(binned per thread starter’s request) MM (Money Maker) Detective | Page 276 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Money management: Capital needed per 0.07 USD per microlot 0 replies

The Volume Detective 411 replies

Attachments: (binned per thread starter’s request) MM (Money Maker) Detective
Exit Attachments

(binned per thread starter’s request) MM (Money Maker) Detective

  • #5,501
  • Aug 3, 2019 4:51pm Aug 3, 2019 4:51pm
Quoting jdva
R4M-BD-ADR EA v19214 Hello R4M, there must be something wrong with the exit conditions either on the Strategy Tester or the EA. I did a BT on GBPUSD with CC entry and exit Midline cross, SL ADR fraction=3 and got these results as an excerpt: {image} As you may see for example on O.Nr. 30 sell opened at 1.30202... SL was 1.32372 but got closed by Midline Cross (?) at 1.31334 for a loss of 100.41. I do remember on a earlier post you explained this: "2. Hard-coded MAX Stop We use a StopLoss. The StopLoss is computed as a fraction of the ADR and registered...
Hi @jdva.
All your open orders 30 - 33 were closed at the same time. The price closed 110 pips above your first trade. Which means the center midline band, shifted above all your sell trades, and hit the midline to exit all trades.
I could be wrong, but it looks that way.
  • #5,502
  • Aug 3, 2019 5:51pm Aug 3, 2019 5:51pm
Quoting Robot Trader
{quote} The AllPairs EA does not at a glance show me what the EA is doing with every currency pair that it is trading and the Single EA does BT's, but does not do them in visual mode with Tick data, because the Indicator will not display, which I find odd, because my correlation EA that needs to get data from Currency Pairs A & B and do a lot more calculations because of those unnecessary Money Management parameters that also need to be calculated and checked before each and every trade is opened, so with your EA one needs to babysit it in it's...
Hi @Robot Trader.
Most EA's won't display the Indicator's while in visual mode in statagy tester. You can drag indicators onto the screen, while it's running to get a visual look.
The trouble is with @robots4me R4M-BD-ADR indicator, while running the EA is it's already loaded, and your doubling up by adding it again. You can see the signals and bands, but the bands don't move, also the spread and Percent ADR get disabled as well.
I tried renaming the Indicator, but you'll get a warning and cannot run.

@robots4me did suggest

Quoting robots4me
{quote} I have alternative idea -- if the indicator detects that it is attached to the same chart as the EA then it won't display its text -- i.e. only the EA's text will be visible. Would that work?
Or, have @robots4me rename the Indicator as a dummy indi for the tester only, and remove everything except the bands and signals. And drag that in.
You will have to be nice...
You can drag most of the other R4M's Indicators onto the screen while it's running in visual, and they seem to be working.
Hope it helps.
  • #5,503
  • Aug 3, 2019 6:45pm Aug 3, 2019 6:45pm
Quoting FiveRows
{quote} Hi @Robot Trader. Most EA's won't display the Indicator's while in visual mode in statagy tester. You can drag indicators onto the screen, while it's running to get a visual look. The trouble is with @robots4me R4M-BD-ADR indicator, while running the EA is it's already loaded, and your doubling up by adding it again. You can see the signals and bands, but the bands don't move, also the spread and Percent ADR get disabled as well. I tried renaming the Indicator, but you'll get a warning and cannot run. @robots4me did suggest {quote} Or, have...
Hi FiveRows,

With the EA's I develop they all show the indicators while in Strategy Tester visual mode, I have them coded that way otherwise testing and verifying they are following the rules is made much more time consuming, therefore it is a must they are coded this way, otherwise we are looking at Demo and Forward testing, which again requires you to be in front of the chart when the trade(s) are being triggered.

That is why having the EA load the indicator automatically based on the settings inputted within the EA, should make it show in Strategy Tester visual mode and do away with having to drag an indicator on to the chart while in visual test mode, but for me using Tick, nothing shows up in Strategy Tester, which I find odd.

Regards

RT
Some people say I'm lucky, but the more I practice the luckier I get
  • #5,504
  • Aug 3, 2019 7:16pm Aug 3, 2019 7:16pm
Quoting Robot Trader
{quote} Hi FiveRows, With the EA's I develop they all show the indicators while in Strategy Tester visual mode, I have them coded that way otherwise testing and verifying they are following the rules is made much more time consuming, therefore it is a must they are coded this way, otherwise we are looking at Demo and Forward testing, which again requires you to be in front of the chart when the trade(s) are being triggered. That is why having the EA load the indicator automatically based on the settings inputted within the EA, should make it show...
Well that's one way of doing it. It's very rare that coders input that into there EA's.
I'm not sure what's involved in the process, but a dummy Indy could be a quick solution, without recoding the EA. Input the same settings as the EA.
  • #5,505
  • Aug 3, 2019 10:55pm Aug 3, 2019 10:55pm
Quoting FiveRows
{quote} Well that's one way of doing it. It's very rare that coders input that into there EA's. I'm not sure what's involved in the process, but a dummy Indy could be a quick solution, without recoding the EA. Input the same settings as the EA.

@Robot Trader, @FiveRows -- I see another issue here. When testing there are (at least) two things you are looking for:

1. Does the algorithm make sense -- from a logical standpoint.
2. Does the EA faithfully adhere to the algorithm's strategy.

To test whether the algorithm makes sense one usually studies the corresponding indicator -- if one is available. The way I've designed things is to first create an indicator and, if the indicator makes sense from a logical standpoint, then proceed with an EA that gets its data from the indicator.

To test whether the EA faithfully adheres to the algorithm's strategy one can use Strategy Tester with OHLC data or run the EA in a live Demo or Real account.


It sounds like @Robot Trader's Correlation EA does not have a corresponding indicator, so his only way of testing is via Strategy Tester. Since that is what he has done in the past then he is attempting to validate both the logic of the strategy and faithfulness of the BD strategy at the same time, using tick back testing.

I think validating the strategy's logic was discussed long ago -- we're way past that. At this point I am only interested in bullet point #2 -- does the EA faithfully adhere to the strategy. And, with the exception of a few bugs, I'd say that it does.

I am not going to add code or re-code to accommodate everyone's different preferences for testing and performing due diligence. Some people will find ways to wrestle with software -- intent that it must conform to their way of doing things. Others are more willing to go with the flow. When developing an application (or indicator or EA) there are decisions that are made along the way -- and most people understand that. When someone doesn't have confidence in the programmer then they will question why things are done a certain way. Also, people who don't have a clue about how software is developed will propose changes and suggest why can't it "JUST" do this or that -- as if their way is the only or best way, and not even considering maybe the programmer had reasons not to do it that way.

What I should have stated early on is that if you find yourself wrestling with it then consider not wrestling with it. It doesn't mean you have to toss your current approach but, rather, add yet another approach to your toolbox. It really is okay to have more than one tool in your toolbox.
  • #5,506
  • Aug 4, 2019 12:54am Aug 4, 2019 12:54am
Quoting robots4me
{quote} @jdva -- thanks for the report. Could you upload a screen shot of the Strategy Tester Report -- complete with data horizon and settings, and then I should be able to reproduce it and take a closer look.
Sure here is the report:
Attached Image (click to enlarge)
Click to Enlarge Name: ST Test GU.png Size: 76 KB
I never lose - either I win or I learn...
  • #5,507
  • Aug 4, 2019 12:59am Aug 4, 2019 12:59am
Quoting jdva
{quote} Sure here is the report: {image}
@jdva -- excellent, thank you. I'll check it out and let you know what I find.
  • #5,508
  • Aug 4, 2019 1:05am Aug 4, 2019 1:05am
Quoting FiveRows
{quote} Hi @jdva. All your open orders 30 - 33 were closed at the same time. The price closed 110 pips above your first trade. Which means the center midline band, shifted above all your sell trades, and hit the midline to exit all trades. I could be wrong, but it looks that way.
OK...but the SL was still over 100 pips higher and should have taken priority over the Midline Cross exit price...or did I see something wrong?
I never lose - either I win or I learn...
  • #5,509
  • Aug 4, 2019 1:28am Aug 4, 2019 1:28am
Quoting jdva
{quote} OK...but the SL was still over 100 pips higher and should have taken priority over the Midline Cross exit price...or did I see something wrong?
@jdva -- when I click on the image in your previous post I only see order numbers 53-68. Could you try uploading another screen shot.
  • #5,510
  • Aug 4, 2019 1:45am Aug 4, 2019 1:45am
Quoting robots4me
{quote} @jdva -- when I click on the image in your previous post I only see order numbers 53-68. Could you try uploading another screen shot.
Attached Image (click to enlarge)
Click to Enlarge Name: SR1.png Size: 37 KB
Attached Image
I never lose - either I win or I learn...
  • #5,511
  • Aug 4, 2019 2:02am Aug 4, 2019 2:02am
Quoting robots4me
{quote} @jdva -- when I click on the image in your previous post I only see order numbers 53-68. Could you try uploading another screen shot.
The order numbers are at about the middle of the screenshots not at the beginning on the left side...
I never lose - either I win or I learn...
  • #5,512
  • Aug 4, 2019 4:08am Aug 4, 2019 4:08am
Quoting robots4me
{quote} @Robot Trader, @FiveRows -- I see another issue here. When testing there are (at least) two things you are looking for: 1. Does the algorithm make sense -- from a logical standpoint. 2. Does the EA faithfully adhere to the algorithm's strategy. To test whether the algorithm makes sense one usually studies the corresponding indicator -- if one is available. The way I've designed things is to first create an indicator and, if the indicator makes sense from a logical standpoint, then proceed with an EA that gets its data from the indicator....
Hi R4M,

With respect to my Correlation EA the indicator was developed first and once we were happy it was doing what our excel spreadsheets were producing then EA was developed to trade the indicator, but the indicator is not built into the EA, so we made sure the EA launched the indicator with the same parameter values input within the EA. There was a further problem which my coder had to address which was because we are trading currency pairs A & B there were times when the data from both pairs became out of sync causing the indicator to give false readings, so extra code was needed to check the 2 pairs were in sync before making any trading decisions. It would have taken a lot longer to find this issue if we were not able to test the EA and the Indicator simultaneously with the aid of Strategy Tester

You state that your EA takes all of its trading logic from the indicator and members are reporting test results are the same regardless of whether the Midline or Opposite band are selected as the TP target, which to me means the indicator needs more testing to make sure its working correctly. This operation can be done through Strategy Tester with Tick data, the basic settings in Strategy Tester will not find this sort of issue correctly due to the lack of data within these very basic back tests.

At the end of the day it comes down to the time taken to report any issues, you can either do them quickly through Strategy Tester or just carry on with Demo account and relying upon members to be at their charts when they see something that does not look right and then report the issue. However with Strategy Tester I can get my coder to replicate the bugs I find very quickly (he can just run the same test on his PC), but when using a Demo account and an indicator that does not work within Strategy Tester this is not possible.

Like I have said I have been developing EA's for many years and I know the process from start to finish, because I have done it many times, so I look at ways that save time. At the moment with the DB strategy we have an indicator that is constantly shifting the goal posts through its continues calculations of data, we have tried to address this through the use of CC candles, but is this really addressing the issue we do not know because the rules are not being followed correctly e.g we select a TP target Midline or Opposite band but yet we get the same results this should be totally impossible. So when we select CC candles are we just getting lucky and how long will that luck last, will fixing bugs change that luck. I can't do meaningful back tests to understand what the EA/Indicator are doing, so I will leave it to others to assist you with the development of your EA, I have tried to help a number of times, but I find you to be very hard work.

Regards

RT
Some people say I'm lucky, but the more I practice the luckier I get
  • #5,513
  • Aug 4, 2019 4:12am Aug 4, 2019 4:12am
I trust you R4M. Not opening up any cows (can of worms) here.

@Robot Trader.
My plan is to run the statagy tester on all pairs for a good 6 - 12 months of data, with 6 or so optimisation options selected, to get a ball park figure of what works best for each pair. Sort the best pairs, run the mean average, and use them in the Allpairs EA. I run 4 - 5 Demo screens at once, running optimisations, you only need 1 demo account to achieve it. I copy the history data from the live account into these demo accounts for consistancy. I also have been a subsciber of Bert's data for a number of years, and use that to verify the broker data, and surprising enough, there's only a 5% difference.
You can get clean broker data through a process, but with limited time span. Some don't trust BTing, but having good data can help a great deal.
I'm hoping @robots4me has finally got to a stable build, with all bugs sorted and not introduce anything dramatic. Which means running it all again on a new build.
I'm not here to nit pick down to the tick whether a trade enters or exits in the right place. I guess I have to trust @robots4me theory and coding, and things are looking good sofar.
I don't have the luxury (or would want to), to sit down looking at charts all day. If @robots4me and @BishopDotun's system can gain a 10% profit with a 10% DD or better per week, I'll help where I can. Everyone's money is precious. And time will tell.


Quoting Robot Trader
{quote} Developing EA's is a process, where you need to check out each and every parameter and this for me with your EA is made more difficult due to not being able to see what it is doing during a BT, for other members this may not be how they do their due diligence, but when I look at a BT graph and see a spike down, I normally do a small BT during that period of the spike down (the loss) in visual mode to see what was happening, I do not have the ability to do that with your EA.
Well the only other way I can think of @Robot Trader is to run the EA on your Live account using broker or Bert's data. You can't use Demo data.
You would only need a week's worth of data. Run the EA first in Statagy Tester for a week's data, whether open or every tick, and save the results to spreadsheet. Now run the indicator in Statagy Tester (dropdown button top left), for the same period, tick or open, and use the time stamp of the spreadsheet data with the Indicator running. You would only need to verify 10 - 20 trades to validate entry/exits, and trust the EA coding.
You may find a flaw in the system. All the better for everyone.
Hope it helps.
  • #5,514
  • Aug 4, 2019 7:15am Aug 4, 2019 7:15am
Quoting Robot Trader
{quote} Hi R4M, With respect to my Correlation EA the indicator was developed first and once we were happy it was doing what our excel spreadsheets were producing then EA was developed to trade the indicator, but the indicator is not built into the EA, so we made sure the EA launched the indicator with the same parameter values input within the EA. There was a further problem which my coder had to address which was because we are trading currency pairs A & B there were times when the data from both pairs became out of sync causing the indicator...
Quote
Like I have said I have been developing EA's for many years and I know the process from start to finish, because I have done it many times, so I look at ways that save time. At the moment with the DB strategy we have an indicator that is constantly shifting the goal posts through its continues calculations of data, we have tried to address this through the use of CC candles, but is this really addressing the issue we do not know because the rules are not being followed correctly e.g we select a TP target Midline or Opposite band but yet we get the...

@Robot Trader -- sorry, but some feathers are going to be ruffled...

You are way out in left-field -- it's almost as if you haven't been following along.

From the beginning it was understood that this was a "community project" -- that is, I shared indicators and EAs hot out of the oven and, in return, people would help with testing and provide feedback. You've watched how things have evloved, right? And the hope was slowly, slowly the software would become more stable and reliable. And it has.

There was a bug related to Midline and Opposite band Exits -- well, excuse me. Okay -- shoot me, there was a bug.

With regards to repainting versus recalculation versus shifting -- OMG, haven't we beaten this to death? EVERY indicator shifts. I will repeat -- EVERY indicator shifts. You may see it, you may not. But, again, EVERY indicator shifts. Shifting is also called recalculating. The reason why it is more apparent here is because when you draw a W1 band on a lower time frame chart then that is equivalent to zooming in an order of magnitude. This has been discussed ad nauseum.

The CC candles were NOT added for the purpose of addressing shifting. Where did you get that from?

Quote
Like I have said I have been developing EA's for many years and I know the process from start to finish
Sorry -- no you haven't. You've paid someone else to develop EAs. You've never developed an EA. When I take my pickup to the shop and tell the mechanic I want such-and-such done that does not make me an automotive design engineer. The fact you do not understand the difference between repainting and shifting / recalculation tells me that. Also, the fact you think tick backtesting is the key to validating logic and flushing out bugs is not consistent with someone who has done this themself many times before.

I was thinking back on your remark about performing "due diligence". And it occurred to me your approach here has been to treat this as a commercial product and you aren't interested until you can prove beyond a shadow of a doubt that it can pass your tick test backtest. Others have understood this has always been a work in progress and that I was requesting / pleading for help with testing. Your "due diligence" remark reflects you are only interested in yourself. When others spend time testing and reporting bugs then they are helping to improve it for everyone -- not only themselves. For someone who claims to be so familiar with developing EA we've spent a lot of time posting responses back and forth about various issues, but I don't think you've reported a single bug. If the shoe was on the other foot and you had asked for my help in testing your EA, I'd have submitted a list a page long. Not because your programmer wasn't good, but because when I decide to test something then I know how to test. I'm afraid you don't. Due diligence is not the same as testing.

EDIT: With regards to testing... @FiveRows found a really good bug -- the one where Exiting at midline and opposite band gave the same result. All he did was run a short OHLC back test two times -- one time using midline as the Exit and another time use opposite bands. He was actually testing to see if the different settings worked. It probably took less than 5 minutes. And in one sentence he reported the bug, and it has since been fixed. You've been running tick back tests that last for days -- and how many bugs have they revealed? None.
  • #5,515
  • Aug 4, 2019 7:18am Aug 4, 2019 7:18am
Quoting jdva
{quote} The order numbers are at about the middle of the screenshots not at the beginning on the left side...
@jdva -- thanks. Yes -- I was searching the left-most column looking for 30-33.
  • #5,516
  • Aug 5, 2019 1:02am Aug 5, 2019 1:02am
Quoting jdva
R4M-BD-ADR EA v19214 Hello R4M, there must be something wrong with the exit conditions either on the Strategy Tester or the EA. I did a BT on GBPUSD with CC entry and exit Midline cross, SL ADR fraction=3 and got these results as an excerpt: {image} As you may see for example on O.Nr. 30 sell opened at 1.30202... SL was 1.32372 but got closed by Midline Cross (?) at 1.31334 for a loss of 100.41. I do remember on a earlier post you explained this: "2. Hard-coded MAX Stop We use a StopLoss. The StopLoss is computed as a fraction of the ADR and registered...
@jdva -- I was able to reproduce your trades using Strategy Tester in visual mode -- below is a screen shot. I don't think there is a bug -- the EA seems to be correctly executing signals from the indicator. I will explain further below...

Attached Image (click to enlarge)
Click to Enlarge Name: jdva-bug-01.png Size: 123 KB


In the screen shot above you can see the trading lines generated by Strategy Tester. I then overlayed the BD indicator on top of that.

The Order #1 closed due to a StopLoss -- nearly 150 Pips (your StopLoss was 3x ADR). And the other trades closed several bars later -- I believe due to crossing the midline. But from the visual it looks like the midline is below, right? But look at that huge spike on May 3 -- it was nearly 180 Pips. It might have pulled the midline upwards (along with the Upper and Lower bands). I can't say for sure, but a rapid 180 pip change in price could cause significant shifting in the bands.
  • #5,517
  • Aug 5, 2019 2:15am Aug 5, 2019 2:15am
Hi R4ME, I was wondering if you could share with the community your past history with software development and in particular with trading. I believe you have been a software developer for many years. Also, you worked for some large US companies specializing in commercial software development for portfolio management. It definitely influences what is written here and the level of respect.
  • #5,518
  • Aug 5, 2019 2:19am Aug 5, 2019 2:19am
I'm confused too. I look at the pic and analyzed what should have happened. I put green lines where there should be the midline target to close the trades. There are "X" symbols where the targets are but did not close there.
Attached Image (click to enlarge)
Click to Enlarge Name: jdva-bug-01_2.png Size: 62 KB
  • #5,519
  • Aug 5, 2019 2:25am Aug 5, 2019 2:25am
Looking at the way this works would it be correct to say this is by and large a 'return to mean' type strategy if traders are trying to catch the pips from outside a ADR extreme back to median line. Before you all decry blasphemy, think about it. The ADR bands are basically laying either side of a mean value based on ADR periodicity.

Isnt it the same principle as using a reversion to mean with e.g. keltner/bollinger/Moving Average bands except here we are using ADR to set band limits?
  • #5,520
  • Edited 2:56am Aug 5, 2019 2:27am | Edited 2:56am
Quoting robots4me
{quote} @jdva -- I was able to reproduce your trades using Strategy Tester in visual mode -- below is a screen shot. I don't think there is a bug -- the EA seems to be correctly executing signals from the indicator. I will explain further below... {image} In the screen shot above you can see the trading lines generated by Strategy Tester. I then overlayed the BD indicator on top of that. The Order #1 closed due to a StopLoss -- nearly 150 Pips (your StopLoss was 3x ADR). And the other trades closed several bars later -- I believe due to crossing...
@R4M...thanks, maybe you are right maybe not I don't know.
I'm currently running a forward test...
Attached Image (click to enlarge)
Click to Enlarge Name: AU R4M.png Size: 52 KB

Attached Image (click to enlarge)
Click to Enlarge Name: EA R4M.png Size: 44 KB

In these examples AU get closed at Midline for loss of 35.6 pips and EA (EURAUD) did not...both wirh same settings TP/SL ADR=2
I never lose - either I win or I learn...
0 traders viewing now