Simple High Profit low Drawdown Triangular Arbitrage | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Triangular Arbitrage 279 replies

Triangular Arbitrage 16 replies

Triangular Arbitrage and carry trade 6 replies

Triangular Arbitrage 7 replies

Question: Trade Mechanics of Triangular Arbitrage 6 replies

Attachments: Simple High Profit low Drawdown Triangular Arbitrage
Exit Attachments

Simple High Profit low Drawdown Triangular Arbitrage

  • #61
  • Jun 18, 2017 11:26am Jun 18, 2017 11:26am
Quoting jvbJacques
@EeZeeTrader For some reason is there no commission and very high spreads and very high swaps on the demo account I opened at FXPRO. This is compared to my fxchoice accounts. According to their website they do charge commission and have low spreads and swaps so I am confident if I take more time to find that demo account it will do better. I probably did something wrong and will check into it a bit later, but for now I am using it as a worst case scenario. I like to make sure my programs can work with it so that the better accounts can make even...

how many pips is the typical profit?
  • #62
  • Jun 18, 2017 11:54am Jun 18, 2017 11:54am
Quoting jvbJacques
The very nature of triangular arbitrage is the small up and down movement and therefore if you start too low then you might never reach the profit point, seeing that unlike normal forex trading this have got a top and bottom level that changes very rarely and if so then very little so NONO to spreads. Rather go for commission and low spreads.
Agreed. My response was purely RE the 500:1 leverage. You definitely need a true ecn broker and NO dealing desk to make your strategy feasible. Will search around and PM you of my findings.
Ask and you look stupid for 5 minutes. Don't ask and stay stupid forever.
  • #63
  • Jun 18, 2017 12:05pm Jun 18, 2017 12:05pm
@dagoods
At the moment I close when the profit equals the starting loss (spread+commission loss) PLUS ($0.5 per 0.01 lot on the GBPUSD symbol).
Profit is seen as anything above what the equity was before I opened the first position(remember I can open 2 position groups at most.).
I found this to cover whatever spread and commission and even some big slippages seeing that I have got a 400ms slow server to test on.
BUT this might change seeing that I found other ways to increase the profit even further, which I am testing now ....


@EeZeeTrader
Thanks.
My tests on post #1 was with fxchoice and I have found that I can increase the spread quite a bit before it becomes a problem, it is just not build into this EA yet.
I also found quite a few other brokers offering similar or even better spreads than them, which I would like to test soon as I get time.
  • #64
  • Jun 19, 2017 6:32am Jun 19, 2017 6:32am
Interesting findings, jvbJ, and hats off, if you get it to work.

So far, I believe you are still in the Fata Morgana stage of backward testing a multiple currency pair spread dependent system. Even if you'd have tick feed for each currency pair, it still means nothing.

I'd like to share some experiences, as I have been where you are now: I assume you are placing sequential market orders to get in and out.
I once coded a TriArb EA combo with the master instance to be placed on the pair with the most frequent incoming ticks (e.g. GBPAUD) and searching for arb possibilities across all 28 pairs. I further set up 3 slave instances on one chart each (i.e. in total 4 charts open), which served nothing else than receiving calls from the master, as to when to enter and exit. This arrangement allowed me to achieve near synchronous order placement, which I forward tested via several ECN brokers and VPS providers co-located at LD4, LD5 and NY4. The best times I achieved were +/-1 ms for calculations, order placement and ping back/from the trading servers and then between 10-140ms (!) for the trading server itself. I reconfirmed those times with some of the brokers directly.

Since MT4 doesn't cater to parallel order processing and the biggest lag and therefore risk actually derives from the brokers' execution, I gave up my HFT and hedging aspirations. The lags caused 2 major challenges with the usability of a chosen triangle: asynchronous filling (or no filling at all) of orders and slippage in all directions. In order to cover the commission, swap and slippages, you need to have 2 (!) very good execution cycles (in/out), which happens very rarely.

Nevertheless, I'm always keen on hearing other people's experiences and success.
Augmenting Intelligence
  • #65
  • Jun 19, 2017 7:28am Jun 19, 2017 7:28am
Quoting jvbJacques
@dagoods At the moment I close when the profit equals the starting loss (spread+commission loss) PLUS ($0.5 per 0.01 lot on the GBPUSD symbol). Profit is seen as anything above what the equity was before I opened the first position(remember I can open 2 position groups at most.). I found this to cover whatever spread and commission and even some big slippages seeing that I have got a 400ms slow server to test on. BUT this might change seeing that I found other ways to increase the profit even further, which I am testing now .... @EeZeeTrader Thanks....
So the bigger the "starting loss" the better, then right
Why or why not use 10 pips to start with


"At the moment I close when the profit equals the starting loss (spread+commission loss) PLUS ($0.5 per 0.01 lot on the GBPUSD symbol)."
  • #66
  • Edited 8:00am Jun 19, 2017 7:45am | Edited 8:00am
@!iii!
You are correct and I have a lot more testing and work to do to get this to work.
One thing I build into the system is that it must be about to handle large slippages.
Problem is if the slippage/delay happen during the closure of the 3 pairs, but even there I have put in slippage calculations and buffers to cater for it, but further testing is needed.
As you said. Fata Morgana stage.
I have a long rode ahead, especially with normal life and working for food coming in the way. Sleep is already a strange word for me, but this is fun so far



@dagoods
No. Bigger starting losses is bad.
The market does not move much from the ZERO point and if the starting loss is too big then there is a chance it might never move far enough back up to get into profit.
Unlike the normal forex market movement where a top and bottom can always be broken an arbitrage like this can not, which is what allows for these profits, but also suffers when spreads/starting losses are too big.

Also big starting losses will take longer to move into profit by an order of magnitude from small starting losses.
This strategy benefits from low profit, high frequency trading.
Even though the profit per 3 pair trade is extremely low it adds up in the end and it has the added bonus of a predictable low drawdown, except during weekend gaps, which is still low, but higher than normal.

Also my profit calculation is working, but ultimately is not even close to the best option and just a placeholder for something better later.
As stated earlier this is one of those duct tape strips that need to be removed and replaced with something better, but it is serving its purpose for now.

I tried to stay away from pips and focus on actual USD profit.
This way I do not have to care about pip this or pop that and can ultimately make this EA broker independent.
I already made some major improvements in the EA (will post it later) to automatically calculate average spread and monitor and recover from spikes in the market.

I grew tired of all the "MONEY MAKING" EAs out there that promise the world, but fail miserably and then blame the user settings.
Also.
I understand wanting money for development, but once your EA is working for a few years then why do you still charge for it seeing that you must be rich already if it works the way you say it does.
My goal is ultimately to release this EA for free, but also it must have NO USER PARAMETERS if possible.
It must detect everything automatically.
Just place on chart and it will work. NOT THERE YET
  • #67
  • Jun 19, 2017 7:56am Jun 19, 2017 7:56am
Quoting jvbJacques
@!iii! You are correct and I have a lot more testing and work to do to get this to work. One thing I build into the system is that it must be about to handle large slippages. Problem is if the slippage/delay happen during the closure of the 3 pairs, but even there I have put in slippage calculations and buffers to cater for it, but further testing is needed. As you said. Fata Morgana stage. I have a long rode ahead, especially with normal life and working for food coming in the way. Sleep is already a strange word for me, but this is fun so far...

TY JV. Please lets consider carefully before we reject any idea. Humor me. I'm trying to picture in my mind what would happen if you purposefully allow the starting loss to be 10 pips.... what happens? In my dream mind, it still hovers around zero, but it goes up much further and down much further, than your current settings, allowing you to take profits much easier. I could be wrong, just using my imaginations. yes? no?

also, curious what happens if you start with EJ rather than GU? Just throwing some ideas that come to mind, possibly will spark imagination for betterment.

I'm so glad you want to share with the world. I always try to share freely as well. If you do release the EA I'd appreciate a copy of the source code. TY again.
  • #68
  • Jun 19, 2017 8:06am Jun 19, 2017 8:06am
@dagoods
Not sure what EJ and GU is.

Problem is that you never know ahead of time exactly what the start loss is as there is slippage involved.
Sometimes it is below 10 pips and others above.
The reason I include it in the profit calculation is just a placeholder until I find a way to automatically deduce the commission from the current open orders and to have room for slippage.
That is coming, but I only have 2 hands and time is my enemy
But it is coming.

Later I also want to adapt the EA to work on triangles from other currency pairs and then 10 pips will have a different meaning so there again I try to stay away from it.

In the mean time I am trying to gather enough funds to sponsor this project so I can focus more on it and speed up development.
Go slow is the word at the moment, but my slow is not that slow
  • #69
  • Jun 19, 2017 8:14am Jun 19, 2017 8:14am
I theorized once a slightly different approach, maybe it helps you:

1) You wait until each one of your chosen pairs runs a very low spot spread+commission (=most liquid pair at that point in time) and enter with a long and short order in order to "stock" that pair at a good price (from spread perspective).
2) Now you wait for the whole ring (USDJPY, EURUSD, EURJPY) to invert (spread compression) and exit with the "stocked" pairs in the direction of the profit.
3) The remaining stocked orders are still canceling each other out and you can either restart with step 1) or wait for them to run a profit.

The above makes use of liquid positions to pile in, while profiting from liquidity imbalances.
Augmenting Intelligence
  • #70
  • Jun 19, 2017 8:16am Jun 19, 2017 8:16am
Quoting jvbJacques
Not sure what EJ and GU is.
EJ - EurJpy
GU - GbpUsd
  • #71
  • Jun 19, 2017 8:21am Jun 19, 2017 8:21am
@spyderman
Thanks spyderman.
My ForexFu is weak



@!iii!
1) I am already doing that to a point.
I am more just calculating the average spread over time and get in if below average.

2) If I understand correctly then I am already doing that along with a mass of other calculations and buffers and stuff I experimented with in my other 225 EAs that worked.
Call this the frankenstein EA build from parts of all my other 225 EAs.

3) Not sure I understand, but then again I might be reading points 1 and 2 also wrong, but I do appreciate the input
  • #72
  • Jun 19, 2017 8:49am Jun 19, 2017 8:49am
Looking at your statements you are not doing what I am referring to.

I am saying to build hedged positions on a pair during its most liquid times - e.g. when the spread on EU is zero - and, when you have a fully stocked triangle - i.e. long & short EU, long & short EJ, long & short UJ - to exit the side (e.g. long EU, long UJ, short EJ) that is in profit due to a momentary imbalance.

Betting on imbalances twice - i.e. getting into a ring and out of it - is playing with fire during "real" (not demo) conditions.
Augmenting Intelligence
  • #73
  • Jun 19, 2017 9:13am Jun 19, 2017 9:13am
@!iii!
I will look into it later once this strategy is implemented fully and the EA is stable.
Thanks for the advice.

I'm only betting on imbalances once to exit in profit, and my EA does cater for the times is does not happen to exit with as little loss as possible.
The entry is plain and simple random (So long as the spreads are below average).

For those monitoring this thread think about helping out a bit in advice and other means if possible.

Then for everyone.
I updated the EA like crazy the past few hours.
Many thanks for everyone pointing out problems.
I also added more functionality :
1) Calculate average spread over time and only get into position if below average spread.

2) Remove the 2 level grid entry to recover from drawdown and keep to just one entry, but scale it down for a small loss over time.
So no more grid system at stoploss.
Main reason is to prevent margin call during spikes.

3) Time delay stoploss to cater for weekend gaps and downward spikes and potentially recover.
This completely removed the weekend gap problem over 2 years, but increased maximum drawdown seen during major market instability to 37%, but the adaptive system can cater for this drawdown easily.

The attached chart starts at $1000 at 2015-06-23 ending 2017-06-19 with net profit of 2,550,930.53.
Profit can be magnitudes more if split into different accounts seeing that the tester tops out at 99lots by 2016-04-28.
Before 2016-04-28 it grew exponentially and after 2016-04-28 is grew much slower sequentially seeing that it could not grow the lots size further than 99lots.
Attached Image (click to enlarge)
Click to Enlarge Name: TesterGraphReport2017.06.19.png Size: 18 KB
  • #74
  • Jun 19, 2017 9:34am Jun 19, 2017 9:34am
I had triangles and whole baskets running live while tracking every tick of equity change over many days. Neither bars nor ticks are truly synchronized in MT4 or MT5, so backtests are meaningless. Again, I don't want to put a negative spin on your efforts but direct your attention to what really matters: real execution during liquidity imbalances. I ran the previously described system parallel on 3 brokers (1 STP with 2+ large liquidity providers and 2 well-known ECN) and while it worked occasionally on the STP, it eventually stopped doing so on all 3....due to increasing slippages / execution lags.

I'll keep watching but mostly when people start putting it on real accounts. Demo is useless.
Augmenting Intelligence
  • #75
  • Jun 19, 2017 11:15am Jun 19, 2017 11:15am
why the EA either profit 2 or 3 not working in my acc? where the setting?
  • #76
  • Jun 19, 2017 1:35pm Jun 19, 2017 1:35pm
@!iii!
It will be sorry if this do not work, but here's hoping.
I'm like a dog with a bone and this bones tastes juicy
If it turns out to be a plastic bone I will still have fun.

I have now tested using 2 brokers and both make profit, all be it a massive difference in profit.
myfxchoice = 2,700,000 and fxpro = 300,000 on a $1000 account.
So far it looks like the massive amount of spread difference between the 2 accounts and indirectly because this causes less trades on fxpro.
I have seen other broker like yforex ,but they are only metatrader 4 compatible so I can do no backtesting there.
Demo forward testing will take a long time seeing that anything less than 3 months does not conform to my standards of testing.
I do have demo forward tests running, but it like watching paint dry and still on a old version of the EA which I need to replace and start over

I have got a limited amount of time so as I get more time I will test more.

I do agree that large fluctuations in the triangle does not happen a lot, but this strategy can make use of minute fluctuation, although the big ones do help.
Sometimes I get one trade every 5 days, but that is fine, so long as the cumulative profits is there in the end.
Other times I get 5 trades in a minute.

I'm hopeful this will work, but beacktesting is the best place to develop and it speeds up development a lot.
Once that is done then the demo/live testing starts.

The theory still hold true for me, and since the beginning of this venture the theory has not changed a bit, which to me is a good sign.
The EA is another monster and needs a lot of work.



@sajib0189
The EA has got limitation on the maximum spread allowed per symbol pair.
If your spread is bigger then it will not trade it.

I have got a new version that will work out the average spread and use that rather.
You can add the logs for me to look at.

But please note the EA is UNSTABLE at this stage.
Made on myfxchoice with 200:1 leverage and strict spreads.
Anything else is just luck if it works.

I want this EA to work without options and it takes time to look at everything.
  • #77
  • Jun 19, 2017 4:30pm Jun 19, 2017 4:30pm
Quoting jvbJacques
But for this instance I do have another catch. If I open a position and I see that this position stays open for more than 3 days then I will check for the lowest point over the last floating 3 days time and if this low point is hit again I will open another position in the same direction to counter the ZERO point move. This way the price only have to move half way back to get into profit and close the whole position. You will see this happening in the cart I added above on the bottom DEPOSIT LOAD.{image}
Hey Jacques, can you explain in details what do you do to your position after after the 3 days of "non-profit" because
I would like to use your strategy manually. ("Deposit Load")
So for example; GBPJPY sell @ 1 lot
GBPUSD buy @ 1 lot
USDJPY buy @ 1.43 lot
how would I do it manually for it to go back half way to be in profit and on which pair? (Your recover strategy)
thx dude
  • #78
  • Jun 19, 2017 4:49pm Jun 19, 2017 4:49pm
@savanhws
Before I open an position I save the current equity/balance.
Called EQUITYSTART.

I then hold the position until in profit.

If not profit for 5 days then I check what the lowest point was during these 5 days.
The is a running 5 days so I only check 5 days back.
For example if the position was open for 7 days then I just check for the lowest point 5 days back, not 7.
Call this low point LOW5.

Now I wait again until the price hits or goes below LOW5.
At this point I use to then open another triangle same as previous one and wait 5 days again and if in the next 5 days it is still not in profit then I stopLoss and close all positions.
I changed this part now for a lot more profit.

Now once LOW5 is hit or price goes below it then I just set the EQUITYSTART TO this LOW5 point and let it run for 5 days again and again and again.
The system will then calculate profit from this new EQUITYSTART value, which normally mean a small loss.
So far in my testing I have only hit LOW5 about 8 times in 2 years time.

Use to happen mainly with weekend gaps, but I have a solution for that in that I see it as a spike and ignore it for 1 hour after weekend and normally the price recovers.

I do the 5 day wait because most of the time the drop is just a spike and I do not want to be stopped out on a spike.
If for some reason the spike is massive then I have got a global stopLoss that will close everything at 25% drop, but I have not seen that since my first EA...
Then again I have about 10 other tricks in the system to cater for spikes and drops and slippages and to recover gracefully from them.

I am already thinking of other recovery strategies I want to try and use, but that have to wait till I get time so that I can focus on stabilising the system a bit more.

As for manual trading.
Good luck. You must be fast.
A lot of traders are done with gaps of a couple of days, but then you get those high movement days where the price moves within minutes or seconds.
You got to glue yourself to the screen to catch those and those fast movers make a lot of profit.
Does not happen often and months can go by without seeing that movement.
  • #79
  • Jun 19, 2017 9:41pm Jun 19, 2017 9:41pm
Got it, thx
  • #80
  • Jun 20, 2017 2:34am Jun 20, 2017 2:34am
Quoting jvbJacques
@!iii!...Demo forward testing will take a long time seeing that anything less than 3 months does not conform to my standards of testing. I do have demo forward tests running, but it like watching paint dry and still on a old version of the EA which I need to replace and start over I have got a limited amount of time so as I get more time I will test more. I do agree that large fluctuations in the triangle does not happen a lot, but this strategy can make use of minute fluctuation, although the big ones do help. Sometimes I get one...
The backward tests are good to iron out your bugs in the code, but that's about it.

You can skip the demo forward test right away and go to live. Demo servers are sometimes reflecting the same feed like the live servers, but not necessarily. It depends on the broker. From a filling and execution point of view, demo servers are useless for what you are seeking. The times of liquidity imbalance when your signals shoot 5 times within a few seconds (= market confusion = chaotic volatility = broker confusion = platform freeze ) are a Fata Morgana - you'll get one entry at best and the lag from the first order will kill it for the rest. My personal record was entering and exiting a profitable arb (i.e. 6 transactions) in 50ms. That doesn't happen often - at all.

The smaller the lot sizes, the less precise the hedging, obviously. So keep that in mind, when you go live.

Good luck.
Augmenting Intelligence
0 traders viewing now