collapse

Author Topic: EA Controller  (Read 139437 times)

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #270 on: August 04, 2018, 04:25:04 AM »
Analysing Test 4 Results for VFX LBO GBPJPY
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 112235 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager)
2. Magic 6112235 is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 9112235 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

The EA has a TP of 66 and a SL of 33 pips, a reward:risk ratio of 2:1. Therefore the reverse trade needs to have its TP at 33 - 5 = 28. If we kept the SL at 66 we would have a bad reward to risk ratio; hence I have set SL for these reverse scale-ins to 14 so as to keep the same 2:1 ratio.

The results are summarised in the image below.

The top graph shows the variable profitability of the EA and hence indicates when the normal and reverse scale-ins were activated. This is also evident in the middle graph.

At first glance this experiment is a failure - a loss of about $120 over 4 months (trading all three at 0.3% risk or 0.1 lot). But still better than trading the EA alone losing $243.

But what if we run the EA at 0.01 lot (0.03% risk) and set the scale-ins to trade at 1% risk (about 0.3 lots on a $10K account).

In that case the EA would have lost only $24, the normal scale-in, 6112235, would have made $108 and the reverse scale-in, 9112235, would have made $250. Net $334 profit.   

So even though we had a losing EA,  :(  ,we (would) have made a profit overall with a better choice of lot sizing.   :) :)

Note also that the reverse scale-in trades had a win ratio of only 39%, probably due to the very tight SL of 14 pips. So one other parameter worth testing is the reverse scale-in SL - maybe set that to 28 for a 1:1 reward:risk ratio?

(In Test 6 we set the reverse scale-in to be triggered on SMA 8 of the closed P/L curve - this turned out to be too tight and results in 33 trades with loss of $56)

Addendum

I am running these Tests on the one platform. However the Controller can communicate with the TM across two platforms on the same computer or VPS by making use of the FX Blue QuickChannel technology. So we could trade the EA on a demo account platform and run the scale-ins on our live account platform. That way we don't suffer any losses that the EA might make.

If we examine the data in this way, assuming that the EA was run on a demo and the scale-ins on a live account, we would have the performance as shown in the image.

So the EA, VFX LBO GBPJPY, is in DD which is of no concern since it is on demo.

The live account, where the scale-ins were traded, is in profit. Remember that this was trading the scale-ins at 0.3% risk, so if we upped that to 2% risk we would have made 2% pm over the 4 months with a max DD of around 6%. Not a bad result for trading an EA that is unprofitable!!

The free VFX LBO EA (https://www.mql5.com/en/code/17465) comes with set files for 10 different currency pairs. Since they all operate the same way, one assumes they could all be traded in this manner. With the same sort of results??? I am going to give it a whirl - will take two months or so to generate the first 20 trades for each so testing with scale-ins won't begin till October.
(A colleague of mine has back tested each of these and found that only the EURJPY was profitable over 10 years when traded normally. So this should be a good test of the scale-in system).

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #271 on: August 05, 2018, 06:10:20 AM »
Analysing Test 4 Results for PZ theZone EA AUDUSD
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 98765 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 698765 is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 998765 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

This EA is set with a TP 0f 100 and SL of 50 (2:1 reward:risk ratio) with BE at 15 and a 50% TS. The impact of the BE and TS meant that the average TP was only around 5 to 10 pips and so the actual reward:risk ratio was more like 1:5 since SL was always 50 pips!! So even with an 83% win rate, the EA still loses. That is the problem with scalpers.
For some reason I set the reverse scale-in to TP 50 and SL 50 (1:1 ratio) rather than SL to 25 for the 2:1 ratio. However only 2 out of 26 trades were badly affected by that.
Scale-in trades were set to be closed whenever the EA's trade was closed but an inspection of the data shows that this did not occur for 2 of the 23 reverse scale-in trades, badly increasing the loss.

So the picture is not pretty as shown in the image. The scale-in system delivers a worse result than the EA itself.

This is surprising since an inspection of the SMA 20 on the performance curve, top image, shows that there were clear periods of downtrend and one uptrend; ie not a choppy situation. Hence the scale-ins should have been profitable as indicated by the arrows on the top image. (During a period of downtrend the reverse scale-ins should be making a profit.) I also don't understand why 10 of the reverse scale-ins closed at a profit of around 15 instead of the 50 as set. The TP of a scale-in corresponds to the SL of the EA and the EA has no moderating feature for trades going into a loss. Hence a reverse scale-in trade in this setup should finish either at 50 pip profit or a small loss while the exit is controlled by the exit of the EA's trade. Something is going wrong. Slippage of the signal from the Controller??

So I am going to make some adjustments to the scale-ins. Tighten the TP and SL of normal scale-in to 25/25 and of reverse scale-in to 50/15. And then a second reverse one with same TP/SL but also a TS of 25.

NOTE
In these Tests I am using a reverse scale-in for when the EA is in DD. However we could also of course use a normal scale-in but with much tighter TP and/or SL during such DD periods. Something else we can test. We just need something different for when the EA is in DD since the market during these periods in obviously unsuitable for the EA's optimised strategy. Some form of variation is required to eke out a profit during these periods.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #272 on: August 07, 2018, 11:18:25 AM »
Analysing Test 4 Results for FXAE 48164 EURUSD
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 48164 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 648164 is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 948164 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

This EA was made and optimised with the FX Autotrader Elite some years ago. It has hard TP and SL at 60/60 with BE at 30 locking in 5, and TS of 30 starting at 30. It also has a dynamic exit based on a MA crossover. It does not trade on Wednesdays. Exits for the scale-ins are triggered by the EA's trade closure.

As is evident from the image, the EA performed well during the Test period from 16 April with closed P/L remaining above its SMA 20 so that only normal scale-ins were entered. Since these scale-ins are entered at +5 pips, the profit from scale-in trades is always 5 pips less than that from the EA itself. It appears that 9 normal scale-ins were not triggered during the test period (32 vs 41). Nevertheless, adding scale-ins during profitable market conditions has proven beneficial. All trades are at approximately 0.6% risk. Trading at 2% risk would have yielded almost 6% pm with a 6% DD.

Note:
Since this EA is an FX Autotrader Elite made EA, the Controller can control copies of the EA directly; and the EA can also be set to reverse trades. In other words we don't need to use scale-ins to trade this system. This is being done in Test 2 and has been running one month longer. During that first month the EA was in DD and hence reverse trades (948164) were executed. The data for that Test is shown in the second image, showing that the reversing during DD also led to a small positive result. In this Test 2, the monitoring EA (48164) is run at 0.01 lots and the normal (648164) and reverse (948164) trades at 0.1 lot or approximately 0.6% risk.

All in all, very encouraging results with this EA.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #273 on: August 09, 2018, 08:30:20 AM »
Analysing Test 4 Results for PZ theZone EA AUDUSD
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 98765 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 698765 is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 998765 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

This EA is set with a TP 0f 100 and SL of 50 (2:1 reward:risk ratio) with BE at 15 and a 50% TS. The impact of the BE and TS meant that the average TP was only around 5 to 10 pips and so the actual reward:risk ratio was more like 1:5 since SL was always 50 pips!! So even with an 83% win rate, the EA still loses. That is the problem with scalpers.
For some reason I set the reverse scale-in to TP 50 and SL 50 (1:1 ratio) rather than SL to 25 for the 2:1 ratio. However only 2 out of 26 trades were badly affected by that.
Scale-in trades were set to be closed whenever the EA's trade was closed but an inspection of the data shows that this did not occur for 2 of the 23 reverse scale-in trades, badly increasing the loss.

So the picture is not pretty as shown in the image. The scale-in system delivers a worse result than the EA itself.

This is surprising since an inspection of the SMA 20 on the performance curve, top image, shows that there were clear periods of downtrend and one uptrend; ie not a choppy situation. Hence the scale-ins should have been profitable as indicated by the arrows on the top image. (During a period of downtrend the reverse scale-ins should be making a profit.) I also don't understand why 10 of the reverse scale-ins closed at a profit of around 15 instead of the 50 as set. The TP of a scale-in corresponds to the SL of the EA and the EA has no moderating feature for trades going into a loss. Hence a reverse scale-in trade in this setup should finish either at 50 pip profit or a small loss while the exit is controlled by the exit of the EA's trade. Something is going wrong. Slippage of the signal from the Controller??

So I am going to make some adjustments to the scale-ins. Tighten the TP and SL of normal scale-in to 25/25 and of reverse scale-in to 50/15. And then a second reverse one with same TP/SL but also a TS of 25.

NOTE
In these Tests I am using a reverse scale-in for when the EA is in DD. However we could also of course use a normal scale-in but with much tighter TP and/or SL during such DD periods. Something else we can test. We just need something different for when the EA is in DD since the market during these periods in obviously unsuitable for the EA's optimised strategy. Some form of variation is required to eke out a profit during these periods.

This is more like it!

I had noted that the reverse scale-ins were not doing what they should. So I made some adjustments including adding a TS of 25 to the reverse scale-ins. Instant success (or just good luck!). However it was nice to see that for one of the trades, the reverse scale-in was stopped out at about 15 pips profit and the EA's trade continued to finish with a small profit as well on its return journey. So both the normal and the reverse trade finished in profit!

The image shows this better action reflected in the graph. This is how the system is meant to work. When the EA is in DD and the EA is losing, the reverse scale-ins should be winning so as to negate the loss.
« Last Edit: August 09, 2018, 10:46:23 AM by diyforexskills »

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #274 on: August 10, 2018, 08:20:47 AM »
Analysing Test 4 Results for SQ EURNZD M15
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 112233 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 6112233 is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 9112233 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

This EA has a TP of 41 and SL of 89 for a 1:2 reward to risk ratio meaning that we need a 70% win rate to break even. I had set the normal and reverse scale-ins to TP of 25 and SL of 50 and set close on EA trade exit to true. Meaning that the scale-ins were also closed whenever the EA's trade was closed. In order to trigger enough trades I set Max Spread to 5 rather than the default of 2.

As shown in the top graph, the EA is still in profit starting from 18 Feb 2018. The performance curve is more choppy relative to its SMA 20 than I would like. So might make some adjustment to that after we have had a few more months of trading. For the period when the scale-ins were operating, from 26 May, the EA is at a small loss. The normal scale-in finished the period with a small gain while the reverse scale-in is at a loss.

Given these results and an examination of the individual trades, I have reset the normal scale-in to TP/SL of 40/25 with a TS of 20; and the reverse scale-in to 40/20 also with a TS of 20. (I always prefer to trade with a 2:1 reward to risk ratio.)

So we need a bit more time to test this EA to see if trading it with the Controller system can improve performance.
« Last Edit: August 10, 2018, 08:49:16 AM by diyforexskills »

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #275 on: August 11, 2018, 05:54:45 AM »
Analysing Test 4 Results for Samurai Robot EURUSD H1
(Pls refer to Reply #235 for descriptions of these Tests)

Test 4
1. Magic 102057 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 6102057is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 9102057 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

This EA has TP of 30 and SL of 88 pips with a TS of 23 and a Max Order of 2. So it needs at lest a 75% win rate to just break even. It is achieving 80% since 18 Feb 2018 and so is running profitably with a profit factor of 1.36. Not bad for a free EA run on default.

The stats show that the normal and reverse scale-ins have not been doing what we would like. The normal scale-in, at 5 pips, had a somewhat tighter SL than the EA so I have now increased that to match the EA's exit price on SL. This means setting the SL at -93 (88 + 5). TP is at 25 (25 + 5 =30 so as to match the EA's TP).

I made a mistake of setting a TS of 25 to the reverse scale-ins. It seems that because of that the reverse scale-in trades never reached their max TP of 83 pips (88 - 5); whereas 11 of the 73 EA's trades reached their 88 pips SL. Hence the reverse scale-ins performed very badly - 16 losses at 25 pips and only 3 wins for 18 pips). Since reverse scale-ins start at -5 pips, I should also have set the SL at 35 pips to coincide with the TP level for the EA's trades. Changes have now been made.

Let's see how we go with these revised settings.
« Last Edit: August 11, 2018, 05:58:56 AM by diyforexskills »

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #276 on: August 13, 2018, 07:19:36 AM »
And finally the last of the six .... Test 4 for BFS GBPCAD M1

Test 4
1. Magic 656565 will trade all the time at 0.1 lot (five Controllers are placed on this chart each linked to its own Trade Manager; two of them were used for this Test)
2. Magic 6656565is a Normal scale-in at 0.1 lot and will trade when the EA is in uptrend (Closed P/L above its SMA 20), at +5 pips
3. Magic 8656565 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 7), at -5 pips
4. Magic 9656565 is a Reverse scale-in at 0.2 lot and will trade when the EA is in downtrend (Closed P/L below its SMA 20), at -5 pips

It has become clear to me that this system of adding normal and reverse scale-ins is not suited to scalpers.

That is sort of logical when you consider the typical performance curve of a scalper - many small wins followed by a big loss, followed, generally, by more small wins. Hence the Normal scale-in will also suffer that big loss and when the Reverse scale-in is kicked in following the big loss (closed P/L falls below its SMA), the EA starts winning again and hence the reverse scale-in loses.

This is evident from the table at bottom of image. The reverse scale-in at SMA20 is magic 9656565 and made a large loss. Actually not as bad as it looks since $220 of that loss is due to a failed SL from a MT4 glitch. Magic 8656565 is a reverse scale-in at SMA 7 which gives a marginally better result, ie less negative!

The two top graphs show different SMAs (20 and 10) applied to the same performance graph.

So my thoughts on what we might do for scalpers of this type is as follows.

1. Use a tighter SMA - eg 10
2. When closed P/L is above SMA10, scale-in a normal trade but with SL set much tighter than on the EA; say 50% reduction.
3. When closed P/L is below SMA 10, scale-in a normal trade but with twice the lot size. Rationale here is that after a loss scalpers normally go back to making small wins and so we want to capitalise on that by increasing lot size.

I will set this up next weekend.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #277 on: August 15, 2018, 08:58:02 AM »
Comparing the Different Methodologies

The EA Controller was designed to enable a trader to increase lot size of an EA's trade while the market is favourable and hence the EA's performance is positive; and decrease lot size and/or enter reverse entry trades when the market is unfavourable and performance is in downtrend.

In order to achieve this the EA Controller can be made to operate directly on the FX Autotrader Elite EAs or on the DIY Trade Manager Plus. (It can also control the FX Blue Trade Copier, which in turn can vary trade parameters and/or invert trades, but that is another story.)

The TM in turn can operate in one of two ways.
a. It can be used to close unwanted trades from the various copies of the EA, by setting a very tight SL.
b. It can be set to trigger scale-ins of the type we need.

I have been testing these three methods for one of my FXAE strategies, FXAE V48164 running on EURUSD M15. The results for trading from 11 March 2018 till present are shown below.

So the EA running on 3 different MT4s all on the same VPS, yield almost identical results for the EA itself (magic 48164).

In the top image (Direct Control) the Normal copy of the EA (648164) was triggered 46 times for a healthy profit; and the Reverse entry copy of the EA (948164) was triggered 13 times for a small loss.

In the middle image, both the Normal and Reverse copies of the EA were running all the time with the TM being used to close "unwanted" trades. ie Normal trade were closed early (at 5 pips) during DD; and Reverse trades were closed early during uptrend in performance. I have no idea how the extra 7 (948164) trades were triggered. There should only have been 59 trades, not 66.

In the bottom image, Normal scale-ins were set to be triggered at 5 pips in profit and Reverse scale-ins were set to be triggered at -5 pips in loss depending on whether the EA's performance was in uptrend or DD. Hence we see fewer Normal trades (648164)  and zero Reverse trades (948164) when compared to the other two tests.

Note also that in the top two test results the EA itself (magic 48164) was run at 0.01 lot compared to 0.1 lot in the bottom test.

In summary, with this EA to date, adding reverse trades during a DD (closed P/L below its SMA 20), is not helpful. But otherwise adding Normal trades is beneficial and any of the three methods can be used.

My preference is for the first method, direct control, but that can only be done with FXAE EAs which have the Signal code embedded in their code. For third party EAs, my preference is for the scale-in method since we get a higher win rate for the Normal trades that way - 71% vs 56%. (I suspect that that is due to a scale-in being triggered at only 5 pips in profit of the EA, and so if price goes against the trade almost straightaway on opening, the scale-in will not be triggered and hence a loss will be avoided. But of course the TPs will be 5 pips less if we use the same TP levels. So the benefit could go either way.)


http://diytrademanagerplus.com/
http://exceptionalfx.com/ea-controller/
http://www.fxautotraderelite.com/

You can purchase the licence for 2018 from here (please see Reply #154 for a description of this Package)
https://secure.avangate.com/order/checkout.php?PRODS=4721645&QTY=1&CART=1&CARD=2

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #278 on: August 17, 2018, 08:41:46 AM »
Analysing Test 1 and 2 Results for FXAE V482220 USDCAD H1

These tests involve running the Normal entry version of the EA when the market conditions are favourable (ie performance is in uptrend); and running the reverse entry version when the EA is in downtrend. And then the two ways of doing that:
1. Controlling the EAs directly (EA on or off depending whether in uptrend or downtrend)
2. Controlling the EAs via the use of the Trade Manager (using the TM to close unwanted trades at -5 pips in loss depending on uptrend or downtrend)


This EA has a TP of 120 with BE locking in 15 at 60 and a TS of 30 starting at 60. The reverse trades have TP and SL both at 60. The results are shown below for the two systems. For some unknown reason we had an extra 7 trades on account using the TM so the results for the EA on that account are skewed positively.

Basically the EA's performance by itself is marginally negative over this period and has been in DD (closed P/L below SMA 20) for most of the time as shown in second image.

So it is pleasing the the reverse entry trades (magic 9482220) have generated a profit.

The normal entry trades (magic 6482220), only 5 during that period, however have generated a loss. The reason for that is obvious to see from the second image - the EA was in uptrend for only two brief periods and so the 5 trades taken during those brief periods all ended in a loss.

This again highlights how a choppy performance curve causes problems with this method of trading. We need performance curves with longer periods of ups and downs. In this example we are in a drawn-out downtrend (DD), but still making money with our reverse entries - that is what we want. So until we see a sustained return to favourable market conditions for this EA, it would make sense to switch off the Normal entry trades.
Or perhaps run the normal entries with tighter TP and SL.

Food for thought.




Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #279 on: August 20, 2018, 02:44:37 AM »
Let's also try something else..

In one of the other threads (Volatilty Factor) mauricejack wrote "in my experience this is one of the classic problems of using an EA. Deactivate it as soon as it enters a phase of dd (which is physiological), and reactivate it as soon as it has recovered, risking other losses. This EA in particular can be very frustrating (more or less long periods of gain and few trades with large losses), but in perspective if left to work it recovers rather quickly (based on tests that are one of the few objective data available). And it is precisely the perspective (if we overcome this psychological block of the phases of DD), it makes me notice that in live it has obtained 100% gain in 4 months ... Chapeau!
...)."

This observation led me to wonder how successful it would be to help an EA get out of DD sooner by adding to the lot size when the EA goes into DD. I am trialing an EA for over a year which does just this. After each loss, the lot size is increased by 50%; and then back to starting size after a win. This has resulted in a performance curve for this EA as shown in first image. The orange shaded areas show the big trade that goes to full TP and recoups all the losses; the red shaded areas show how sometimes the recovery from a large trade is small because the EA has a TS. Still, not bad.

So I am thinking that we can do something like this with the Controller and TM.

Run an EA with 5 copies of the Controller linked to five copies of the TM set to scale-in trades.
Set the MA of the Controllers to 3, 6, 12, 24, 48 EMA and the TM set to scale-in a normal trade at 5 pips in profit for the EA whenever the closed P/L of the EA is below its EMA. ie is in DD.
Eg. When Controller EMA 3 signals DD, the related TM will scale-in an extra normal trade thus increasing overall lot size. If DD continues, Controller EMA6 will also be in DD and signal its related TM to scale-in another extra normal trade... and so on. As shown on the second image showing Controllers with EMA 3, 6, 12, 24, 48.

When the EA starts to meet favourable market conditions again and moves out of DD, ie into uptrend, the scale-ins will be successively removed as closed P/L moves back  above EMA 48, 24, 12 etc.

In this example the Controllers would be set to measure the performance of the EA only. We could run a second test of this setup, but make the Controller measure the performance of the EA and the scale-ins. After all, we can use up to 10 magics in the Controller. So the Controller would be signalling uptrend or downtrend based on the cumulative performance of the EA and scale-ins rather than on the performance of the EA alone.
« Last Edit: August 20, 2018, 02:46:13 AM by diyforexskills »

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #280 on: August 21, 2018, 06:07:28 AM »
A new Multiple Controller Test
Pls see previous Post for the genesis of this idea.

Thinking further about this idea of using the EA Controller to respond to changing market conditions, I am now ready to set up the following Test.

1. I will use the same EA whose results I showed in the first image of the previous Post. This is FX11 2401 running on USDJPY H1 (free bonus EA for purchasers of the EA Controller). The settings used include a lot multiplier of 1.6 to increase trade size after a loss. For my Test with the Controller I will set this parameter to 1 since I want to use the TM to effectively increase lot size after a loss rather than the EA itself. (The EA is a simple OB/OS strategy using CCI with a 1:1 RR and a TS.)

2. The EA will trade with a 0.03 lot size and the TM normal scale-ins will be set to add 0.02 at EMA 3 DD, a further 0.03 at EMA 6 DD, a further 0.05 at EMA 12 DD, a further 0.08 at EMA 24 DD - so total now is at 0.21 lot (this mimics the 1.6 lot multiplier setting of the EA when used alone; and then...

3. If the DD were to continue then this marty style of increasing lot size during DD could end us up in big trouble! Therefore I have decided to enter a counter or reverse scale-in of 0.21 lot at EMA 48 DD. So if the DD goes beyond/deeper than EMA 48, we will be fully hedged for subsequent trades. At this stage I will need to be ready to intervene manually. If and when the EA starts to pull out of DD, say back past EMA 3,  I will need to deactivate the reverse scale-in manually. Because the remaining normal scale-ins will successively stop as the uptrend in performance continues past EMA 3, 6, 12, 24 and it would be some time before the 0.21 lot reverse scale-ins would stop at closed P/L going above EMA 48. Alternatively, at this point, I could change the EMA on the Controller for the reverse scale-in TM from 48 to 3. In either case I will have to reactivate/reset this Controller's EMA at some point. So not totally automated. A bit of "tinkering" may be required.  :)

4. I will set up a copy of this test but with the Controllers measuring the sum of EA plus scale-in results, rather than just the performance of the EA itself. (I have a feeling that this will yield better results, but we will see.) From previous optimisation back tests I know that this EA will lose badly without some form of lot size multiplier. Note also that this EA, when used with lot size multiplier activated, has a built-in protection so that any one trade will not be at a higher risk than 20% of account balance.

5. By using the FX11 2401 EA for these Tests, we will also be able to compare the results of this type of intervention using either lot size multipliers built into the EA vs using the Controller and TM to achieve a similar type of intervention. The final reverse scale-in for my setup at EMA 48 is meant to serve as an equivalent of the 20% risk limit of the EA (refer Point 4 above).

This should be interesting. I will have the EA on one chart with 10 copies of the Controller, and then a further 10 charts each with a copy of the Trade Manager. Most of the time they won't all be trading so should not be too much of a strain on VPS resources.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #281 on: August 22, 2018, 07:33:18 AM »
In my previous Post I wrote:
At this stage I will need to be ready to intervene manually. If and when the EA starts to pull out of DD, say back past EMA 3,  I will need to deactivate the reverse scale-in manually. Because the remaining normal scale-ins will successively stop as the uptrend in performance continues past EMA 3, 6, 12, 24 and it would be some time before the 0.21 lot reverse scale-ins would stop at closed P/L going above EMA 48. Alternatively, at this point, I could change the EMA on the Controller for the reverse scale-in TM from 48 to 3. In either case I will have to reactivate/reset this Controller's EMA at some point. So not totally automated. A bit of "tinkering" may be required.  :)

Of course we could also do something a little out of the ordinary. We could add a copy of the Controller to the reverse scale-in TM and set it to monitor the performance of those reverse scale-ins. And then link it to another copy of the TM which we set to close any unwanted trades at -5 pips in loss. So when the market dynamics become suitable again for the EA and hence the reverse scale-ins start losing, say when closed P/L of these reverse scale-ins falls below its EMA 6, the extra TM will close any such unwanted trades at -5 pips.

If the market continues to be favorable for the EA and hence the performance uptrend of the EA continues, then reverse scale-ins will no longer be activated when closed P/L rises above EMA 48. And hence there will be no unwanted trades to close any more and so this extra TM will just stay dormant until the next major DD and recovery.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #282 on: August 26, 2018, 04:28:37 AM »
Steady performance

Pleasing to see the FXAE V48164 continuing to perform well; and that doubling lot size during favourable market conditions (by using the Controller and TM to add magic 648164) is paying dividends; delivering around a 40% bonus in profit while market conditions are favourable and the EA is winning. All done on auto - no need to tinker manually.

Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #283 on: August 27, 2018, 04:14:26 AM »
Control Trade Exits with the Trade Manager

In trading most attention is paid to getting the entry points correct. Equally important is getting the exits correct. Mostly we just do this by setting TP and SL levels and maybe BEs and TS, usually all pip-based. And hope for the best.

In my testing of the Controller so far we have focused on using the Controller with the TM to add normal or reverse scale-ins depending on market conditions.

But we can also use the Controller with the TM to adjust/override the exits of our EA's trades depending on whether the market dynamics are favourable (EA is winning), so probably best to leave as is, or unfavourable (EA is losing), time to intervene.

The DIY Trade Manager Plus provides many ways of setting TP and SL and TSs - from ATR-based settings to using RSI levels, and TS based on fractals, or PSAR or MA crossovers, or %.

We can use the Controller to signal when the TM should be activated - eg when the EA goes into DD, say closed P/L below its SMA20. At this stage we might just do something to tighten the exits. If closed P/L falls further, say below its SMA 40, we might use a second copy of the Controller to activate a second copy of the TM to just close trades at -5 pips.

I am currently trying a variety of settings with the BollingerBand EA. Just by itself, without use of the Controller, as described in this thread https://donnaforex.com/index.php?topic=20211.msg379730#new (Replies 98 to 101).


http://diytrademanagerplus.com/


Online diyforexskills

  • Verified Vendor
  • Sr. Member
  • ***
  • Posts: 848
Re: EA Controller
« Reply #284 on: August 30, 2018, 08:35:38 AM »
Control EA trades based on market indicator?

The EA Controller we have now is based on measuring performance of the EA in order to determine the signal - yes or no (eg switch TM on or off for adding a positive or negative scale-in).

But a simple coding change could have the Controller signalling on the basis of a market indicator.

For instance, if we have one setting of an EA that generally works well when say the Daily ATR 14 is higher (say above 100 or whatever), and another version which works better when the Daily ATR14 is lower (say below 100), then if we had an ATR-based Controller we could control the TM to close trades at say -5 pips on whichever EA version is less suitable under the prevailing ATR condition.

Suggestions anyone as to which indicator may be most appropriate for such an approach?

 

browse forum

* Recent Posts

Re: LEGEND by anjelo
[Today at 06:53:59 AM]


Re: LEGEND by reinerh
[Today at 04:52:18 AM]


Re: LEGEND by Humble Trader's Fx
[Today at 02:42:31 AM]


Re: Darwins by outsidetheboxhk
[Today at 01:25:24 AM]


Re: EA Controller by diyforexskills
[Today at 12:55:44 AM]


Re: Apogeum Price Action (via MQL5.com) by ivanvp
[Yesterday at 10:55:19 PM]


Re: Forex Manager DBA by ForexManagerDBA
[Yesterday at 10:22:11 PM]


Re: Forex Manager DBA by ForexManagerDBA
[Yesterday at 10:09:27 PM]


Re: Darwins by petersurrey
[Yesterday at 09:06:58 PM]


Re: Will cryptocurrencies like Bitcoin ever gain mass adoption? by Humble Trader's Fx
[Yesterday at 05:44:51 PM]