one to one coaching

Custom EA by stovedude

stovedude

Active member
I've been increasingly aware that we spend so much time looking for EA's that ultimately don't survive forward testing. We see the ones (and have them too) that make multi-millions in a few days, but the fact remains that they will not do that on your account (even demo). A high win rate is also accompanied by either low frequency trades or short duration (scalping) trades. It seems that scalping doesn't always work with great success as we have found out, because the market whips around so much. I believe the real money belongs in the higher TF (ask any fund manager), but I've been wrong before.

I'm devoting this thread to starting our own EA. I just gone done building a simple one for another guy that has a 400% gain in 3 months on backtest and has had winning trades for a month on forward test. It has a 20-60 pip SL (depending on the pair) with 19% DD on 10% risk and only makes 6-7 trades each month. I promised not to share it, but that doesn't mean we can't use similar ideas with decent indicators. It's only based on 2 standard indicators and one custom, which I'm sure we all have anyway. Even with that, I did not spend much time to improve the simple logic, and I still don't understand why it doesn't trade more frequently when all the conditions exist. What I did learn was that with a super low SL, and a high R:R ratio, we don't need a high success rate to make loads of money. We just need consistency, persistence, and a logic that works hopefully half of the time.

The point I'm making is, that with a simple idea great things can be accomplished. I invite everybody to contribute to this project by sharing ideas that we can all consider, but let's focus on the end goal.

I have all the code already. We just need to plug in the ideas to get the ball rolling. I believe simple is better, because extra weight only slows down a winner.
 

rwrz

Member
Hello,

I strongly agree!
I know a little about Harmonic Trading and know that Harmonic Patterns is good system for manual trading... but I think we can create something for Automated too.

Know a lot of things too about forex... but never got good results from forex. So, I can help coding... to MT4 and Dukas if necessary.

Even if need something outside those platforms, I can program too.

Lets get results!

Cya,

RWRZ
 

Tinojazz

Professional Trader Since 2007
VIP Member
Hello,

I strongly agree!
I know a little about Harmonic Trading and know that Harmonic Patterns is good system for manual trading... but I think we can create something for Automated too.

Know a lot of things too about forex... but never got good results from forex. So, I can help coding... to MT4 and Dukas if necessary.

Even if need something outside those platforms, I can program too.

Lets get results!

Cya,

RWRZ
You can't trade Harmonic automatic, You must feel the moment to enter and you must look the other Tme frames. But if you are serious to make EA i am here for you all to give you explanations for that strategy.I am Using already one year, my percent of wining trades is over 70%.
Regards
 

rwrz

Member
You can't trade Harmonic automatic, You must feel the moment to enter and you must look the other Tme frames. But if you are serious to make EA i am here for you all to give you explanations for that strategy.I am Using already one year, my percent of wining trades is over 70%.
Regards
Sure im interested! =) Lets begin.
 

Tinojazz

Professional Trader Since 2007
VIP Member
For begening read the roules, there is a all explanation.Later will talk about strategy.
 
stovedude,
I am not sure about this all.
You are not allowed to share the EA nor the strategy, but we can make suggestions ?

There are so many manual systems out there, where can we start ?

I suggest the Ultimate Forex System (from Jeff Wilde). On some forums it is not allowed to be shared.

It is 100% mechanical and I think it is simple enough to be coded into an EA.

You cannot download the strategy because the site is protected, but you can learn it from the website.
Code:
http://www.forextrendfinder.com/utstest2.html
Let us know if such suggestions is what you are looking for or what else it is.
Thanks.

I made a screenshot (scrolling window) from the site.
You can magnify it in your picture-viewer and get good quality for offline-reading of the strategy.

Sorry, the editing-function has no button for wrapping code-tags around the link.
 

Attachments

stovedude

Active member
I can't share my other EA's, but the ones that we will build here will be shared to all members here, because we are like family here. Nothing will be held back. We will all profit not only in money (I hope), but also in wisdom.

Let's get started...

I just came home now. My time here is GMT-8, so take that into consideration when we are communicating here, because we will all be going to bed at different times, so don't be too frustrated when posts don't come right away.

Now, I'll take a look at the strategies here, and we can discuss them...

I guess we'll need more info on Harmonics. Sounds complicated, but if indicators take care of most of the work, we can handle that, or if its feasible in an EA, we can do that too. Maybe rwrz and Tinojazz can lay down the complete rules here. We can try different strategies, because we don't want to waste time on just one. I think we can code things pretty fast here, because we have more than one coder. I'll get to work on the Ultimate Trading System EA and see how that goes.
 

radicaltour

Moderator
VIP Member
starting to sound exciting, guys!

I have a few general thoughts and suggestions as to what the behavior of the EA may turn out to be like:

(1) use a pattern or system that has 60-70% accuracy rate, which means most of the time the entries will go into positive, giving at least 2:1 risk/reward ratio. I'm sure Tinojazz will agree that trading Harmonic patterns alone will meet this criteria. :)

(2) assumption #1 is really important because if we have such good expectation for a system then we can average-in a few more positions as price continues to go against us. Of course we can make all of the parameters configurable (ie. lot multiplier, max number of positions to average-in, etc...)

(3) when price starts heading in our direction then we can scale-in a few more positions on small retracements to take advantage of the momentum. Again, we'll put parameters in place for things like max number of scale-ins and lot sizes.

These are just a few things I've picked up from other successful manual systems i've come across.
But let's continue the discussion!
 

stovedude

Active member
We can work on that if we have all the system rules. Meanwhile, here is the UTS EA. It might have some bugs still, because my platforms are optimizing other EA's. It needs to be backtested visually to make sure it meets all the strategy rules. Some of the parameters are:

Diff = #pips above/below signal bar to place order
SLDiff = #pips above/below S/R for SL
StopLoss = only used if UseSRStopLoss is false
TakeProfit = only used if UseDynamicTP is false
UseSRStopLoss = uses S/R for SL
UseTrailingStop = standard TS
UseSRTrailStop = resets SL to each new S/R
UseCandleTrail = resets SL to lowest low/highest high of the last 3 bars
UseDynamicTP = uses 0 TP
BreakEvenPips = #pips gained before BE
CloseBar = #bars after entry to close trade
PrevBarFilterPercent = if signal bar is #% larger than previous bar, don't take trade (according to rules)

P.S. Not sure this is the ultimate trading system that we're looking for, but it's a start. Maybe something can be tweaked and optimized to make it successful. If somebody can do that, because my computer is busy right now. I'll try it as soon as I can. If you find any bugs or errors, please post.
 

Attachments

Back testing, I get the error
2010.03.01 02:47 Cannot open file 'C:\Program Files\Prime4x\experts\indicators\SRBarry.ex4' on the EURUSD,M5

SRBarry.ex4
I want to share indicator;)

"SL20 TP10" I think it is a very interesting challenge
 

stovedude

Active member
Sorry about that. Forgot about the SRBarry. Here it is. It doesn't really use it, but you will probably need it for other versions later.
 
I backtested the UTS-EA and barely can tweak it to not lose.
According to the manual strategy description, that system should yield much more than break-even.

I have to test other pairs ( tested EURUSD so far ).

One thing: the time-management does not work. Orders are opened all around the clock although
I limited the time-period for London-session.
 
Im not understading nothing anymore... We will use those EAs to create our custom?!
No,
the UTS-EA is an EA which stovedude coded himself.
The strategy is from Ultimate-Trading-System by Jeff Wilde.

But so far the EA is not profitable, which means that the whole strategy is not profitable - although the author claims that it is.
 

stovedude

Active member
I've found that most strategies are not profitable unless you have other factors involved, such as human intervention (which involves both human error and intuition). Maybe the time filter will weed out some bad trades. A news filter would also help, but I think I would like to try a different strategy and approach. Let me come up with something in a few hours. Does anybody have any good indicators to make into an EA?
 
We have "good indicators" shared in the forum, but we are not sure if they are that good.

I have something which should be easy to be coded
and as usual the author makes good profits ( it is from ForexFactory, not a commercial product ).

The system uses 2 indicators:
- Adaptive Laguerre
- TVI

ALF (Adaptive Laguerre Filter) was created by John Ehlers, a true visionary who adapts Digital Signal Processing techniques for use in futures trading. You may want to visit his website at . In the Technical Papers section of the site are many interesting and free examples of his work. ALF functions as a continuously self-adjusting overbought/oversold line. Buy when price goes above the line, sell when price goes below (similar to a Stochastic except that ALF is both the 20 and the 80).

TVI (Tick Volume Indicator) was introduced by William Blau in his book "Momentum, Direction and Divergence". The book contains many other excellent indicators and delves deeply into smoothing techniques that introduce only minimal lag. TVI measures the relationship between incoming ticks for a given volume. If upticks outnumber downticks the TVI rises, and vice-versa.

The thread is here:
Code:
http://www.forexfactory.com/showthread.php?t=226154
My suggestion:
We use the CCI 20 as an additional filter.
CCI 20 (Commodity Channel Index, Period 20, Typical Price)

Rules for entry:

LONG: Price-action above ALF (Adaptive Laguerre Filter) + TVI = green
+ CCI 20 above Zero-level

SHORT: Price-action below ALF + TVI = red + CCI 20 below Zero-level

EXIT: no fixed rules ( bad for an EA, I know ), therefore options for
Trailing stop, fixed TP , cross of Moving-Averages etc....
 

Attachments

stovedude

Active member
I'll get to work on it. Looks like the TVI_2color repaints the previous bar, so we'll just use TVI, ALF, and CCI with adjustable inputs, along with the time filter.

Update: Finished, but with the CCI filter, it is essentially only a CCI EA, because the TVI and ALF always confirm when CCI crosses 0. It might be able to scalp a couple of pips (will need to modify it to scalp less than your stoplevel) on 5m or 1m.
 

Attachments

rwrz

Member
Artificial Inteligence? Neural Networks? Genetic Algoritms?
( to study a little:
Who wants to enter on that idea?

I like to code Artificial Inteligence for games... and I think is the same for market. Im creating an IA API on Java for jForex, and could be easily ported to MQL4.

I was thinking on Inputs and Output. I have tried supervised training and it doesnt works because this random market. I was thinking on create some mixed its learning with Genetic Algoritms, the idea is:

  • Create a big population of agents
  • Those agents, will can do 3 things:
    1. Nothing
    2. Buy
    3. Sell
  • The algoritm that will choose WHAT TO DO will be a Neural Network.
  • We wont force MAX TRADES, we will let this agent open freely how many trades it want.
  • After a year or when every agent lost its money, we will finish this GENERATION and will get the best ones:
    1. who made the best money
    2. who stayed alive most
    3. who losted lesser
    4. there is a possibility of FITNESS calculation (here say more about it: ), we can choose one agent per his actions... but need to focus on what actions! hehe
  • then we make a crossover with its neural networks "weigths"
  • try again until get better agents
Create a big population of agents.

This is my idea... and I was thinking on the better inputs for those agents LEARN to take their decision:

  • SMA 200
  • SMA 100
  • SMA 50
  • SMA 25
  • SMA 15
  • SMA 10
  • SMA 5

Why only SMAs? Well, on one SMA I have as many values as I want to.... So, I can resume a big amout of values in just one value... this can reduce the numbers of inputs... What do you, guys, think about it?

Lets try?

Cya
And sorry about my poor english...
 

stovedude

Active member
Yes, we can explore this area. Go ahead and lay out a basic plan to start with. Granted, it is complicated, and like you said, it is a random market, and as we all know, a random market does not follow a system, but rather a system follows a random market (usually blindly).

Let's get to work on it, and I'll try to find some other options soon.
 
I'll get to work on it. Looks like the TVI_2color repaints the previous bar, so we'll just use TVI, ALF, and CCI with adjustable inputs, along with the time filter.

Update: Finished, but with the CCI filter, it is essentially only a CCI EA, because the TVI and ALF always confirm when CCI crosses 0. It might be able to scalp a couple of pips (will need to modify it to scalp less than your stoplevel) on 5m or 1m.
stovedude,
you are a great coder and very fast.

I have backtested the EA visually and it is not doing what it should.
I modified the code to get better entries, but since I don?t know mql , the results were mixed.
I cannot tell it to do what I want - but you can.

I will make screenshots and explain where and why it should behave differently .
But now I have to get some sleep.
Regards

P.S.: If several EAs are discussed, the thread soon will be a mess.
If possible, a MOD should create sub-threads for each EA, then it is more organized and better to read.

@stovedude,
maybe it is a good idea to code the millionaire-strategy.
That german guy Sebastian who sold it, traded a 5k live-account to 100k within 5 months.
So we can say that it is a proven strategy.

Actually , he uses at least 2 strategies, one with doubling of lot-sizes when SL is hit, the other with fixed lot-sizes. If possible, both strategies should be compared in backtests of the EA ( or 2 EAs ).
The basic idea is the same, chosing a range and trading the breakout.
 

sandman4x

New member
I've been increasingly aware that we spend so much time looking for EA's that ultimately don't survive forward testing. We see the ones (and have them too) that make multi-millions in a few days, but the fact remains that they will not do that on your account (even demo). A high win rate is also accompanied by either low frequency trades or short duration (scalping) trades. It seems that scalping doesn't always work with great success as we have found out, because the market whips around so much. I believe the real money belongs in the higher TF (ask any fund manager), but I've been wrong before.

I'm devoting this thread to starting our own EA. I just gone done building a simple one for another guy that has a 400% gain in 3 months on backtest and has had winning trades for a month on forward test. It has a 20-60 pip SL (depending on the pair) with 19% DD on 10% risk and only makes 6-7 trades each month. I promised not to share it, but that doesn't mean we can't use similar ideas with decent indicators. It's only based on 2 standard indicators and one custom, which I'm sure we all have anyway. Even with that, I did not spend much time to improve the simple logic, and I still don't understand why it doesn't trade more frequently when all the conditions exist. What I did learn was that with a super low SL, and a high R:R ratio, we don't need a high success rate to make loads of money. We just need consistency, persistence, and a logic that works hopefully half of the time.

The point I'm making is, that with a simple idea great things can be accomplished. I invite everybody to contribute to this project by sharing ideas that we can all consider, but let's focus on the end goal.

I have all the code already. We just need to plug in the ideas to get the ball rolling. I believe simple is better, because extra weight only slows down a winner.
Hi stovedude. This is really great trying to put our efforts together to create an EA.

I have put together a project on google code named "Worldwide-Invest" and I put a library that we can add usefull functions to going forward.

I think it's really important that we have re-usable library.

We will be able to checkin/checkout using svn repository.

Lets put there all indicators we will be using and the EA.

I already put the news filter as part of the indicators.

Please anyone participatiing, PM me to give you full access to this project.
 

stovedude

Active member
We can look at the millionaire strategy too. What do you think about an EA based on this indicator? I'll get to work on this one now and see how it goes on 5m TF. It should keep us in the trend longer. I'm not crazy about scalping right now, so we'll explore that later. Meanwhile, I think we'll make a new thread on the main page and then we can do subthreads from there if MJ approves.

Edit: Ok, here is the first draft. Same code as the prev EA's, but with the doubled stoch instead. Let's see if we can optimize and improve entries and exits. Also, the indicators are attached. I call this one ForexDevastator1.5 (prev versions are still in progress - don't know if this name exists, but it sounded good).

Edit: Removed EA to protect from marketers. PM me if you are interested.
 

Attachments

one to one coaching
Top