10points 3.mq4 - page 285
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
First, thanks to everyone who has contributed here.
I'm forward testing the latest version of FXA0.
I'd just like to make a comment.. there's way too much emphasis being placed on backtesting.
Backtesting is necessary and useful to determine whether the EA is functioning correctly, but is of very little value in determining performance.
Backtesting belies real market performance.
Unfortunately it requires months of forward testing to determine how an EA may ultimately perform. I've had to learn this the hard way.
First, thanks to everyone who has contributed here.
I'm forward testing the latest version of FXA0.
I'd just like to make a comment.. there's way too much emphasis being placed on backtesting.
Backtesting is necessary and useful to determine whether the EA is functioning correctly, but is of very little value in determining performance.
Backtesting belies real market performance.
Unfortunately it requires months of forward testing to determine how an EA may ultimately perform. I've had to learn this the hard way.You are probably right overall. But absent current results, backtesting provides some information about how the EA works under certain past market circumstances. That is not to say that the EA will work as well in a live environment but it will likely not work any better.
I have to agree, I have yet to see an EA which performs better on live or demo forward testing than it does on backtests. They may be out there, but I would venture to guess that they are the exception to the rule. There still is a lot of potential to be uncovered however, I am particularly interested in the statistical analysis methodology.
.. i also agree.. to make any forward progress
you have to forward test. i have been playing with all versions of the ea for almost a year. you have to forward test and you need to know what times of year are best for this type of ea. i would have to say that an Ea will need a full calander year to gather all of the data needed to see what it is capable of.
i have tested this on a live micro account this past year and the account has had it's ups and downs, but the account is still going.. i would encourage that all that are testing ea's to invest in a tiny account and go live.
...backtesting provides some information about how the EA works under certain past market circumstances.
Given the MT platform's backtesting algorithm's shortcomings, I don't think that this assumption is necessarily correct. Everything is suspect unless we have 100% complete and accurate tick data. Anything short of that introduces a very large margin for error.
First, thanks to everyone who has contributed here.
I'm forward testing the latest version of FXA0.
I'd just like to make a comment.. there's way too much emphasis being placed on backtesting.
Backtesting is necessary and useful to determine whether the EA is functioning correctly, but is of very little value in determining performance.
Backtesting belies real market performance.
Unfortunately it requires months of forward testing to determine how an EA may ultimately perform. I've had to learn this the hard way.Yep 100% true, only shows if its doing what you want it to, not if it will make a profit.
Given the MT platform's backtesting algorithm's shortcomings, I don't think that this assumption is necessarily correct. Everything is suspect unless we have 100% complete and accurate tick data. Anything short of that introduces a very large margin for error.
I understand your point. However, from a personal perspective if I have the opportunity to view several years of backtesting data at 90% modeling quality and 3 days worth of live testing I find the backtesting info more informative. Now of course herein lies the real problem. These aren't 2 good alternatives. One year of live data would be great, and that is the alternative I would like, but I rarely see that alternative. SO I grab what I can. The bottom line is that backtesting data should be neither overvalued nor undervalued--imho.
This is going to be my last comment.
My point is that 3 years of backtesting ESPECIALLY at 90% modeling quality is BAD information to begin with. You are basing whatever opinion you may develop on fundamentally flawed data. We all need to develop the patience to test a system for several months (at least) before making any judgments.
We're going to have to pay one way or another, time or money.
Fini.
It may seem like I didn't understand your point. I do understand your point.
veni, vidi, vici and I quit