Jag's Phantom 6 System

Hi Guys,

I modified MetaCoder's Phantom 6 EA and played with it the way I feel would be more profitable long term.

How is mine different you ask?

-Made it a bit more dynamic... I didn't like some of the code it was inefficient using global vars.. I used arrays and did a bit of housecleaning.
-Logically it is pretty different. I trade 2 baskets 1 Asian and 1 London. I clear the phantom trades at the start and end of each session instead of every 8 hours.
-If trades span across multiple sessions I just try to break even instead of a % profit. The idea is to try to escape the potential and ever lingering death basket.

If you don't know what the Phantom 6 EA is about please read:
http://www.forexfactory.com/showthread.php?t=262936

For background info on the theory read the Basket 14 EA:
http://www.forexfactory.com/showthread.php?t=252521

Have a look at the code its pretty self explanatory. I tried to comment how it works aswell as variables.

It was pretty easy to modify because the code was already pretty clean thanks to MetaCoder!

Attach to a JPY pair which updates frequently I did Geppy @ 30 mins. No indicators needed!

I know MC uses 1 hour timeframe( regardless of TF you attach to it trades every hour) but I made it dynamic (you can trade every minute if you want). I'm playing with M30...you get trades every 30 mins...it is to be determined if this is a good idea or not.

This is a non losing strategy that with unlimited margin will keep bringing in the cash... couldn't we all work for the fed?

Don't use it LIVE until we test that the EA is sound and profitable long term!!!

To trade the EA with both baskets your broker must offer the following pairs(top 5 buy and bottom 5 sells):

LONDON
1. GBPUSD Real
2. EURGBP
3. GBPJPY
4. USDCHF Real
5. EURJPY Real
6. EURUSD Real
7. USDJPY Real
8. GBPCHF Real
9. CHFJPY
10. EURCHF



ASIAN
1. GBPAUD Real
2. EURGBP
3. EURJPY Real
4. AUDJPY Real
5. GBPNZD
6. EURNZD Real
7. EURAUD
8. AUDNZD
9. NZDJPY Real
10. GBPJPY Real

But of course you can come up with your own baskets that work for your broker and that follow the criteria that they must cancel each other out (buys and sells should be hedged)

VERSIONS:
V1x - Added filter to trade (only if all current trades of same pair are already in profit).
This will stop it from trading on a retrace/trend change on your phantom runner...

PRO: Less trades, only add to winning winner
CON: The pair has to move further for you to reach your profit goal. If it was only a retrace then you should have traded. If it was a change of trend then you saved yourself drawdown.

V2 - Adds upon V1x plus changed time (not modifiable)
Changing time will help us properly detect strong vs weak phantoms. You cannot pick an arbitrary time and then say lets trade the strongest phantom.. unless you want drawdown first and then go into profit later. This tries to optimize and skip the drawdown phase and go directly into profit (highest chance to go into profit right away.. not guaranteed obviously).

It will open new baskets in the first 2 hours of each session. After that it will try to reach the profit goal or BE if session ends.

London Open Prep: 8 GMT
London Trading: 10:30 GMT
London Close: 14 GMT

Asian Open Prep: 20 GMT
Asian Trading: 0 GMT
Asian Close: 6 GMT
(No trading for Asian session on Monday)

I also added a fundamental principle of "Don't let winners turn into losers". So I added a BE to trades that go more than 200 ATR (Hourly) in profit. This will help the case where you added a bunch of trades that almost gets to your goal but then reverses suddenly... atleast you don't lose much you will have minimal lurking trades in DD (only ones that didn't reach more than 200 ATR pips in profit). Since we only add to winners... this will cut down amount of bad trades significantly.

CONS: It will take a longer time to reach your profit goal.. you will have BE's and more lurking trades possibly across multiple sessions. You have to hope that market picks a direction and goes with it.

V2a - Placed on M15 chart only!
----

This version uses some suggestions given to me by Timmy.
-Trade on the M15 chart using the bar at 15 minutes and 45 minutes only.
-Volatility filter (SMA of ATR with the indicator provided - MA-ATR). This is modifiable through the EA settings.
-Changed time to trade...
- New baskets will be opened at the start of session (11:15GMT) now... and will reset the phantoms at 11 GMT... This usually is the trend for London/NY overlap
- We can open new baskets for the first 2 hours.. ie: London (11:15 GMT until 13:45 GMT) thats 6 bars to find a potential entry.
- If basket is open you can trade until 17 GMT... at which point you will look for an exit (atleast BE)

- If symbol already has trade open... add another only if minProfitBetweenTrades is met (ie: $10 dollars set it to 10)
- Use the real pairs as anchors to start a trade.. and thereafter the EA may open trades on any phantom pairs. Idea is that anchor pairs (real pairs in the EA settings) dictate the direction of the basket... so use them for optimal entry of baskets. Versions prior to this only traded the anchor pairs (real pairs) and not the phantoms.

-There was also a bug in the code that allowed it to trade on Monday Asian session.

Thanks,
Jag

will follow this with great interest...3b is in great drawdown at the moment

I re-upped the EA since I was playing with the strategy tester and modified the time code to give me Hour() instead of the GMT time... I fixed it.

BTW im using FXPRO for testing because their quote's are supposed to be reliable aswell as the fact that they have the symbols I mentioned above.

Jag

thanks for starting this new thread

but i have one qustion on selection of basket. if you see 2nd basket, NZD is in denominator 2 times but in nominator ony once. same thing with EUR. this could lead to imbalance.

can you explain ?

ok, got the answer from EA code. may be you need to mention the real tradeable pairs for each session, for those not reading EA

updated the post thx but its pretty self explanatory look at the external vars when loading the ea

You are obviously speaking only of YOUR 03b demo. Mine is doing fine. Although it has not closed since 11/9, it has only 16 open trades and margin level is 1162%. Part of the difference is that I am using 03bx.

I am interested in this tactic of trading only during specific Market times, jag. If I can run another demo without overloading my VPS I will test this one too. Have you considered using the x factor? I'm using it on all my AddSix demos now and it does help limit the number of trades by not opening a trade on a winner unless all previous trades on that pair are winners, i.e. the pair has not reversed direction enough to cause previous trades to be losers.

Also, just for the record, MC does not "use 1 hour time frame". His EAs will work on any time frame, any chart and he tells you that in the initial instruction section of each EA's source code.

Hi Frazer,

Being someone who has modified the code I suspect either your coding skills are not that great or you do not understand the way the EA works. If you refer to line 1371 in your modified code for 03bx you will see the reason why you are wrong about 1 hour time frame. Yes you can put it on any timeframe why not? But when will it trade? Only on minute 0... what does that mean? Every hour.

The EA is really not sophisticated it is simple... but now comes the tricky bit to help us get better decisions on where to enter and when. Some say its random but I say there's some fundamentals to it. You did a good job in deciding to only add to a winner thats winning (who cares about retraces). Now we still have to judge if this will be a divergence towards the same path of failure or helps get us on a new path of an EA that remains profitable enough for you to at least regularly withdraw funds before the death trade occurs... which it will one day I guarantee it. The market will remain irrational longer than you can remain solvent... ask LTCM.

Your line of thought makes me think about possible logic changes or adding complexity(never a good thing but sometimes necessary).

Firstly we have to agree if adding to only winners is a fundamental flaw that will statistically do more bad than good if omitted.

On the plus side we will have better chance of getting on a strong trend that's not retracing or changing.

On the negative you have less trades since you're limiting entry and higher chance of longer periods of drawdown spanning multiple sessions which may whipsaw each time and cause your death basket.

If we decide that is a fundamental necessity like air food or water and you should only add to winners meaning its pretty much heading in one direction with no retrace then we should look at the daily ATR and trade if its atleast less than half of the ATR on average for statistical probability that the move isn't over yet.

Now comes timing. When do you decide to calculate the ATR. Go back to fundamentals. What does the Asian session do? It usually follows what end of NY does.. because news is released and then digested by Asia(usually futures are affected and the sessions picks up where it starts off and continues in that direction). Should we start our phantom basket near the back half of NY and then pick the strongest winner at our normal Asian time? Why not weigh this decision by the number of times the winner won at each decision step... whatever time frame you may choose.


Althought I am complicating it I don't think I'm over-complicating as I'm trying to model the psychology between different sessions and decisions that lead up to it. Is this modelling needed? Just like S/R trendlines and pivots are respected I think fundamentals have to be respected... but we have to get our assumptions right and thats VERY hard.. who knows the truth unless your the one pushing the market!

I propose the following:

1) Start the phantom trades for Asia earlier.. give time for winners to run and find out whos the strongest based on highest PL and frequency of highest PL at each decision step since phantom trades were reset.
2) You will have one winner aslong as profit is positive. If highest PL phantom spiked 2 hours ago and now its retracing you should fallback on the next strongest and steadiest winner with higher frequency of decisions aswell as almost as high absolute PL. This will help the trading being limited issue.
3) Only start the basket for real if your winner is less than half the ATR for the day starting from when you reset your basket. After you start... you really shouldn't limit trading because you may get very few trades and it will span across sessions which is a no no(you can't predict with any certainty that the next session will go your way).
4) Start the next phantom basket prior to London Open and trade for real maybe pre US or midday London... moves usually go until NY start then the likely hood of a new trend or retrace is high since a lot of money gets poured in again. Follow the same rules as above.

We WILL have to have a heuristic applied to the decision of which pair is the best phantom at trading time.. ie: if iwe fall back to a phantom that only won the highest PL once or twice and is below the bottom half of the absolute PL chart then we simply don't trade yet. This is the fuzzy logic that will make or break it I think.

Long one lol hope you can digest all that!

I should add that technicals would be the phantom logic(its like trading on a MA Cross) and fundamentals would be timing between the phantom resets and when to decide to trade.

Jag

I guess I misunderstood your statement. We were talking about two different things. I thought you meant that MC trades only on a H1 chart, but you actually meant that his EAs only trade hourly, and that is certainly correct.

Just so you will know where I'm coming from, I am not a coder/programmer. The only modification of code I do is cut and paste from one EA to another. I understand the code well enough to find the section I want to copy and then after pasting it and compiling I can usually figure out how to fix errors in unbalanced perenthesis and undefined variables. That is the extent of my coding ability.

I understand MC's EAs, probably better than anyone on this forum, other than folks like yourself who can read the code. I have been working with MC for well over 2 years now, long before we came to the FF forum. I was a part of his initial investigation into the T101 basket trading method and we developed and tested numerous basket strategies before he brought the Basket14 EAs to this forum. So I can assure you that, although I'm not a programmer, I do understand MC's EAs.

BTW, years ago the army taught me to program in Cobalt and Fortran, so I have a pretty good grasp on computer/programming logic which helps me to understand the code. But I am way too old to try to learn a new languge, although I may eventually learn the mq4 language just from my attempts at modifying the code in these EAs.

hi thx for clarifying well i have a degree in comp sci and work as a software engineer so i guess that makes a big diff lol not to harp on u but i felt your statement would be interpreted incorrectly and may mislead someone reading this thread (I also fixed the first post). Now how about improvements lets focus on those shall we what do u think about my proposed changes?

I too have studied T101's system and came up with my own strategies.. problem is testing. It probably works if you stick to the fundamentals but again its highly subjective..

BTW you only need a base of C/C++ to learn any language... they are all the same pretty much just recreating the wheel.
I know MQL5 follows the object oriented approach and uses polymorphism and objects to do its thing... aswell as Dukascopy's Java API.Along that note we can speed up testing by porting to Dukascopy's platform since it can backtest multi-currency pairs with spread. MQL5 doesn't work yet... spread is always 0.

I know it actually auto converts mql4 to dukascopy's platform I tried it quickly with 03b and it didn't work.. but porting over should be an easy excercise for this EA... probably the global variables need to be dealt with.

This way we can test in the past and actually see the validity of this system.

hi sidhujag,

i am a java programmer for over 12 years and write coding for living. learnt MQ4 recently (6 months back), i cant write a new EA from scratch, but can copy paste from similar EAs and create my own EA that works

i have read your EA code, you are still using some global variables. i dont know why. writing/reading from disk is expensive. my undrstanding is you need them only if you have multiple EAs and they share these variables. in any case speed is not an issue for this EA as it checks for adds only once in a hour, but closeall logic needs speed as PL can change quickly.

about frazer nothing more to write, its all well documented in this FF. he has a brilliant mind, objective analysis, tons of patience to test 10s of EAs on multiple accounts and report the results in a way everyone understands and analyze those results to tweak the EAs for better performance. i am sure MC couldnt have created these EAs without what frazer did (MC himself acknowledged this fact).

i will be testing your EA. but i am worried about the spread for some pairs especially the ASIA pairs. i may remove some from asia , i can understand from your code you need equal number of tradeable pairs in both sessions. i can make them 4 each. or modify the code to support 4 in AS and 6 in EU

problem is i cant run EA on my live account as they dont support MT4. i trade 02ax manually.

i can test and try your EA on my live account manually as i can trade each session. its not a 24X7. i can do one session at a time. but i am not sure how the logic works out. i would still think x factor has to be coded, since it eliminates losing trades. lets see

hey man bear with me im typing from my phone but are u saying globals are read and written to the os pagefile? this is news to me all vars are stored in memory on the stack or heap maybe if global but globals actually increase speed since stack doesnt need to pop globals cause the scope is across the life of the program and yes this is done because if you reset the ea you can continue from where u left off imagine ur mt4 closes on u . but i think ur right in this case u dont need globals for the phanton or equity calculation you can simply look back bars and its fast since you only look at the state of the pairs at start of a new bar. metacoder used globals for flexibility i think but we can prob remove them

about the spread u may be right but there is some logic in my selection i chose the most volatile pairs to complete the hedge basket while using nzd jpy and aud. So while sprwad is high they make up for it with daily range. Remember the point of pair selection is to complete a synthetic hedge to reduce margin req during dd situations so using 4 would work if it completes a hedge

Properties of GlobalVariables



As distinct from other variables, GV can be not only created from any program, but also deleted. GV value is stored on a hard disk and saved after a client terminal is closed

ive thought more about choosing the right phantom and i think timing will improve this. What i mean is choosing when to reset the phantoms and when to start trading this will filter out bad trades.

To solve the pair retrace problem we should revisit the fundamental principle of cutting loses and let profit run in this context meaning closing trades that retrace on u and adding better phantoms this will roll our basket and keep optimal pairs at all times without hoping for a direction. This whole business of only adding pairs if prev trades are in profit only identify a local optima while doesnt solve the problem completly although it is a start bwcause it can block trading and cause problems later on makes me think theres still a global solution to optimize this algorithm. If you have a cable trade and it retraces enough to change trend then we should cut it and only add to it if all trades in profit

wow interesting thx ya should get rid of them if it makes sense to then

any other pairs to substitute GBPNZD?

I'm not sure you would have to recalculate the basket... try it... and only focus on AUD JPY AND NZD... its not such a trivial task to come up with an effective basket.

Following is an excerpt regarding a working EA. It may or may not provide programmers with some ideas. The EA, in some ways works like the Phantom EA's but only trades 2 pairs and is not a hedging strategy. This info is either useful or useless. Make of it what you will.

“The EA trades two currency pairs. Based on the frequency setting, the EA checks to see if the conditions to open an order are met. It does that for both the currency pairs used for trading.


If there are no previous orders, it opens one in the direction suggested by the trend algorithm. If there are already other open orders, and if the distance from the last open order is greater than the number of pips set in the order_step setting, it opens a new order in the same direction with the same lot size.


As soon as the profit for all the open orders for each pair is equal to or greater than the target_profit it
it closes all the open orders for that pair.


On average, around 2 trades per day are opened and closed. But you can usually expect between 0 up to 4 trades a day with the default frequency setting (4 = every 12 hours).


If the total $ value of all the open orders (both currency pairs) is in loss of more than the max_total_DD percentage value, all the orders are closed. This is a “safety net” in case of a huge, sudden change in trend.


This has already happened during the demo and live testing period but the max_total_DD has never been reached, thanks to the fact that the lot size used for each order was small enough. The EA also recovers from these situations quickly, usually in a matter of a few days.


In the counter-trend phases it is crucial to have enough margin to withstand even the largest moves. For example, during the big drop during the first few days of May 2010, the EA suffered a temporary draw down of around 20% and it didn't even touch our initial investment as it was the 3rd month of trading and we were already in profit of about 40%.


Note that the profit check and draw down check are made in real time. The frequency setting determines only how often the EA will check the market for conditions to open new trades.


Strategy Settings


initial_balance (1000)
This is the balance that the EA will consider as its only available amount for trading. Many of the money management settings depend on this value. If you have an account balance of $1,000 you should leave it at the default setting of 0.01 lots (10 cents per pip m/l). If you have $2,000 you can change it to 0.02. For $10,000 you can change it to 0.1 lots. It is strongly suggested that you use this lot size (or less) as it is the right balance between risk and performance. (TR note; this is a nice feature. For example, if you have a 5k account but only want the EA to trade with 2k.......).

risk_percent if you want to have the lot size automatically adjusted set this value to (suggested) 1.


The frequency setting is probably the most important for the EA strategy. It is strongly suggested that you leave it at the default value (4). Setting values below 4 will make the EA open orders more frequently. They may increase profits but may also increase risk.


Based on our back tests and forward tests, the setting of 4 has proved to be the best balance between performance and risk. Using a more conservative setting (5) will instruct the EA to only open orders once a day which can lead to lower performance but also lower risks.


max_total_DD (30)
This percentage is the maximum loss of the actual balance that the EA can stand before closing all open orders for all pairs."


peace/out


TR

Sounds interesting.. where is the EA, roadrunner?

The name's TradeRunner sonny It's a cheap, commercial EA (yeah, go ahead and laugh). I'm not a salesman and don't want to be taken as such but I think the name is Crescendo. It really is interesting.