Poll: balance of the winner - page 6

New comment
 
Michel_S:
Mak:
13000

I do not believe in systems that give 100% in 3 months.
IMHO they are the result of adjustment of parameters on the history.
IMHO, such systems even on demo (not real) will sell out.

I don't understand what the fitting on the history is? I understand what you mean, but I don't understand how you are going to build a workable system without optimization on the history. And if you have optimized it at least once, then you can say that you have also fitted the input parameters of your system on history!
Why should you lie? The testing on the history and optimization are not created to fit, but to find some regularities in the market. If you optimize your EA, then you are looking for those patterns. And if others are doing the same, then you're claiming that it's a fit with the story.
I've never done any optimization on history in MT :)
In TradeStation yes, I even have a genetic optimizer for it
(which is slightly better than the same in MT)

All systems in MT I test on demo online.
This is one of the big pluses of this platform.
The performance of the system in live is quite different from what you can see in the tester.

What is CurveFitting.
This is tuning of parameters to historical data (optimization),
when parameters are adjusted to maximize the target function (NetProfit in our case).

Take a simple mouwing from the distribution kit and run it in the optimizer.
You will get good results.
Run it online and it will start to fail.
This is the result of CurveFitting.

IMHO, all the fabulous results shown here are the result of CurveFitting.
Online it will be much worse.
 
I should add, CurveFitting is not an offensive word,
but a mathematical term.
 

Eh, if only to reach the final balance in the region of 15.000 (50% of the profit of the initial deposit for 3 months) - it would be a success!

 
I, too, can't wait to drain on a very simple system -Shi Channel and about
 
Mak писал (а):
I've never done any optimization on history in MT :)
In TradeStation yes, I even have a genetic optimizer for it
(which is slightly better than the one in MT).

All systems in MT I'm testing on demos online.
This is one of the big advantages of this platform.
The work of the system in live is not what you can see in the tester.

In principle, I agree with what they say. But only in principle. Because it's not clear to me what you achieve by testing online? Do you think that everyone else hasn't thought of racing their EAs online? The online testing only allows you to detect flaws in the Expert Advisor in its incorrect response to all sorts of deviations from the ideal trade (requotes, server crashes, power outage, Internet crashes, etc.). The effectiveness of your EA in online trading you can see only after some time (probably a long time). It is for acceleration of this process and there is the run on historical data. And the Expert Advisor optimization is an art. If you want to blindly look through Expert Advisor's input parameters providing maximum profit, efficiency of such optimization will be low. I think most Expert developers with experience understand this. Therefore, they carry out optimization by including their brains (intellectual approach in optimization). And if you don't do optimization at all, how do you speed up the process of finding market patterns?
 

To avoid excessive fitting on historical data, divide the data in half - in the first half fit, in the second half test the quality of the fit.
I tested on the 2004-2006 period, and fitted (optimized) on the most downhole section for my system (autumn 2005). I found the slaughtering area after the first pass through the history with the parameters set "by eye".

 
I have about 8 years of experience in writing experts and optimising them :)
So I already know it all ...
 
Mak писал (а):
I have about 8 years of experience in writing experts and optimising them :)
So I already know it all ...

Do you have Expert Advisor, which gives confident profit after these 8 years? If you don't have one, then maybe we don't need to spend 8 years to find a solution to an unsolvable problem? I've been trading forex for 4 years, but I have not developed a reliable Expert Advisor yet, although I've written dozens of them with different ideology. For me this Championship is just of interest to see if there is sense in continuing this business or it is an unsolvable task and spontaneity of the market cannot be compensated by any Expert Advisor... And not just by Experts (who actually help in trading), but in general the currency market is a lottery in principle.
 
I don't have such an expert and, what's more, no one else does.
I'm doing it out of sporting interest, and there are some side profits from it.
For example, my GO is being slowly sold, and I even got an award at the conference
of TradeStation users in Las Vegas

I.e. the income from forex can be not direct.

IMHO, fully automated trading systems may exist,
but they will be big monsters that MT can't handle ...
The ideology will be different there too ...

And what is simpler and also makes sense are the so-called DSSRs
(decision support systems).

I think it would be difficult for any expert to survive in purely automatic mode for even 3 months.
The real trader still trades on/off the Expert Advisor, sometimes changing parameters, etc.

Currency market is not always a lottery.
But IMHO, it shows up only on large times and timeframes.
Anything less than a day or maybe a week is random wandering and a lottery.
 
Mak wrote (a):
I don't have such an expert, moreover, no one else does.
I do it from a sporting interest, well, plus it has a side income.
For example my GO is quietly selling, and even got an award at a conference of
of TradeStation users in Las Vegas.

I.e. income from forex may not be direct.

Congratulations, you are another one of those who create a fuss around forex to get indirect income from the naive trying to get direct income and who have no idea that they are goats (meat), to put it mildly.

IMHO, fully automatic trading systems can exist,
but they will be big monsters that MT cannot cope with ...
The ideology will be different there too ...

And what is simpler and also makes sense are the so-called DSS
(decision support systems).

I think it would be hard for any expert to survive in purely automatic mode for even 3 months.
The real trader still trades on/off the Expert Advisor, sometimes changing parameters, etc.

Even knowing this, you still participate in the Championship to share (at random, of course) the free money of MetaTrader. Although, you do not believe that anyone will ever get this freebie.

The currency market is not always a lottery.
But IMHO, it only shows up on long times and timeframes.
Anything less than a day or maybe a week is random rambling and a lottery.

And since the currency market is a lottery, then there is no need to mess with people's minds with clever statements about fitting EAs on historical data, and advertising your GO at the same time. At the end of the day it is still a LOTtery. And the winner is, to a large extent, the lucky one, and not the one who has won the award for his CS in Las Vegas.
1234567891011
New comment