someone like it.

You are very nice guy, and I would like to see your next version. Please don't let me interrupt your great job.
- Ray Tracy - Novice forex trader from GMT+8
  • #126
  • Mar 8, 2007 7:52am Mar 8, 2007 7:52am
aicccia, based on your original system, how may trades should have the EA taken in 2007?
By the way thanks for the updated versions, they're working fine for me! Great!
  • #127
  • Mar 8, 2007 8:37am Mar 8, 2007 8:37am
Raytracy,

Thanks so much for the optimization articles, I will start reading these and get familar with optimizatiuon process.

This information is valuable as look to try to aid in contributing to this EA.

Theo
  • #128
  • Mar 8, 2007 8:56am Mar 8, 2007 8:56am
Quoting Nupe1911
This information is valuable as look to try to aid in contributing to this EA.
You are welcome!

By the way, aicccia has made a good strategy to design this system which worked with "Open Prices only" model. You may find that backtesting or optimized with this model are super fast and still accurate.

I believe many other EA cannot get accurate result with this model, most of them must deal with "Everytick" model only.

I did not realize about it at first, until Acccia remind me in above post. This design save all of our time. Thanks to aicccia!
- Ray Tracy - Novice forex trader from GMT+8
  • #129
  • Edited 11:30am Mar 8, 2007 11:16am | Edited 11:30am
Quoting MarkGavalda
aicccia, based on your original system, how may trades should have the EA taken in 2007?
By the way thanks for the updated versions, they're working fine for me! Great!
13 trades as of this morning.
  • #130
  • Edited 11:39am Mar 8, 2007 11:37am | Edited 11:39am
Quoting aicccia
13 trades as of this morning.
Okay then I got everything right. I've made a little change to the code, because the automatic lot size determination didn't work for me, only the fixedlot one. I made the change to the version of raytracy, should I post it, or just toy with it myself?

I'm currently testing if the every tick - open price are really the same, if yes, that would be great I haven't looked into the code deeply enough yet, I guess the EA is only using completed bars, that's why "open price" method is enough.

EDIT: Ok i just read through the posts, and found that this has been already answered. Great work guys!
  • #131
  • Mar 8, 2007 11:50am Mar 8, 2007 11:50am
Quoting aicccia
. The one big fault I have with this systems reliability is that all of this is based on just 16 trades. But they're long, logical trending trades so I think it's okay.
Hello Aicccia,

I have just started reading this thread but you are right, 16 trades is definitely not enough if you want to judge the profitability of a system, even if it took 50 years for a system to generate these 16 trades.

I have seen FX systems make tons of cash after 300 trades, only to give it all back the next 50 trades, even though the logic behind these systems looked absolutely perfect and sound.

Ok, I am going to read the rest of this interesting thread now...
  • #132
  • Mar 8, 2007 1:05pm Mar 8, 2007 1:05pm
I have made some quick backtests to optimize the performance, the best result was +3266 pips from 1st September 2006 to 8th March 2007 (last trade was closed on 6th March). Total trades: 35.
I think this is the maximum of this system for this period, I ran 75 937 500 000 variations (genethic algorythm of course ).

The biggest and most important question is that how often should one optimize this system to achieve maximum (or at least medium) live results? And how long the sample period should be?
If anyone can answer these simple-looking questions, we're on our way to get rich, automatically....
  • #133
  • Mar 8, 2007 2:19pm Mar 8, 2007 2:19pm
Quoting MarkGavalda
I made the change to the version of raytracy, should I post it, or just toy with it myself?
If don't mind, you may tell me the changed part or code, I would like to integrate into my version, and then publish it to all. Maybe Aicccia would also like to do the same change.

Or, you may also publish your complete version, just remember changing the filename to reflect your modification.
- Ray Tracy - Novice forex trader from GMT+8
  • #134
  • Mar 8, 2007 2:22pm Mar 8, 2007 2:22pm
Quoting MarkGavalda
I ran 75 937 500 000 variations (genethic algorythm of course ).
By the way, I do not understand what do you mean about the number? Would you like to teach me more? Is it about external variable? But I have found 5 related external, not 4?
- Ray Tracy - Novice forex trader from GMT+8
  • #135
  • Mar 8, 2007 3:10pm Mar 8, 2007 3:10pm
Quoting raytracy
By the way, I do not understand what do you mean about the number? Would you like to teach me more? Is it about external variable? But I have found 5 related external, not 4?
It means the number of setups I ran to get the best result, (75 thousand billion different variations of the external variables!), but I think I don't need to teach you anything, you already know much more than me, I just read through the links you posted below, for a long time I was searching for a genetic algorythm explanation, thanks for that link -- but to tell the truth I don't understand a word from that article

By the way I changed the original code to this one:
Inserted Code
if (fixlots==0)    lot=NormalizeDouble(AccountFreeMargin()*MaximumRisk*0.001/100,1);
The "100" at the end should be an external variable and it was meant to be the user-set maximal Stop Loss, but I hardcoded my decision instead. You can change that of course.

I still can't figure why doesn't the original want to work for me!? I can't think of anything that works on one platform and doesn't work on an other!? Anyways, this is 100% works on every platform.

Keep up the good work guys, this system has a very good potential! (And I'm really curious about that optimization question I brought up earlier! What is your opinion, guys?)
Have a good night everyone!
  • #136
  • Mar 8, 2007 4:04pm Mar 8, 2007 4:04pm
Quoting MarkGavalda
It means the number of setups I ran to get the best result, (75 thousand billion different variations of the external variables!),
OK.....I did not realize that you running so much variation at one pass.....Since I often separate them into several difference group to help me get a quick glance.

You have created a good idea about risk management. I just start to thinking about how to reduce or limit the drawdown and stoploss with this EA. I will coding you idea into my code as next version. Thanks for this great thought.

By the way, maybe Aicccia will have difference thought regarding your idea. I would also like to hear any advice about stoploss from Aicccia.
- Ray Tracy - Novice forex trader from GMT+8
  • #137
  • Mar 8, 2007 6:24pm Mar 8, 2007 6:24pm
Hello All,
My first post as a FF member!

I'm new to Metatrader, and like the idea of autotrading.

I've read this thread, downloaded the ea's and found the same good results.
When I look thru the trades the ea makes, there are a few that don't make sense to me according to aicccia's descriptions.

I've attached a chart showing one of the trades. It seems that when a trade goes in the wrong direction, sometimes (not all) it goes very wrong and many pips lost. Obviously I don't completely understand how the exit is supposed to occur, but maybe the chart picture will allow someone to see a potential fix for this problem.
Attached Image
  • #138
  • Mar 8, 2007 7:15pm Mar 8, 2007 7:15pm
Quoting jackinabox
Hello All,
My first post as a FF member!

I'm new to Metatrader, and like the idea of autotrading.

I've read this thread, downloaded the ea's and found the same good results.
When I look thru the trades the ea makes, there are a few that don't make sense to me according to aicccia's descriptions.

I've attached a chart showing one of the trades. It seems that when a trade goes in the wrong direction, sometimes (not all) it goes very wrong and many pips lost. Obviously I don't completely understand how the exit is supposed to occur, but maybe the chart picture will allow someone to see a potential fix for this problem.
Jackinabox, I can't see the dates on that window, when was that trade?
  • #139
  • Mar 8, 2007 8:14pm Mar 8, 2007 8:14pm
Quoting MarkGavalda
I still can't figure why doesn't the original want to work for me!? I can't think of anything that works on one platform and doesn't work on an other!? Anyways, this is 100% works on every platform.
On a full size account most brokers consider 1 lot = 100k and if offered .1 lots to be a mini lot = 10k. There are accounts from some brokers where .1 lots = 1k. IBFX for example uses both types depending on if you get a standard or mini account. Several other brokers trade like the IBFX mini using 1 lot = 10k and .1 lot = 1k etc. for all their accounts.
  • #140
  • Mar 8, 2007 8:49pm Mar 8, 2007 8:49pm
I did a backtest from 6/1/06 to now. This trade was on 6/21.
I know the ea isn't optimized for that far back.... My thinking was that maybe you can get some added information by analyzing the results without optimization?? I don't know if that makes any sense or not.

I only brought it to attention to see if an improved exit strategy on losing trades can be thought out.

Maybe something like- If the entry was for a buy and price is above all 4 ema's then any time price crosses below all four ema's get out of the trade... Just a thought.
0 traders viewing now