Simple High Profit low Drawdown Triangular Arbitrage

@EVERYONE
WAIT FOR VERSION 18.
I discovered a but in the MT4 version of the EA in that it does not calculate the SELL trades correctly.


@eatrader75
@Post #174
The logs ..... ooooo ... the logs
I went log crazy

There is a lot of information.
The stuff about the Illuminati, Bermuda triangle and JFK assassination you can ignore.

Look at the "*** OPEN" and "*** CLOSE" lines.
They contain what the 3 prices were before the tickets were opened and then after.
Then it also contains the amount the price slipped "SLIP" after the trades were opened.
I can see that GBPJPY slipped by 1.6 pips, GBPUSD by 1 pip and USDJPY did not slip, which is not to bad.

You can also look at the actual opening of trades.
Look at the first log line before the first trade opened and its timestamp.
Then look at time of the last trade and you will see how long it took to open the 3 trades.
In your case it took about 100ms to open a trade so 325ms to open all 3, which is not to bad from what I have seen, but compared to FIX API it is bad.

I am still looking for a way to improve on the open and close time delays.
One way is to communicate with the trade server every 25 seconds to prevent the connection from timing out so that there is no login delay before trades, but I do not think brokers will like 2880 signal per day that is not actually open or close trade signals, but other than that or the FIX api I am not sure how to speed it up ... YET.
Suggestions are welcome.

According to me the time to open/close a ticket should not be longer than 2 times your connection speed to the server.
I have developed a multi threaded version of the EA, but it runs even worse, because each thread then need to sign on seeing that they run in seperate metatrader instances, where a single thread only signs on once.

@Post #180
Nooo...
Why did you restart it
Yes you missed it because of the 100 minute start delay.
Too bad.
Next time
You might want to look for a broker with a better spread and commission also.
The distance between the top white channel and bottom red channel is a bit big.
It will work, but only once in a while like mine

@Post #174
Looking at the logs I still see a problem in that the trades contains the same amount of lots.
Is this version 17 of the EA ?
That is wrong.
I will look into it a bit.
Found the bug.
Will be fixed in version 18

Fixed
Attached is the new EA.

Someone please tell me how to remove the trade delay on trades in metatrader.
It is really frustrating.
500ms is a long time to wait for a trade to go through, even 100ms is a long time if you have got a 6ms link to the server.

Lock and load

"Someone please tell me how to remove the trade delay on trades in metatrader" _ you may want to ask XMESS7, he is a really good guy and a awesome coder!

https://www.forexfactory.com/xmess7

Jacques,

@Post #174
Thanks for help with the logs!
I run 2 MT's: FxPro and Global Prime. FxPro is quicker (9ms) and has higher spread rate on these pairs: 1-3 pips or more. GP has less than 1 pip with 7$ commission per lot but the ping is 79ms and had no trade. Anyway, if we manage to go live with, will need a VPS. Maybe I'll try Darwinex too, suggested by Fxentropy.

@Post #180
The lots weren't well balanced (v16) so I just wanted to save the 38$ demo profit and try your v17. As you see the attachment in Post #166, the first set of trade was ok only the second set went wrong.

I'll try multiple pairs with reduced account size so possibly will get more trades to analyse.

Use a good vps with good ms speed. There are a lot in market connecting less then 1 ms speed, first check ur forex mt4 location from ur broker.
for example : USA, U.K. Eu, then buy the appropriate one.

@eatrader75
Yes the bug was in the 2nd part of the coding also.
There is no way to backtest the MT4 code to make sure it runs correctly so it only revealed itself during the forward test.
Thanks for the log

MT5 I could backtest and it is running correctly since version 16 using the new coding.



@aszain02
I helped someone with a VPS which have got a 17ms connection to the MyFxChoice servers, but the trade delay is still about 500ms for 3 trades.
Maybe there is something else in play here.
Maybe 17ms is just to the MyFxChoice MT4 server and then it still need to go to the actual market from there, but that speed I can not test seeing that I do not know how they connect.
Then again it is pure speculation.

I just need to find a way to reduce the delay, because a lot can happen in a second.

Then I helped someone else with a VPS connecting to another broker with a 3ms to 5ms connection that gets a trade delay of a WOPPING 2 seconds for opening 3 trades.
That is crazy slow.
So something else must happen in the background.
Same metatrader version.

Maybe it also depends on the broker themselves.
Just not sure how to find out which broker is fast.

I was thinking of testing on Dukascopy, but for that I need to convert the EA to java to run on JTrader, which I am planning to do seeing that Java is my speciality.
From what I heard Dukascopy have got very fast order execution.

Yet another theory is that the delay is done on purpose by the broker on demo accounts so as not to slow down live accounts.
That is extremely difficult to test.

can u plz give a try on other brokers :
tickmill
jfd
paternosters
Regards

@aszain02
I live in South Africa.
My ping to most brokers are about 600ms to 800ms.
For me to test without a VPS close to each of those brokers is next to impossible.

The VPS I did use is someone else's and I can not do with it what I like out of respect.

I would love to test it.

Still havent got one trade yet....its normal?

Jacques,

I might be wrong trying with other pairs (e.g. EURJPY, EURCHF, CHFJPY) - I know, you've designed the ea for GBPJPY, GBPUSD, USDJPY. Simply, I wanted to see more trades on the forward test. I've got the same good results on MT5 backtest (V16 backtest) with these new pairs. It's true, no drawing. No drawing on forward test either.

@Fxentropy
Yes that is normal.
Days can go bye with no trade, but I found the past few days to contain a couple of hits.
Just give it time.
According to your chart there was no hit yet, but they did come close a couple of times.

I am working on a new model to hopefully cater for more slippage and more profit in a shorter period of time.

Also keep in mind that once the EA is stable then I will be adding all possible triangle as @eatrader75 already started to test
That should speed up the number of arbitrage hits exponentially, but it is easier to test and develop using one triangle for now until I know everything is working correctly rather than multiple triangles and then also trying to figure out in which triangle a bug is.
This strategy is extremely expandable.



@eatrader75
My original idea was to run it on all possible triangles at the same time so that you get more hits.
So yes it should work on other triangles like your pairs, but I did focus on GBP first until I can stabilise the EA.
Once stable then I can incorporate the other pairs.

The nice thing about this strategy is that you can add as many pairs as you want with no extra drawdown.
Just more hits and profit, but only as long as the same EA handles all the pairs seeing that they can not be opened at the same time or risk hitting margin call.

Just make sure when you backtest to use EVERY TICK BASED ON REAL DATA and not EVERY TICK.
I suspect you used EVERY TICK.
It is not as accurate, but it does look very good.

As for not seeing the chart.
You must attach the EA to the 2nd symbol, which in this case is EURCHF.
It is drawing the cart on EURJPY, but somewhere off screen close to Jupiter.

I made the backtest now with "every tick based on real data", attached to EURCHF. I've got back the drawings. Just observed, there's an extra red line on top of the price. Only the V16 backtest works, the V17 and V18 don't.
The results aren't so consistent as with "every tick" only.

On forward demo I can run only 1 ea per MT4 at the same time. So I may need multiple MT's in different folders or you just code the option to change the magic number.

Thanks
Attila

ok, for the sake of testing....i added these pairs lol

3 trades of 0.01 i had in the morning....

hello
if you need to test something in my VPS, i also have same cent account with Justforex, FortFx and XM.com


we are here to support you
Don't worry ask me if you need my help

ciao

@eatrader75
Strange that you can not run multiple EAs in your metatrader.
I have no problem.

But in any case.
Do not run multiple copies of this EA on the same account, even if on different terminals.

The EA uses a lot of margin and if 2 of these EAs opens positions at the same time then you are guaranteed a margin call and a failed trade.
At the moment it uses almost 50% of the available margin so 2 EAs might work, but with just a tiny drawdown and you hit margin call.

It need to be coded into the EA to handle this situation so that it does not open trades at the same time, but still use all triangles as faster entry points.


@Fxentropy
Read above message to eatrader75.

This will not work.
Well it might, but if it tries to open trades at the same time then it will crash seeing that it will hit margin call.
I still need to build in the checks to prevent this situation, but this EA can only handle one triangle at a time.

My focus at the moment is to decrease the slippage by decreasing the time it takes to open and close tickets.
So far no reasonable solution.


@raffus
Thanks.
I need all the help I can get and I have already some wonderful patrons that are lending me their VPS.
Problem now is getting the time to do the actual development.
Not cheap so I must lick boots to save up for free coding time

That server does look good, but I also have access to a 6ms server connection terminal, but even on that one the trade delay is upwards of 600ms.
I need to solve this delay dilemma first before I can really make use of further vps tests.

Thanks for my VPS providers seeing that I would not have found that problem without a vps.
Now just to solve it.


@dagoods
I do not care for metatrader much.
Especially after the delay problems.
I know more than 25 different programming languages and learning more APIs unique to brokers like oanda is no problem for me, especially if it can do better trade execution.
I heard the same about JTrader on Dukascopy and the FIX api.
Problem is getting a demo account to test on them.
FIX normally require a live account.
oanda last I checked also requires some sort of payment for using their API if you do not have a live account with them.

Jtrader is free, but dukascopy only have got a 14 day demo account trial.
Luckily Java is easy to program so I should be able to do it quickly .... well ... during my free time, which is limited.

I does seem like the only option is to make use of the proprietary broker platform for speed.
Metatrader seem to have fallen behind on development and is slow, even on a fast broker.

Problem with metatrader is that even if you have a fast connection to the metatrader server it does not mean that metatrader have got a fast connection to the market.
From what I heard is that what makes matters worse is that metatrader is a single threaded application, meaning it can only do one trade at a time and therefore if it becomes busy then it will slow to a crawl, relatively speaking.

Had first set of trades on GU GJ UJ

@Everyone
Hang on.
This slippage problem is killing this strategy.

It can handle a couple of pips slip, but 5 or more pips will eat up the profit.

I am working on an idea that just might kick the slip between the legs.
Wait a bit for me to create and test it.

watching the image of @Fxentropy and mine

if the value of lot of USDJPY was the same of other 2 currencies, the trades will be closed with profit.

Actually not

i know u told me that is due just to balance/hedge the arbitrage but i see i many case .....this "issue"

@raffus
Not quite, but I am working on a theory where an unbalanced arbitrage where all lots are the same might help.
Bigger draw down, but bigger profit and more trades.
Still testing it.
Requires quite a bit of changes.

Hopefully will have something to test tomorrow.

If it wasn't for slippage then the trades you saw would have been in profit.
Metatrader is way too slow, a direct broker api will work better on this strategy, but I might have a solution that can work even with slippage.

Look.
Slippage is never good, but I might have a way to counter it.