Jag's Phantom 6 System | Page 10 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Phantom 6 System 1,540 replies

JAG Trade Journal 0 replies

JAG's Journal 0 replies

Phantom high, phantom low 0 replies

JAG fx Trade Journal 10 replies

Attachments: Jag's Phantom 6 System
Exit Attachments

Jag's Phantom 6 System

  • #181
  • Dec 7, 2010 2:05pm Dec 7, 2010 2:05pm
Quoting Saidar
And I thought I knew what forex is all about. Way confused over here will have to do some research. Thanks for the info!

By the way JForex don't have genetic optimization, this is not good news optimizing will be a problem.
It has optimization genetic optimization is just a optimization on the optimization process

It basically tries to get to the same answer quicker... the brute force optimizer still exists... however with this strategy it is tough because you are trading on a tick basis across multiple pairs... you would need a powerful computer to optimize for any benefit.

However optimizing across multiple pairs is something new since I only used Mt4 There's nothing much to optimize in v1x except the timing...

You need to understand that this is just for testing... to see what works and what doesn't... then you can use the strategy anywhere with any amount aslong as the ratio between equity/lot size is preserved.
  • #182
  • Dec 7, 2010 3:15pm Dec 7, 2010 3:15pm
Quoting sidhujag
It has optimization genetic optimization is just a optimization on the optimization process

It basically tries to get to the same answer quicker... the brute force optimizer still exists... however with this strategy it is tough because you are trading on a tick basis across multiple pairs... you would need a powerful computer to optimize for any benefit.

However optimizing across multiple pairs is something new since I only used Mt4 There's nothing much to optimize in v1x except the timing...

You need to understand that this is just...
That's true, it's just that genetic optimization is very important when testing a strategy with a lot of input settings.

Brute force testing 10000000 iterations will take a life time compared to the 10000 when using genetic optimization.
  • #183
  • Dec 7, 2010 7:14pm Dec 7, 2010 7:14pm
Quoting sidhujag
Tim,

Yes the goal is to test and see where the hiccups lie and how to prevent them. The backtests take a long time (6 month test ran overnight still not done). But that is expected because of the amount of ticks it has to go through.

Only issue I saw from 06/01 to 11/01 (I left for work here) was around end of 9 (september). It posed a problem for the EA... it went from 70k to 25k equity but closed out in profit later in a few weeks... you can visually see this to see how it can be avoided.

Note that with the platform you must trade atleats...

Hi Jag,
Ok, so jForex is Duka's platform. No problem with playing ball there. i understood their liquidity/trading model pretty much as you explained it - but you cleared up a few querys, so thanks for the explanation.

I would be happy to go with the jForex option, and eventually run the bot live if it can deliver.

I am concerned with the DD you mentioned in your running backtest - a 65% max DD is definitely a no go. I assume that we can analyse the periods price action and refine the parameters (optimise to some degree).

Also it would be good to test out my variation and charvo's to see how they handle this period.

Can you provide me with the link to download there demo. And I will do some testing if/when you can develop the variation that I was keen to test out.

Is this possible for you?

Thanks
Tim
  • #184
  • Dec 7, 2010 8:50pm Dec 7, 2010 8:50pm
hi Tim its based on v1x so that shows that dd can build up like a proof of concept the demo is on their website im busy right now typing on my phone but its not hard to find. can you msg me the full rules you want implemented?
  • #185
  • Edited 11:55pm Dec 7, 2010 11:19pm | Edited 11:55pm
Jag, sorry to bother you again.

i know that you are working on Jforex now, but would appreciate if you could spare a minute for this question?

i keep running your v1, it is profitable so far. however, whenever i restart the MT4 with a basket of floating profit, the EA will close the basket upon re-opening MT4. the EA should've ONLY close the basket on DollarTarget. (BE seems working fine, and not the cause of problem here.)

i checked
****************************
if (!CloseAll && valLastFlatEq != 0.0 && AccountEquity()-valLastFlatEq > DollarProfitTarget)
****************************

this is the only line setting "CloseAll" bool value as true. however, i cannot understand where is wrong if the EA is restarted and retrieving GVs.

I mentioned this problem 2 days ago, now I am pretty sure that the EA will always close a basket with floating profit upon restarting MT4.

@@@@@@@@@@@@@@@@
Edited:

In "init()", you have a DeleteGlobals(), what's purpose of this subroutine? I doubt that this sub caused the unexpected closure.....
what will go wrong, if i delete this "DeleteGlobals()"?

many thanks!
  • #186
  • Dec 8, 2010 12:10am Dec 8, 2010 12:10am
Quoting charvo
Jag, sorry to bother you again.

i know that you are working on Jforex now, but would appreciate if you could spare a minute for this question?

i keep running your v1, it is profitable so far....
Hi Charvo,

There are two possibilities.. one valLastFlatEq (last flat equity) got set to 0 or something so that AccountEquity() - 0 > DollarProfitTarget(something smaller than AccountEquity().

The second is that DollarProfitTarget got set to 0 so that anything in profit is considered the target.

This gets set in SetExitTarget()

DollarProfitTarget = StrToDouble(DoubleToStr(valLastFlatEq*ProfitPct/100.0,2));

if(GMTsHint > LCHour || (GMTsHint > ACHour && GMTsHint < PrepLOHour))
{
Print("Outside hours resetting profit goal to breakeven");
DollarProfitTarget = 0;
}
if(TotalTrades > 0 && (GMTsHint == PrepLOHour || GMTsHint == PrepAOHour))
{
Print("Entering new session with lurking trades... just trying to breakeven");
DollarProfitTarget = 0;
}

You can see that if you are outside session hours it will set target to 0. When you start Mt4.. you can see in your log if the EA thinks its outside hours. Also check your globals to ensure that lastflatequity is right (equity before basket is open)... If your basket is in profit then this number should be SMALLER than accountequity but (AccountEquity() - lastflatequity is still < than dollarprofittarget).

Hope that makes sense.

Good luck!
Jag
  • #187
  • Dec 10, 2010 6:25am Dec 10, 2010 6:25am
sidhujag,

Any update on the JTrader strategy ? I have mastered solo pair strategies but not sure how to approach multi pair strategies in Jtrader.
  • #188
  • Dec 11, 2010 12:59pm Dec 11, 2010 12:59pm
JTrader's tester is useless, it takes forever to run a test.

Yes it is accurate, but no one can use it because it takes too damn long
  • #189
  • Dec 13, 2010 10:21am Dec 13, 2010 10:21am
Lots of discussion full of assertions of greater wisdom, very few results shown. Where's the fruit of all the superior thinking?
Open to new approaches.
  • #190
  • Dec 13, 2010 10:45am Dec 13, 2010 10:45am
Quoting sidhujag
I didn't like some of the code it was inefficient using global vars.. I used arrays and did a bit of housecleaning.
I looked at some of your code. You haven't significantly improved efficiency. You will disagree I'm sure and if I had more time, I'd be pleased to open up graduate-level algorithm analysis textbooks and go through our respective versions of the code function by function, block by block, loop by loop, line by line. If users were frequently seeing that the EAs weren't executing all of their actions within real-time constraints, we'd have an efficiency problem. However, I've never seen any reports of that, nor seen them myself with the EAs I've posted on this forum.

You've reused far too much of the structure of my software to legitimately claim you've made significant improvements that have any appreciable effect on how the EA's execute on someone's computer.

Some decisions, such as using global variables when bar lookbacks might have served a similar purpose, were admittedly arbitrary. To remove them and claim you've upgraded the efficiency is silly.

Same with employment of arrays. Certainly I could have elected to use them in many places I did not, but it so happened I first wrote certain sections of code using individual variables. Since it worked perfectly well, there was no compelling reason to change. I preferred to stay focused on distilling the trading tactics.

It looks to me like you are introducing complexity for the sake of complexity, though I'm sure you believe there will be payoff. You appear to sincerely believe the complexity is necessary. Until the greater wisdom of your ways is proven to me, I'll maintain my current set of opinions.

You appear to have a good deal of knowledge and I hope you can focus it in a way that really does move us forward.

I'll be the first to congratulate you if your ideas prove successful.

Good luck.
Open to new approaches.
  • #191
  • Edited 8:17pm Dec 14, 2010 8:06pm | Edited 8:17pm
Quoting MetaCoder
I looked at some of your code....
"I modified MetaCoder's Phantom 6 EA and played with it the way I feel would be more profitable long term."

What part of the "way I feel" don't you understand? Keep reading.

I have no assertions until they are proven by tests (which is what I am doing).

I never claimed to "significantly increase efficiency". You seem to be stuck on that. You missed the whole point of the thread. I used your base as a starting point to implement some of my ideas.

If you want to talk about efficiency lets see what the definition from wikipedia says:

"The term inefficiency has several meanings depending on the context in which its used:

 

  1. Algorithmic inefficiency - refers to less than optimum computer programs that might exhibit one of more of the symptoms of:-

    1. slow execution
    2. excessive resource consumption (computer memory, instruction path length, auxiliary storage , power consumption"


By using the globals you not only decrease execution time but more importantly make it a much more complex testing environment when all you needed to do was use an API call. Think a little bit about what I said in my other posts if you read some of them. I talk about the ability to backtest. JForex allows it but doesn't allow global vars like MQL so as a part of the porting exercise I worked my way backwards towards the new architecture.

I am grateful for your code I really am, because without it I would have dropped basket trading for good. In fact I thanked you on numerous occasions regarding your initial EA's and its' structure however there were many things that just seemed out of place. I did not complicate it I simplified by using an API call and removing the need to use the global variables for phantom PL checking. I don't care if this reads/writes to HD or not like someone said but it is more about readability and expectancy and to me it was a "complicated" way to look at a simple problem.

To each his own your view will probably always be different which is what's so wonderful about forums where anyone is welcome to contribute his/her own idea's based on their personality.

I may have added some complexity in terms of choosing what to trade when to trade and how to trade but that's based on my own personal experiences. Some people on here expressed that they did not like that and I was and am open to change by adding switches for users to play with.

During my testing I am using the basic idea and working my way up through extensive backtesting to ensure that anything I add to the EA is backed by solid understanding as well as testing to prove the understanding.

The ideas I put into it are to make me feel comfortable trading this strategy because in the end if its true that the market is random then money management and psychology are the only thing preventing you from being successful.

I'm glad I had the starting point for your code... I merely changed what I thought was unnecessary for the environment we are testing with. For me this EA is not suited to go live as it stands and in order to properly test it I needed to make some changes.

The biggest thing for me in this strategy was the ability to test it. I simply would not accept wasting time on a strategy I can't properly test... this is why I've stayed away from trading basket strategies live... with the some of the new platforms this opens those doors for me and allows me to see what works and what doesn't (as well as try to avoid the situations to cause the strategy to fail).

If you were to show me an algorithm book and compare our code I would show you a Clean Code and Design Patterns book and tell you to go fish. The point of this whole exercise is not to see who has the better code.. but to see which idea(if any at all) derive a more profitable strategy in the long run. If any of what I said makes you think that I am belittling your work and claiming mine is better I'm sorry for that and hope you accept my apologies. If I were you I'd rather concentrate my time on improving the idea instead of personally attacking someone who's just trying to do the same thing as you. You say you're not personally attacking? Sure go ahead and waste 15 minutes on a paragraph post focusing on a simple thing like "hey you said you made it more efficient when it really isn't.. I'll bring a textbook to prove it to you! blah blah all talk no action all assumptions but no proof!". If that's not singling someone out for no reason other than that then I don't know what is!

MQL4 itself doesn't do itself any favors to help programmers to help create "good" testable code. My goal is the ability to find a branch off the tree that is actually strong enough to stand on for a long time.

Without looking to the passed the future is not possible.

Jag

  • #192
  • Dec 14, 2010 8:28pm Dec 14, 2010 8:28pm
Quoting Saidar
JTrader's tester is useless, it takes forever to run a test.

Yes it is accurate, but no one can use it because it takes too damn long
Saidar

you think dealing with multiple symbols on a tick basis would be fast? On top of that Java is interpreted so it is a little slower. However you can speed up the test to get an idea of your strategy by using OHLC ticks. I agree the platform is not any better than MT4 but for our intents and purposes it is. It really is better because it's closer to reality.

You need to test for time periods that prove difficult and then figure out either a filter or a way to avoid the period or change your strategy. After you've done that you can run over a long period over night. That's what I'm doing. However I may be just "fitting a curve" and might not actually create a strategy that works in future but I think that's an ongoing exercise. Being a programmer definitely comes with its perks since you can easily go in and tweak it as you see fit when you come across bumps in forward testing.

So far I'm getting 35% DD max in the bad times (flash crash, early 2010) and very positive results in between... I'm talking > 10-20% each month.

The EA is basically the basic phantom EA which trades 6 pairs out of 10 phantoms and trades every bar regardless of profit of open trades until x% profit is reached. I find with the amount of leverage we are using we don't need a perfect synthetic hedge but just opposing baskets that go up and down in tandem. Anyways by doing this I actually noticed better results withtout all the filters we talked about except (cutting big losers atr*3 pips in loss) and choosing right hours to trade. Other than that... your basket will pretty much be even in DD until a trend kicks in... during ranging times the DD will add up which is what you have to filter out.

There will be a case where the trend will turn to a range and your DD will not recover. This is what will seperate some traders as they will know to get out... while others won't and a margin call happens.

That's my findings so far. Still testing. Do you have it all set up? I can give you the script if your good to go.

Jag
  • #193
  • Dec 15, 2010 12:09am Dec 15, 2010 12:09am
Quoting sidhujag
Saidar

you think dealing with multiple symbols on a tick basis would be fast? On top of that Java is interpreted so it is a little slower. However you can speed up the test to get an idea of your strategy by using OHLC ticks. I agree the platform is not any better than MT4 but for our intents and purposes it is. It really is better because it's closer to reality.

You need to test for time periods that prove difficult and then figure out either a filter or a way to avoid the period or change your strategy. After you've done that you can run...
Hey thanks for the reply, if you don't test with every tick the results are extremely inaccurate. Metatrader exits and enters at the exact specified price even if you use open prices only, java trader only executes the price you select so if your take profit is 20 pips you will get anything between 20 and 100 pips depending on the size of the bars.

I did reprogram the EA to JTrader but it is not working too good think I missed something, it will be thankful if you could share yours with us.

I contacted JTrader they are working on a better strategy tester at the moment. If they can improve it (fast optimization,genetic optimization, multi core support) I will definitely move my account to Dukascopy.
  • #194
  • Dec 15, 2010 1:14am Dec 15, 2010 1:14am
Hello.
I put this post on the line of this method and got no response I understand that it's boring to answer in short I try my luck on this line because I really want to understand but not speak English as I c is very very difficult and in addition there are many terms that can translate goole translate (eg MMTU etc. ..)
So if you could take a little time it would be nice so I shall give my post here.

.................................................

can you explain how you take the leading position I read through google translate and I found that this method seemed very interesting but some had great difficulty understanding it, so imagine a person who does not speak English difficulty so I sought from the creator or any other person assistance to understand this method I did a lot of effort, but frankly with google is not clear so you can tell me how to use this method of course if you wish otherwise, well too bad I could understand you have no time for that but one can support would be very nice of you.
thank you for your help

FRENCH

Bonjour.
pouvez vous expliquer comment on prend position j'ai lu la file grâce a google translate et j'ai constate que cette méthode avait l'air très intéressante mais que certains avaient beaucoup de mal a la comprendre alors imaginer pour une personne qui ne parle pas anglais la difficulté donc je sollicite auprès du créateur ou tout autre personne une aide pour comprendre cette méthode j'ai fait beaucoup d'effort mais franchement avec google c'est pas évident donc pouvez vous me dire comment utiliser cette méthode bien sur si vous le souhaiter sinon et bien tant pis je comprendrais que vous avez pas de temps pour cela mais un peut d'entraide serais très sympathique de votre part .
merci de votre aide
  • #195
  • Dec 15, 2010 2:10am Dec 15, 2010 2:10am
Quoting Saidar
Hey thanks for the reply, if you don't test with every tick the results are extremely inaccurate....

Sure I'll post it as soon as its ready. Price is executed on each bar or tick... so if you choose OHLC it will generate 4 ticks to check for your TP or SL. I don't understand what you mean it's inaccurate compared to MT4.

It seems like its doing a ok job emulating even at 1min OHLC testing very fast.

MT4 open prices means it only checks 1 tick (Open).. while this would check 4 (OHLC).. every tick is every tick(but slow) and better than Mt4 because of tick quality and bid/ask spread availability.

Can you clarify what you meant?

Jag
  • #196
  • Dec 15, 2010 2:13am Dec 15, 2010 2:13am
Quoting panthere
Hello.
I put this post on the line of this method and got no response I understand that it's boring to answer in short I try my luck on this line because I really want to understand but not speak English as I c is very very difficult and in addition there are many terms that can translate goole translate (eg MMTU etc. ..)
So if you could take a little time it would be nice so I shall give my post here.

.................................................

can you explain how you take the leading position I read through google translate and I...
Bonjour panthere,

Comment ca va? Sa va bien! lol

I take the best phantom at the london open(until profit is met). You can play with the times but say 1 or 2 hour before london opens you check to see the best performing phantom and trade that for say the first 4 hours... if you hit profit you are allowed to open within the first 4 hours from london open.... this way you can hit multiple profit targets... or you can specify to only trade for the first 2 hours to be conservative.

Do you understand?

Jag
  • #197
  • Dec 15, 2010 2:44am Dec 15, 2010 2:44am
Hello and thank you very much for answered
I have several questions if I bother you said to me I understand no problem: D

what EA should be taken?
should open as many as there are pairs graph?
I put the EA on the chart or on any one?
orders are manually?
TP and SL manual? how determined?
where is information about which operation to do?
TF is important because I am often 5MN TF?
my broker is to zero GMT this matter?
I would have other issues if it does not bother you too: D

cordialy

FRENCH

Bonjour et merci beaucoup beaucoup de m'avoir repondu
j'ai plusieurs questions si je vous embête dite le moi je comprendrais pas de problème

quel EA il faut prendre ?
il faut ouvrir autant de graphique qu'il y a de paires ?
je met l'EA sur tout les graphique ou sur un seul ?
les ordres se font manuellement ?
TP et SL en manuel ? comment on le détermine?
ou se trouve l'information pour savoir quel opération faire ?
le TF est important car je suis souvent en TF 5MN ?
mon broker est a zero GMT cela a une importance ?
j'aurais d'autres questions si cela ne vous embête pas trop
  • #198
  • Dec 15, 2010 2:59am Dec 15, 2010 2:59am
out of curiosity I opened 6 graphs because I do not know if that's what to do next I put one of EA, the information appears on the left or there are certain pairs or there I wrote TRADING then place orders on his peers but I'm not sure this is the right approach here is a screen capture that will be clearer

FRENCH

par curiosité j'ai ouvert 6 graphiques car je ne sais pas si c'est ce qu'il faut faire ensuite j'ai mis un des EA , des informations apparaissent a gauche ou il y a certaines paires ou il y a écrit TRADING j'ai donc passer des ordres sur ses paires mais je suis pas sur que ce soit la bonne démarche voici une copie écran cela seras plus clair
Attached Image (click to enlarge)
Click to Enlarge Name: ecran1.jpg Size: 162 KB
  • #199
  • Dec 15, 2010 7:41am Dec 15, 2010 7:41am
Quoting sidhujag
Sure I'll post it as soon as its ready. Price is executed on each bar or tick... so if you choose OHLC it will generate 4 ticks to check for your TP or SL. I don't understand what you mean it's inaccurate compared to MT4.

It seems like its doing a ok job emulating even at 1min OHLC testing very fast.

MT4 open prices means it only checks 1 tick (Open).. while this would check 4 (OHLC).. every tick is every tick(but slow) and better than Mt4 because of tick quality and bid/ask spread availability.

Can you clarify what you meant?

Jag
What I meant was if I test a strategy with a fixed TP and SL, I get different TP and SL pips at each order. For example if my TP is 15 pips, I will get anything between 10 and 100 pips (when not using every tick).

In MT4/MT5, even if you only use open price MT4/MT5 still executes your TP or SL at the price specified, even if it only generates a tick at the open of each bar.

Hope that makes any sense to you. Use the visual mode and see at how many pips the trade was closed each order and you will understand what I mean.

And again thanks for you effort looking forward to your JTrader version.

By the way I contacted Dukascopy, the minimum lot is 0.001, so it is basically the same as a normal standard account at any "normal" Forex broker like IBFX. Since they use 1million units per lot 0.001 lot is the same as 0.01 lot at IBFX.

This is good news!
  • #200
  • Dec 15, 2010 11:12am Dec 15, 2010 11:12am
Quoting panthere
out of curiosity I opened 6 graphs because I do not know if that's what to do next I put one of EA, the information appears on the left or there are certain pairs or there I wrote TRADING then place orders on his peers but I'm not sure this is the right approach here is a screen capture that will be clearer

FRENCH

par curiosité j'ai ouvert 6 graphiques car je ne sais pas si c'est ce qu'il faut faire ensuite j'ai mis un des EA , des informations apparaissent a gauche ou il y a certaines paires ou il y a écrit TRADING...
Hey,

The EA must be put only on ONE chart, not all 6.
0 traders viewing now