All forward testing on Metatrader is flawed - page 3

 

Just one example with Alpari and North Finance broker.

D1 bar. 22nd of October.

Open:

Alpari: 2.0536

NF: 2.0542

High:

Alpari: 2.0538

NF: 2.0548

Low:

Alpari: 2.0257

NF: 2.0258

Close:

Alpari: 2.0306

NF: 2.0329

Do you see the differencies? it is the same D1 bar.

And the ddata for NF and Alpari is very similar.

Can you imagine what will be if we compare Alpari and IBFX?

 
el cid:
jlpi

I have a trader in the oilfxpro D T S S private room who uses FXDD.

He tested Alpari data and proved it is also flawed for testing and forward testing.He used Alpari for forward testing and then later compared it with backtests.The results were flawed from the same broker.I do not need to prove it.If you want to communicate with this person I will pm you his name

That leaves very little choice for reliable testing

Regards

El cid

Again I think that you are reasoning wrongly.

You say "He used Alpari for forward testing and then later compared it with backtests. The results were flawed from the same broker."

To compare forward testing and backtest results and get different results doesn't mean at all that forward tests are flawed.

I can more probably mean that the backtest results are wrong because of an EA written in a way that is let's say not compatible with the backtester.

Anyway, forward tests on a demo account are obviously not fully representative of what you will get on a real account because mainly the execution will be different. But it is the testing that is usually the closest to the reality.

And there is a difference between "not fully representative" and "flawed", no?

So in my opinion it is very important that people new to Forex don't believe from this thread that forward testing is useless. It is useful but like for everything in life you have to know what you do and what are the limitations.

All the threads and posts that say "this and that don't work, it is useless" should be considered with some distance. There are of course things that are useless, but also many people find things useless because they don't know how to use it.

If you would have given a car instead of horses to people in the 18th century, they would have say, what is this useless crap (surely with different words) , but nowadays some of us find cars useful, no?

 

Now something about forward testing and backtesting.

I backtested this TPE EA (elite section) to find good settings. It was profitable for all the pairs. Then I attach this EA to the chart for forward testing. same broker, same settings.

Do you see the differencies?

It is forward testing statement ... It is profitable but not for all the pairs.

Why?

Because his breakout EA is using high/low of the bar on H1 timeframe ...

If you attach this EA to the chart IBFX chart so you will get different forward testing results with mine. Slightly but different.

Files:
tpe12_1.zip  155 kb
 
newdigital:
Forward testing of the EAs are the profession. It is not as attach EA to the chart and forget about. Person who is doing this forward testing should "visit" Metatraders at least 2 hours per day in the beginning and, if everything is fine so it may be 4 hours per day. If someone is forward testing some EA so the reasonable question is the following: "when do you sleep?". So, forward testing of the EAs are not easy job. I am testing EAs duirng the many years and I know what I am talking about. As to the broker's data feed so the data feed may be different with many broker's mistakes. And if someone is selling EA so this person is responsible. Not the broker.

ND I just saw your post and it illustrates perfectly what I said:

You have to know what you do and know the limitations.

If you don't, don't complain.

 

If someone said: "my EAs are very good because it is backtesting very good" so it is totally mistaken. Because there are no connections sometimes between good backtesting and good forward testing.

If EA is backtesting good so it means nothing. It means only that EA may work. May not.

 

Just the other example.

Don't want to post the statement from elite section sorry.

StepMaExpert_v1.45.

Alpari broker and Fibo Group broker.

Data feed is very similar.

It is profitable for both brokers: about 2000 pips in profit for GBPUSd since the end of 2006.

But the tresults of this week:

Alpari Broker:

GBPUSD: +206 pips.

USDCHF: -10 pips.

Fibo Group broker:

GBPUSD: 0.

USDCHF: -38.

It is the same settings. And the brokers with slightly different data feed.

 

And it is the last example.

Goldwarrior EA.

Public version and public settings. EA was posted in some thread long time ago together with the settings. I am testing exact the same public version.

It is forward testing results since january 2006 with same account, same broker and with same settings. Attached.

Why it is good example.

Because the creators did not use any backtesting and did not optimize the setting. Not even single time.

They attached indicators to the chart and traded manualy just to find good settings and to improve the EA.

This EA was not backtested. The creators decided that no reason for that as backtesting will not help.

Files:
 

What is the general conclusion?

Backtesting can not say to use anything about how EA is profitable. If EA is backtesting good so it means nothing.

Of course if we can look inside the code so we will understand that EA is using high/low of the close bar and because of that we can not rely on backtesting. Or EA is using some indicator and this indicator is using high/low of the bar in calculation and we can not rely on backtesting.

But in some cases even the coder does not know why backtesting and forward testing is diferent.

But just inmagine that someone is selling EA and we can not look inside the code? Can we rely on backtesting results in this case?

Absolutly not.

Only forward testing can prove something. Just something.

And almost all EA will have different results with diferent brokers. Slightly different, or very diferent - it depends on the EA (how it was coded) and the broker itself.

If someone is telling that his EA is very good and not telling the brokers and telling about backtesting results only, and we can not look inside the code so it may be 2 cases:

- EAs must be forward tested formany weeks and we may have just general opinion about it;

- or we should simple belive to the creator but in this case the reasonable question may be, for example: "why are you, creator?". If the creator is ... Igorad for example ... or davidke20 ... or some other coder or developer so we may believe in them. May be. May be not.

 
jlpi:
Again I think that you are reasoning wrongly.

You say "He used Alpari for forward testing and then later compared it with backtests. The results were flawed from the same broker."

To compare forward testing and backtest results and get different results doesn't mean at all that forward tests are flawed.

I can more probably mean that the backtest results are wrong because of an EA written in a way that is let's say not compatible with the backtester.

Anyway, forward tests on a demo account are obviously not fully representative of what you will get on a real account because mainly the execution will be different. But it is the testing that is usually the closest to the reality.

And there is a difference between "not fully representative" and "flawed", no?

So in my opinion it is very important that people new to Forex don't believe from this thread that forward testing is useless. It is useful but like for everything in life you have to know what you do and what are the limitations.

All the threads and posts that say "this and that don't work, it is useless" should be considered with some distance. There are of course things that are useless, but also many people find things useless because they don't know how to use it.

If you would have given a car instead of horses to people in the 18th century, they would have say, what is this useless crap (surely with different words) , but nowadays some of us find cars useful, no?

Jlpi

They are flawed because they are not perfect .A diamond might be good but it isn't perfect if it has flaws.It is still a diamond

flawed - definition of flawed by the Free Online Dictionary, Thesaurus and Encyclopedia.

If these bucket shops mess with data feeds on real accounts and they do ,then the forward tests are flawed.They are only forward tests but flawed forward tests.

How do requotes come into the forward tests if they are not executed?In real accounts many trades are not executed.So this forward testing is flawed.

How does new time spreads widening come into the forward tests?These are not included in historical data.The spread may be widened higher than the price your order was executed at .This is not accounted for in back tests and does not give clearer picture of the true market spread used on forward tests.

El cid

 

I think this thread may be renamed to the following:

"All backtesting testing on Metatrader is flawed"

--------------

Sometimes some EA is showing different results during the backtesting and forward testing. And sometimes we may ask the coder: "why?". In many cases the coder will say "i don't know" and he will spend many hour just to understand why. It is concerning EA with open source code.

But if we took EA as ex4 file so no one can say nothing based on backtesting.

Based on backtestng we can say just the following: "i may work, may be, may be not, stop loss ... take profit ...".