collapse

Author Topic: EA Controller  (Read 8418 times)

Offline donnaforex

  • Administrator
  • Hero Member
  • ******
  • Posts: 4155
EA Controller
« on: March 04, 2017, 10:01:54 AM »
This is a different kind of EA, brought to us by the same company that created Autotrader Elite.

Visit Controller EA website

Note: This is not a 'normal' EA, it's more like a trading tool. It does come packaged with 4 EA's to use with it but the applications of the EA extend way beyond just this and I feel it will interest more advanced EA traders. I have described more below.

Donna's live myfxbook stats, using the provided EA's

Here is the demo account that feeds into the live one (see below for why this is needed).

Important note: I feel this one is NOT for newbies in any way, not because of the way it trades, but because 1. It is tricky to set up correctly, and 2. It is most useful if you already have a collection of EA's you can use it with (controller will really come into it's own i believe if you have a number of EA's which are very 'up and down' in their performance).

EA Controller is an indicator which monitors your trading. It creates a moving average of your equity curve. If the equity curve is currently above it's own moving average, then it allows trades to be copied to a trade account. If the equity curve is currently below it's own moving average, it stops trading.  You need two accounts to use it - a demo (usually) for monitoring purposes, and then a live account where you get your filtered trades happening.

The idea is, if your system is currently performing well and gaining then EA controller allows these trades to be copied, and if your system is going through a period of drawdown, it cuts off the trades and doesn't allow them through to your live account.

For obvious reasons, this is not suited to every EA. For instance, a grid type system that opens multiple trades over long periods of time that hang around for weeks or months is not going to work well here. But, a system that gets in and out of trades over the short term will work better. Also, systems which have stretches of winning, but then long periods in drawdown, will excel here since controller will filter out the large drawdown sections. There are other subtleties here which come into play too, but if you have been trading for a long time you will have an idea of what strategies might work with this (or make a post here, we'll help you).

If you are still confused after reading my description, take a look at their website which has some more visual representations of how it works. Like i said, this one is tricky to operate so bear that in mind before plunging in!


Follow DonnaForex on Facebook and Twitter.
 
If you need to contact me, email admin@donnaforex.com or Skype 'donnaforex' or PM via the forum.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #1 on: March 04, 2017, 11:48:19 AM »
Donna's test will take a while before we have enough trades to see what is happening, so I thought I would share the results from my first month of trading with the EA Controller system. Trading the four bonus EAs that come with the package.

(When you use these particular EAs, you don't need to use a Trade Copier to stop or start trades being copied to your live account. Instead the EA Controller on the demo account sends the Yes/No signal for trading directly to a copy of the EA on the live account.)

I started this test on 1 Feb. So I am running these 4 EAs on both a demo account and on my live account; and then added the EA Controller to the charts of the EAs on my demo account to manage/control the EAs on my live account

So now we have a comparison of the EAs running freely, uncontrolled on demo, vs running them only when the market dynamics are suitable, that is, running them in a controlled manner on my live account.

The results are shown in the image below.

So my live account has made a profit in the month of February; my demo account went OK for a while but finished the month with a big loss. And that was because one of the EAs was in a longish drawdown on the demo account. And hence a copy of that same EA was not being allowed to trade on the live account. And hence I avoided a big loss for the month and finished in profit.

The philosophy behind this new trading method is simple.

Don't let your EAs try to cope with constantly changing market dynamics. That too often leads to failure. After all, EAs based on lagging technical indicators, cannot constantly adjust to new market dynamics.

So instead, let the market come to your EAs. In other words only let your EAs trade when the current market conditions are favourable for the strategy that has been coded into the EAs.

Trading with the EA Controller system does just that - all on autopilot.

And it can do so in one of two ways.

1. In Copier mode - By using a free Trade Copier from FX Blue to copy or stop copying trades to live accounts.

2. In Signal mode - By using FX Autotrader Elite EAs or the free EAs provided with the Controller

Online nwboater

  • Sr. Member
  • ****
  • Posts: 1382
Re: EA Controller
« Reply #2 on: March 04, 2017, 02:42:10 PM »
This is a fascinating concept!

The last two months have brought poor results for many strategies. During this period the controller might have been a big help.

It will be very interesting to watch Donna's demo's over a period of time. Does the vendor have any MyFxBooks that show this?

Also the Controller should work with signals that are controlled via an EA such as the SimpleTrader ones.

Cheers,
Rod

Offline donnaforex

  • Administrator
  • Hero Member
  • ******
  • Posts: 4155
Re: EA Controller
« Reply #3 on: March 04, 2017, 05:04:28 PM »
Agree, i'm really curious to see how this goes. My account (the one the 'best' trades go on) is live (the one i'm copying from that takes all trades is a demo), so will be cool to watch.
Follow DonnaForex on Facebook and Twitter.
 
If you need to contact me, email admin@donnaforex.com or Skype 'donnaforex' or PM via the forum.

Offline sydap2003

  • Jr. Member
  • **
  • Posts: 9
Re: EA Controller
« Reply #4 on: March 05, 2017, 05:06:39 AM »
This is a fascinating concept!

The last two months have brought poor results for many strategies. During this period the controller might have been a big help.

It will be very interesting to watch Donna's demo's over a period of time. Does the vendor have any MyFxBooks that show this?

Also the Controller should work with signals that are controlled via an EA such as the SimpleTrader ones.

Cheers,
Rod

Hi Rod, glad to hear that you find this a fascinating concept.

I have been advocating this method of trading for over a year on a manual basis. I couldn't generate any interest and I ended up not doing it myself either. Required too much effort and a solid dose of discipline - which I lack.

Luckily FXBlue.com was happy to work on my ideas and agreed to modify their Trade Copier for this purpose. I then developed the Controller to work with the modified Copier (v7.10) so that the whole process was automated. And then also used their quickchannel dll technology on my FX Autotrader so that the Controller can also control those Autotrader EAs directly on another platform without the need for a Trade Copier.

(Trade Copiers tend to be a bit hungry for CPU and memory resources, thus requiring VPSs with enough grunt if you want to run several EAs this way. Using the direct, Signal, mode, I am currently running 12 Autotrader EAs on a medium grunt VPS without any problem and will soon increase to 20.)

Note to other EA developers - making your EAs Signal ready for use with the Controller, requires only a few lines of code to be added to the source file - code that I would be happy to provide if you wanted to do some testing with the Controller; or perhaps as part of Donna's testing suite.

And yes you can view my live account as per link below. Just note that I only introduced the Controller mode of operation late in January on that account. A shame that I did not have it earlier since I would then have avoided most of that 50% DD that I suffered from July through to December!!

But, better late than never.

http://www.myfxbook.com/members/sydap2003/sydap-pro/1675402

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #5 on: March 06, 2017, 12:41:54 AM »
Just to be clear.

The EA Controller is NOT an EA. It is NOT a product that you can trade by itself.

It is a product to be used in conjunction with your EAs - to make them more profitable by eliminating most of the drawdowns that they all suffer from time to time.

To use this product with your EAs is very simple.

1. Open a demo account and transfer your EA from your live account to the new demo account. Keep the live account open.

2. Add the FX Blue Trade Copier Sender to your demo account and the FX Blue Trade Copier Receiver to your live account (The FX Blue Trade Copier is free and comes with a simple installer  https://www.fxblue.com/appstore/2/mt4-personal-trade-copier)

3. Add the EA Controller to the chart of the EA that you are running and enter the magic number of the EA and an ID#.

4. Add that same ID# to the Trade Copier Sender on your demo account and the Trade Copier Receiver on your live account.

5. On the EA Controller, select a value for the MA of the EA's closed profit (balance). Default is 10.

6. Repeat Steps 2 to 5 for each additional EA.

And you are set to go. That is all there is to it.

So with the settings we have made, when closed profit from the EA running on the demo account is above the moving average of the closed profit from the previous 10 trades (ie the EA is running profitably, market conditions are suitable for this EA), the next trade will be copied to your live account. When it falls below its MA (ie it is in drawdown), copying will be halted until it comes back up again (ie comes out of drawdown).

In other words, copying of trades from your EA running on the demo account to your live account, will only occur when the market conditions are suitable for your EA. And when your EA on the demo account goes into drawdown, the copying will be stopped and you will not suffer those losses on your live account.

Let the market come to you. Don't try to fight it and suffer losses.  Let the EA Controller manage this for you - all on auto.

 

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #6 on: March 07, 2017, 04:06:45 AM »
Does this sound familiar?

How many times have you purchased an EA, watch it make some money for a few weeks, and then promptly lose all of those winnings and more a month later. Yes, you are in a drawdown, and you wont know how long that drawdown will last. And how much money you will lose.

So you switch the EA off in disgust. You have spent $200 buying the EA, and now you have lost another $300 of your $1000 trading account.

A few months later you are tempted to try the EA again, but by now you have also seen another EA which appears to be much better. So you buy that one for $250 and trade it on the $700 that is left in your trading account.

But much the same thing happens. This time you go into a short drawdown but then the EA starts winning and you are making money. Happy days are here again. Unfortunately, six months later, the EA goes into a drawdown and you start losing all the profits you had made.

Grrrrrrrrrrrrrrrrrr!!!!!!

What do you do? Stop the EA? Keep it going and hope it turns around soon?

Sound familiar?      [Read on... http://www.diyforexskills.com/let-market-come/     

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #7 on: March 08, 2017, 06:18:13 AM »
Performance Update

I have been running several EAs made with the FX Autotrader Elite for a while and since 1 January, have managed these with the EA Controller in its Signal mode of operation.

So now we have some data for an EA running normally, UNcontrolled, on a demo account and an EA running while Controlled to avoid drawdowns, running on a live account as shown in image attached.

We have generated 2.4% return on our live account while our demo account essentially broke even.

In other words, trading with the EA Controller system has generated a 2.4% return from our EA which otherwise would have returned zero %.

A few points.

1. The Controlled account is a $1K live account; the UNcontrolled account is a $1M demo account. So the comparison should be in terms of % Return and % Wins and profit factors. These comparisons are shown in the attached image.

2. The graph at the bottom of the image is the performance graph of the EA on the demo account as measured by the EA Controller. The red line is the closed profit of the trades and the blue line is the MA of the closed profit.

It shows the the Controller's action in switching the live EA on and off as required and was based on a MA of 10 of closed profit. That is, the MA of closed profit was based on the previous 10 closed trades:
  • When the red line was above the blue line (performance on demo account is in uptrend), the EA on the live account was enabled to trade.
  • When the red line was below the blue line (performance on demo account is in drawdown), the EA on the live account was prevented from trading. As shown by the flat horizontal line on the "Controlled" graph. We were kept out of trading during the drawdown.

3. Now that we have generated some trading history, it would make sense to switch the MA to say 15 or 20. Just like individual trades need room to breathe (ie SLs should not be too tight), equity (or more correctly balance) curves also exhibit a normal ebb and flow and hence need room to breathe. We need the Controller to prevent us being ebbed into oblivion while allowing us to ride the flow!   :)








Offline ianj

  • Full Member
  • ***
  • Posts: 251
Re: EA Controller
« Reply #8 on: March 11, 2017, 01:19:22 AM »
I would expect that you might be able to see the theoretical effects of this mechanism on historical performance without actually needing to run it for ages to see what it might do

You would need something to parse a statement (say from myfxbook), reconstruct a balance curve only including those trades that qualify  - just as it would do going forward. Some adjustment to mimic the effects of copying would help

I don't know if the vendor has such a tool but it would help his case no doubt help convince all ..
« Last Edit: March 11, 2017, 01:23:45 AM by ianj »

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #9 on: March 11, 2017, 06:12:35 AM »
I would expect that you might be able to see the theoretical effects of this mechanism on historical performance without actually needing to run it for ages to see what it might do

You would need something to parse a statement (say from myfxbook), reconstruct a balance curve only including those trades that qualify  - just as it would do going forward. Some adjustment to mimic the effects of copying would help

I don't know if the vendor has such a tool but it would help his case no doubt help convince all ..

Thanks for the suggestion. I don't have such a tool but like all things it is no doubt something I could learn how to do.

However the tricky thing would be how to decide which trades should qualify - at a first stab you would eliminate x% of losing trades when they occur in a sequence of say 5 or more (ie sufficient number of losing trades in a row to qualify as a DD rather than a normal ebb and flow of wins and losses)?

Interestingly this is also an issue with the EA Controller, especially at the start. Here the issue in question is how many trades we should use to determine the MA of closed profit/balance. And hence the trigger point for copying or not copying.

On the EA Controller we can set the MA to any number of closed trades to be measured in the calculation of the MA. I normally start with 5, which can lead to whipsawing of YES and NO copying and potentially make things worse. So I like to move the MA to 20 asap so as to allow the balance curve to breathe.

The image below shows how copying of trades would vary if we had an SMA of 20 instead of an SMA of 5. Note the two red rectangles on the SMA 5 chart where we are whipsawed in and out of copying. In that example, the copied trade results would be worse than letting all trades be copied.
(The flat lines on the graphs are due to the fact that if, for example, you want to measure a MA of 5 previous values, you need at least 5 values to start.)

As to the suggestion we started with, you can of course do a crude analysis of how the EA may have performed if it had been "controlled".
Simply look at your gross profit and gross loss over a period, and then recalculate net profit by reducing gross loss by say 70%.

Eg, UNcontrolled.  gross profit = 1000; gross loss = 400; net profit = 600
      Controlled.      gross profit = 1000; gross loss = 70% of 400 = 280; net profit = 720

And hence you would be 120 better off if indeed the EA Controller had been able to restrict your drawdowns by 70%. This is very crude since it depends on whether there were just a few largish DDs, or whether there many small DDs. In the latter case you could be better or worse off depending on what value you had used for the MA of closed profit on the EA Controller.

Offline ianj

  • Full Member
  • ***
  • Posts: 251
Re: EA Controller
« Reply #10 on: March 11, 2017, 06:50:46 AM »
Yes - i can see it's a tuning issue. You could also "soften" the threshold by manipulating risk continuously (ie not just on/off - vary the lot size) according to how far it is from the MA (or regression or whatever you use)

You might also wish to consider Z score  - it may be possible to change the size of or omit copied trades in a run of failures or successes if the historical analysis indicates a Z score in excess of +/- 2.0. (I have a strategy with a Z score of -3.8 and it DEFINITELY improves the overall payoff/PF if i increase/reduce the size of following trades accordingly)

« Last Edit: March 11, 2017, 06:56:08 AM by ianj »

Offline ianj

  • Full Member
  • ***
  • Posts: 251
Re: EA Controller
« Reply #11 on: March 11, 2017, 06:54:37 AM »
Take a look at this
https://www.mql5.com/en/articles/1492

BTW The z-score on myfxbook is calculated this way
« Last Edit: March 11, 2017, 06:58:45 AM by ianj »

Offline ianj

  • Full Member
  • ***
  • Posts: 251
Re: EA Controller
« Reply #12 on: March 11, 2017, 07:09:58 AM »
I have some (pseudo code) i use to calc a z score in my own code - its in java but its still useful i think:

int zn = 0, zr = 0, zw = 0, zl = 0;
boolean lastInProfit = false;
while (next trade) {
    boolean inProfit = trade profit>=0;
    if (zn==0 || lastInProfit!=inProfit)
        zr++;
    zn++;
    lastInProfit = inProfit;
    if (inProfit) zw++; else zl++;
}
double zp = 2*zw*zl;
zscore = (zn*(zr-0.5)-zp)/Math.sqrt(((zp*(zp-zn))/(zn-1)));


Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 519
Re: EA Controller
« Reply #13 on: March 11, 2017, 08:09:20 AM »
Yes - i can see it's a tuning issue. You could also "soften" the threshold by manipulating risk continuously (ie not just on/off - vary the lot size) according to how far it is from the MA (or regression or whatever you use)

You might also wish to consider Z score  - it may be possible to change the size of or omit copied trades in a run of failures or successes if the historical analysis indicates a Z score in excess of +/- 2.0. (I have a strategy with a Z score of -3.8 and it DEFINITELY improves the overall payoff/PF if i increase/reduce the size of following trades accordingly)

Thanks. That opens up a whole new world of sophistication and possible improvement to what I am trying to achieve.

The EA controller works in one of two modes. Copier mode using the FX Blue Trade Copier; and Signal mode where we have copies of the same EA on both a demo and live account, and where the live EA is switched on and off.

To progress your idea of lot size variation rather than simple yes or no, in the Copier mode, I would need to collaborate further with the FX Blue guys. In current version you can manually vary lot size to be copied. So presumably a z-code algorithm could be incorporated?

In the Signal mode I could imagine the EA Controller's performance analysis to be "z-coded" to soften the signal?, and then have the recipient EA, the one on the live account, also being z-coded to vary the lot size depending on its own history of trading.

Since the EA's history on the live account will always be different to its history on the demo account and since the live EA would have mainly positive history, it would be mainly increasing lot sizes. And hence increasing profits.

Hhmmm.... this all sounds fascinating. Just need a few more hours in the day.  :-)  Will have to talk with my coder.

Offline ianj

  • Full Member
  • ***
  • Posts: 251
Re: EA Controller
« Reply #14 on: March 11, 2017, 11:01:21 AM »
In the Signal mode I could imagine the EA Controller's performance analysis to be "z-coded" to soften the signal?, and then have the recipient EA, the one on the live account, also being z-coded to vary the lot size depending on its own history of trading.

You would CALCULATE the z-score on the demo account and use it to vary the trading lot multiplier into the LIVE account

Analysing the z-score on the LIVE account is unnecessary as it's trading will already have been manipulated

As i hinted, a high magnitude z-score is an exception rather than the norm - it simply indicates that the strategy tends to strings of losses or wins that are outside of the normal distribution. In particular it suggests that having a z-score below -2 means that you drop (or reduce) the trade after a failed trade and increase the size (or at least include) after a success.

The z-score assumes some symmetry - you might consider something more sophisticated that treats an extended string of failures different to a string of successes. You might also consider a string of successes/failures that exceed a threshold as indicating a rolloff in the lot multiplication (If you continue to scale up the successful trades you will inevitable hit a BIG failure) Perhaps a tool that offers a banded distribution saying how often you get runs of 2/3/4/5 successes might be interesting (One for the future)

You are entering territory here normally requiring more comprehensive quantitative analysis

One side effect is that it can increase dd/profit in SOME strategies depending on how lumpy the equity curve - your mileage will vary

Another idea for you:
  • Retrofitting RECOVERY in a copy - this is really just a variation on z-score size mapping. This is NOT to be confused with a martingale which tends to keep the preceding trade open, compounding risk geometrically - they are distinct trades taking advantage of a mathematical distribution
  • Retrofitting Position sizing to add flexible compounding
  • What about currency or pair risk reduction - reducing the size of trades where other trades are already open in that pair or would increase the risk on a given currency more than is comfortable (anyone who has run multi pair Asian scalpers will know what i mean only too well)
« Last Edit: March 11, 2017, 11:04:13 AM by ianj »

 

browse forum

* Recent Posts

Re: FX Autotrader Elite by diyforexskills
[Today at 03:15:32 AM]


Re: Easy Forex (ezefx.com) by IFFTrader
[Today at 02:33:37 AM]


Re: Fidelitas by reinerh
[Yesterday at 11:04:05 PM]


Re: Fidelitas by petermatt
[Yesterday at 10:54:48 PM]


Re: Easy Forex (ezefx.com) by diyforexskills
[Yesterday at 10:03:35 PM]


Re: Easy Forex (ezefx.com) by CanadianPsycho
[Yesterday at 09:52:56 PM]


Re: ECB ending deposit protection? by CanadianPsycho
[Yesterday at 09:48:50 PM]


Re: ECB ending deposit protection? by compujock
[Yesterday at 09:24:23 PM]


Re: Forex Gold Investor EA by wallstreet.forex.robot
[Yesterday at 09:24:22 PM]


Re: Best Scalper by donbon2
[Yesterday at 08:57:55 PM]