Rsmereka Trading Journal

[willhuang]

its been a while... and i finally figured out how to let MT4 do the optimizing.. testings thousands of different settings....

i was doing it manually before....

i calibrated tcs to give 6.3 profit factor, 29% DD, 93% accuracy using backtest data from Jan 1 2009 till now. this is with 20% risk

[rsmereka]

Originally Posted by willhuang

its been a while... and i finally figured out how to let MT4 do the optimizing.. testings thousands of different settings....

i was doing it manually before....

i calibrated tcs to give 6.3 profit factor, 29% DD, 93% accuracy using backtest data from Jan 1 2009 till now. this is with 20% risk

That's Great,

Keep in mind two very important things.

1. Backtests are only a very rough estimate. The next logical step is a demo account which is more accurate but only an approximation.

2. 20% risk for this EA is very high IMHO. Lets put it this way, if the EA hit the hard stop three times in a row (which is unlikely but possible), how much would you loose expressed as a percent of your equity.

Rick

[willhuang]

ur right.. i was being too greedy. setting it back to 12% risk

[willhuang]

in your experience.. how much worse is live trading compared to back-testing?

[rsmereka]

Originally Posted by willhuang

in your experience.. how much worse is live trading compared to back-testing?

A lot worst. In many cases I have seen, backtest results bears no resemblance to demo account performance.

This is why you will hear Pipcop and a lot of others saying that backtests are useless (feel free to interject here Pipcop). My opinion of backtests is more positive than that but not by much. I generally refer to backtests as a very rough indicator .

From there live account performance is also different from demo account performance. In my testing, all brokers fudge the demo feed (especially the spreads). The proof of this is to have two accounts with the same broker , one demo and one live and run the same EA's with the same settings on both accounts using the same computer with the same Internet feed. The difference in the results between demo and live is what will call here the broker integrity factor . I have not yet seen any broker that passes this test.

There are also other factors involved here. The speed and reliability of the Internet feed. You would not believe the performance boost I got just by moving my trading to a good VPS instead of using a computer and Internet connection from home.

The latency of the feed (the amount of time it takes to get the signal from your computer to the brokers server) is also a big factor. The latency is expressed in milliseconds (thousands of a second). To get an idea of this, find the host name or IP address of your brokers server (left-click on the numbers at the bottom right of the MT4 window which will tell you the host name or IP address of the broker server that you are connected to), then bring up a Windows command prompt (Start->run->cmd) and type:

ping server

Where server is the host name or IP address. Here is the data from my VPS to my live account broker server:

C:\Documents and Settings\Administrator>ping uk.mh-mg.classic2.alpari.co.uk

Pinging uk.mh-mg.classic2.alpari.co.uk [67.215.65.132] with 32 bytes of data:

Reply from 67.215.65.132: bytes=32 time=15ms TTL=52
Reply from 67.215.65.132: bytes=32 time=15ms TTL=52
Reply from 67.215.65.132: bytes=32 time=16ms TTL=52
Reply from 67.215.65.132: bytes=32 time=15ms TTL=52

Ping statistics for 67.215.65.132:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip  times in milli-seconds:
    Minimum = 15ms, Maximum = 16ms, Average = 15ms

This tells me that the speed of the data from my VPS to my brokers live server is very consistent and is around 15 milliseconds. This is good but not excellent. I could improve this time by finding a VPS in the UK where my broker is. Right now, all the data has to cross the pond.

Rick

[willhuang]

oh ic... if you were to give a % worse for backtest and live test. how much worse by % would u say a live test is?

also i've done a lot of research on brokers what do you think of http://www.forexpeacearmy.com/public...atcbrokers.com
the review seems good. but its an ECN. do you know the average price for commission for a ecn? it might quite possibly be the brokerage i use.

its 8dollar per lot traded. buy and sell.
http://www.atcbrokers.com/forexecn/t...issionrate.php

[PipCop]

I pretty much agree with everything Rick says. I've seen one developer create an EA which backtested an incredible profit in just one day. And I mean it went from $5000 to $5,000,000,000 in one day. Did it work on a live account? No, and the author pointed out that backtests can easily be exploited by unscrupulous developers. This is why I only forward test robots.

I wish I had an unlimited amount of money and could do them all on live accounts, because that would really shake up the EA authoring industry. But alas, we are stuck with brokers who also exploit their demo data feeds so that's about as close as we can get.

My personal feeling is that a demo account will be about 75% of what a live account can do. It's VASTLY better than backtesting.

I only backtest to get a rough idea of how the settings work and then apply them to a live account. Even then, it's somewhat of a crap shoot.

[rsmereka]

Originally Posted by willhuang

oh ic... if you were to give a % worse for backtest and live test. how much worse by % would u say a live test is?

also i've done a lot of research on brokers what do you think of http://www.forexpeacearmy.com/public...atcbrokers.com
the review seems good. but its an ECN. do you know the average price for commission for a ecn? it might quite possibly be the brokerage i use.

its 8dollar per lot traded. buy and sell.
http://www.atcbrokers.com/forexecn/t...issionrate.php

You said the difference between backtest and live test. Is that live demo or live real account. My comments are based on a demo account.

It's so hard to estimate. Some EA backtests are around 75% of what a demo account will perform and other EA's are not even close between backtest and demo account. If you want to take The Channel Scalper as an example, the backtests are about 75-80% of what a demo account would do.

I have heard a lot about ATC. As a matter of fact, I posted a question in the MegaDroid review thread that asked about this broker . I don't use them personally. I gather that they are ECN with charges of spread plus commission . BTW, their commission is $4 per full lot (according to
http://www.atcbrokers.com/forexecn/t...issionrate.php, not $8). I guess it all depends on their spread and the data feed and how much better they are than a normal broker . ATC is not listed on mt4spreads.com, I wonder why?

Anybody using ATC live that can give a thumbs up or down?

Rick

[willhuang]

Originally Posted by rsmereka

You said the difference between backtest and live test. Is that live demo or live real account. My comments are based on a demo account.

It's so hard to estimate. Some EA backtests are around 75% of what a demo account will perform and other EA's are not even close between backtest and demo account. If you want to take The Channel Scalper as an example, the backtests are about 75-80% of what a demo account would do.

I have heard a lot about ATC. As a matter of fact, I posted a question in the MegaDroid review thread that asked about this broker . I don't use them personally. I gather that they are ECN with charges of spread plus commission . BTW, their commission is $4 per full lot (according to
http://www.atcbrokers.com/forexecn/t...issionrate.php, not $8). I guess it all depends on their spread and the data feed and how much better they are than a normal broker . ATC is not listed on mt4spreads.com, I wonder why?

Anybody using ATC live that can give a thumbs up or down?

Rick

they say its $4 per full lot per side. im assuming they mean if you put in an order to buy/sell. thats 4usd/lot. if you put in another order to close it. thats also 4usd/lot so total is 8usd/trade buy and sell.

[rsmereka]

Originally Posted by PipCop

I pretty much agree with everything Rick says. I've seen one developer create an EA which backtested an incredible profit in just one day. And I mean it went from $5000 to $5,000,000,000 in one day. Did it work on a live account? No, and the author pointed out that backtests can easily be exploited by unscrupulous developers. This is why I only forward test robots.

I wish I had an unlimited amount of money and could do them all on live accounts, because that would really shake up the EA authoring industry. But alas, we are stuck with brokers who also exploit their demo data feeds so that's about as close as we can get.

My personal feeling is that a demo account will be about 75% of what a live account can do. It's VASTLY better than backtesting.

I only backtest to get a rough idea of how the settings work and then apply them to a live account. Even then, it's somewhat of a crap shoot.

The more I was writing talking about backtest verses demo and demo verses live, the more I realized that the whole thing is a crap shoot. I have seen really close results but mostly I have seen wildly divergent results.

MyFXBook.com has lots of published systems that fall into the too good to be true category. Take a look at this page. Over 5000% gain...ya...right, dream on.

Rick