Million Dollar Pips diskuse

hi i would like to know it as well :-)

If Algo like MDP can be adopted to Futures or Equities, execution with CME is about 5 ms per order, some other venues like Chi-X claim around 30 MICROSECONDS. Sure to access DMA you need a fat wallet, infrastructure, but all is doable and money can be found is this will work.

   FH1FOREX posted:
   @MetaCoder

My VPS has 2ms of latency to the broker, how can Java reduce it?. You still have to go through a broker to give it your trade and they will pass it as a market order (ECN). My total latency, broker+banks is less than 1 second. Can you reduce it signifincantly from a technical standpoint or do you depend on using little known brokers?

only the backend on the broker side can reduce this latency.
there is a bridge for currenex with MT4 which appear to be really good. (not tested in real account, but demo account is around 100ms opening / modifying a trade, while my best today is 350ms)

The slow part is not your VPS or even MT4 client, it's MT4 server, and then MT4-2-ECN bridge then have to use. Still I consider slippage being bigger enemy than execution time. Too bad brokers where execution has been superior also produced significant slippage. So in this case I prefer bit slower broker but little or no slippage. I can adjust algos to require less speed.

   FH1FOREX posted:
   @MetaCoder

My VPS has 2ms of latency to the broker, how can Java reduce it?. You still have to go through a broker to give it your trade and they will pass it as a market order (ECN). My total latency, broker+banks is less than 1 second. Can you reduce it signifincantly from a technical standpoint or do you depend on using little known brokers?

   willgart posted:
   primeval...
at least he display all information. history, current trades etc..
and there is a bunch of real account on myfxbook using it.
so there is no real need for a real account from him.

and you?
everything hidden.... what do you want to hide?
a killing martingale system (the only real account displaying the current trades confirms this case)?
or a free scalper that you package and resell?
where is your real account using your EA only? you only have demo accounts. and your real ones are pamm accounts with manual trades...

What do you know about primeval??
I have big LIVE accounts unlike this trader who can't show a Live Account simply because his system only works on demos.

My EA is original, it uses different scalping strategies but is not what you think, a simple and unmodified copy of other EAs, you like it or not I'm still in the business and still sell my very great EA.


Hey, look at this image from this system:

I have deposited $23k on live account to test MDP Java. I'll see how it goes

   MetaCoder posted:
   I have deposited $23k on live account to test MDP Java. I'll see how it goes

Meta, what broker? Keep me posted. I am planning to invest to develop MDP-like strategy for LMAX platform. Was planning to develop for JForex for Dukascopy but not happy about their slippage. Liquidity, haha.

below information is correct.
MDP is changing setup when BT ed.
but I must say this is not a honest way.

this is the proof of what I am saying. no correct/wrong setup for MDP. all can work all can fail.

   primevalea posted:
   

   willgart posted:
   primeval...
at least he display all information. history, current trades etc..
and there is a bunch of real account on myfxbook using it.
so there is no real need for a real account from him.

and you?
everything hidden.... what do you want to hide?
a killing martingale system (the only real account displaying the current trades confirms this case)?
or a free scalper that you package and resell?
where is your real account using your EA only? you only have demo accounts. and your real ones are pamm accounts with manual trades...

What do you know about primeval??
I have big LIVE accounts unlike this trader who can't show a Live Account simply because his system only works on demos.

My EA is original, it uses different scalping strategies but is not what you think, a simple and unmodified copy of other EAs, you like it or not I'm still in the business and still sell my very great EA.


Hey, look at this image from this system:

If someone has the time, it would be awesome if they would edit out the changes to settings, and run a BT for the whole time period and post it on myfxbook. I would, but MDP keeps erroring out when i try it.

MDP Back Test means NULL ZERO NOTHING.

when BT ed there is no order execution delay slippage or spread change. the code above is keeping MDP away from blowing account. I have seen some other EA s coded same way. this is not honest for clients.

good luck all.

walker

this my comment to MDP at Bırt s Page while I was testing it on Demo.


***********************************************************************************
I will tell you with which brokers MDP will not work.
FXOPEN
VANTAGE
GOMARKETS
MBT
OANDA
those are for sure.
Fırst 3 slipps no matter what.
FXOPEN sometimes slipps 8 pips on manual order execution at liq. hours.
MBT is then one of the most slower execution has.
OANDA saying they dont use VDP but their order execution is really slow. you can eyeball it no need to count by ms.

there sould be a combination of <800 ms order execution and reasonable slippage. faster order execution does not mean that there will be no slippage. If you read the VDP manual you can understand that.

the main problem is MDP is openning too many orders in a very short period of time and when the price is fast. on the top of that it is modifiying orders and trailing stop chasing after the profit.

when all come togather with more then some hundreds of MDP working with same brokers server it becomes a mess.

so my opininon is using the group orders. and work with most profitable trades option.

MDP BT maens nothing. When BT
- No slippage
- No order Delay
- No VDP
- average tick size is totally different from real data feed even if it is %99 or %200

My opinion is the most important setp for MDp is trailing stop gap! that should be worked on.

I am currently testing it on Pepperstone.

I will share my results here if find a working setup.

walker
***********************************************************************************

That is all nice and wonderfull, but there are obviously different settings coded in the dll. I am interested as to what they are and how they perform. If they can be tested, then they can be hard-coded for use live. Maybe one of the older settings would be more profitable nowdays. That is the reason for a BT......to see HOW it responds....not test the validity.

if there is a hard coded date trick included in the code this means %99 to keep BT away from blowing account. you are right this set up can be useful for the future but if creator would determine the period of hard coded setup change before hapenning he would definetly code a algo to prevent that. but this can not be determined before happening so he used a date hard code.

well... I have good results with FXOpen ECN. the execution speed is good. I'm opening only a max of 2 trades and the group orders is on.

for the code in the robots to exclude some months, I understand this. take a look at the back test in 2011 of any EA, july (i think its this month) has no valid data, this completly kill any back test.
from a back test point of view I ignore month where the history is not correct.
what's the only real test is to setup the EA in a test account for weeks
and its more true in robots like MDP where the spikes are important.
I know... you'll say, get the dukascopy dataset to have a 99% backtest quality... still 99% not 100% and incredible long just to get the data from the site.

finally ... everybody wait for january and the new year, it will be good to see what's the EAs will do...

Sorry guys, because of a lot of 'fake-bids' for my MDP license, I will stop the bidding and simply sell it now for a fixed price of 80$ to the next person that messages me. It´s the MDP for EURUSD, full license. Cheers.

Well....I am surprised that no one other than myself has thought of this, but you can get a fairly accurate picture of the real MDP performance with a dukas data backtest. It requires knowing your slippage specs before rendering the data though, which you can get from any form talking about MDP. People are always willing to complain about how bad their slippage is. Ask about how bad the slippage is for whatever broker, when you get a good average slippage for said broker, then add that to the avg spread of the broker. For instance....on FinFX ECN my avg spread was 0.7 pips, and my avg slippage was about 4.7 pips. So, you can set the spread during the data rendering to 0.7 pips, and then set the commission to {(0.7pip actual commission) + (4.7 pip slippage)}. You get the same triggers from the 0.7 pip spread, the added penalty for the commission, and the added penalty for the slippage. Anyway, just a thought.

you can also use attached exe to change your spread. offline your mt4 and drop symboll.sel onto exe and change. after test when you go online mt4 replace the file.

have fun

   fughe posted:
   Well....I am surprised that no one other than myself has thought of this, but you can get a fairly accurate picture of the real MDP performance with a dukas data backtest. It requires knowing your slippage specs before rendering the data though, which you can get from any form talking about MDP. People are always willing to complain about how bad their slippage is. Ask about how bad the slippage is for whatever broker, when you get a good average slippage for said broker, then add that to the avg spread of the broker. For instance....on FinFX ECN my avg spread was 0.7 pips, and my avg slippage was about 4.7 pips. So, you can set the spread during the data rendering to 0.7 pips, and then set the commission to {(0.7pip actual commission) + (4.7 pip slippage)}. You get the same triggers from the 0.7 pip spread, the added penalty for the commission, and the added penalty for the slippage. Anyway, just a thought.

Not sure how you tested it but execution on Oanda MT4 is actually very fast (around 300 ms), live and demo, and slippage is next to none. The problem with Oanda is different and have two factors:
1. MDP does not work properly on Oanda as it relies on Magic number and Comment while managing orders (if comment set to empty MDP expects it to be empty in order rather omits the check), while magic is ok with Oanda, comment is not as Oanda MT4 overrides your comment and forcibly sets its own comments reffering to corresponding order number in their java platform
2. Once your order sizes with Oanda grow, they start rejecting modification/opening of your orders and you end up trying 10 times to push the order through but it comes back as 'off-quote' or 'invalid stops'. You open a new account and start small order size - no problem again, until they grow.

   walker36 posted:
   this my comment to MDP at Bırt s Page while I was testing it on Demo.


***********************************************************************************
I will tell you with which brokers MDP will not work.
FXOPEN
VANTAGE
GOMARKETS
MBT
OANDA
those are for sure.
Fırst 3 slipps no matter what.
FXOPEN sometimes slipps 8 pips on manual order execution at liq. hours.
MBT is then one of the most slower execution has.
OANDA saying they dont use VDP but their order execution is really slow. you can eyeball it no need to count by ms.

there sould be a combination of <800 ms order execution and reasonable slippage. faster order execution does not mean that there will be no slippage. If you read the VDP manual you can understand that.

the main problem is MDP is openning too many orders in a very short period of time and when the price is fast. on the top of that it is modifiying orders and trailing stop chasing after the profit.

when all come togather with more then some hundreds of MDP working with same brokers server it becomes a mess.

so my opininon is using the group orders. and work with most profitable trades option.

MDP BT maens nothing. When BT
- No slippage
- No order Delay
- No VDP
- average tick size is totally different from real data feed even if it is %99 or %200

My opinion is the most important setp for MDp is trailing stop gap! that should be worked on.

I am currently testing it on Pepperstone.

I will share my results here if find a working setup.

walker
***********************************************************************************

I am back-testing using Dukas data exclusively. Sure execution and slippage will change the picture but unless there are systematic problems with slippage (like in your example) or speed, you can get an overall idea of profit distribution curve and risks. I am currently working with another robot, also a mean-reverting strategy, but slightly different to MDP and add another momentum strategies to it, and it works miracles on demo, producing somewhat 2500% gain with 5% risk during 2011 back test. Of course it can't be expected to happen on a live account with any meaningful investment amount, but even if it will transform into 500% gain compounded and 10% risk over the year, I will be more than happy but time will tell. What back test on tick data show you is whether your algo is flawed or sound.

   fughe posted:
   Well....I am surprised that no one other than myself has thought of this, but you can get a fairly accurate picture of the real MDP performance with a dukas data backtest. It requires knowing your slippage specs before rendering the data though, which you can get from any form talking about MDP. People are always willing to complain about how bad their slippage is. Ask about how bad the slippage is for whatever broker, when you get a good average slippage for said broker, then add that to the avg spread of the broker. For instance....on FinFX ECN my avg spread was 0.7 pips, and my avg slippage was about 4.7 pips. So, you can set the spread during the data rendering to 0.7 pips, and then set the commission to {(0.7pip actual commission) + (4.7 pip slippage)}. You get the same triggers from the 0.7 pip spread, the added penalty for the commission, and the added penalty for the slippage. Anyway, just a thought.