Back Test vs Forward Test: In Picture.

blackmore
434
blackmore  

This balance sheet, says it all. Picture speaks a thousand words.


For a total of 50+ recent trades, the breakdown of approach towards making my EA, is explained in the chart.

I recalled not too long ago, Strategy tester keeps failing me to the point of no answers. Without any answers, I am practically left with no choice other than to leave back-testing out, altogether, for the sake of rolling an EA out.

And the result, manage to pull back into succession of winning trades.

So, in conclusion: Im taking the Strategy Tester, with a pinch of salt. Forward Test is ultimately, the way to go.

Ubzen
5300
Ubzen  
Be careful. The picture you provided could be a snap-shot at a grain of sand within a world full of galaxies.
blackmore
434
blackmore  
ubzen:
Be careful. The picture you provided could be a snap-shot at a grain of sand within a world full of galaxies.

Meaning?

Ubzen
5300
Ubzen  
The sample size is small and meaningless. If you're going to keep modifying the EA on the go then cool. It's more like a semi-automated tool then instead of an EA. However, if you think you've go it correct this time then why not run a 10 year back-test? Perhaps you're afraid it'll not look so pretty?
blackmore
434
blackmore  
ubzen:
The sample size is small and meaningless. If you're going to keep modifying the EA on the go then cool. It's more like a semi-automated tool then instead of an EA. However, if you think you've go it correct this time then why not run a 10 year back-test? Perhaps you're afraid it'll not look so pretty?

(1) Those maybe cheap grands but they real. No further comments.

(2) In trading, nothing is pretty. Figure that one out while taking a stroll on wall street, or Tokyo EDO facility, or some CBD, etc, etc wherever nearest. I really cant help you, on this one.

(3) "why not run a 10 year back-test? Perhaps you're afraid it'll not look so pretty?"

On this: MetaQuotes may have something for me on this matter. I have requisition and in the process of filling. Their office just minutes away from mine and we're neighbours. I will pop into his room, once I'm done.

Thx anyway, for your comments.

Simon Gniadkowski
17400
Simon Gniadkowski  
diostar:


So, in conclusion: Im taking the Strategy Tester, with a pinch of salt. Forward Test is ultimately, the way to go.

I couldn't disagree more . . .

A Forward Test is nothing more than a Back Test but at a much, much slower speed . . . and with many, many few samples and performed at a different date. Take your Forward Test date range, run it on the Strategy Tester under the same conditions, if you can't get similar results you are doing something wrong.

Of course the Strategy Tester has some limitations that mean not all EAs can be tested using it . . . https://www.mql5.com/en/articles/1512 but if your EA can work within those limitations then you can get the same results.

The Ultimate way to go is Live Trading . . . . not Testing.

blackmore
434
blackmore  
RaptorUK:

I couldn't disagree more . . .

A Forward Test is nothing more than a Back Test but at a much, much slower speed . . . and with many, many few samples and performed at a different date. Take your Forward Test date range, run it on the Strategy Tester under the same conditions, if you can't get similar results you are doing something wrong.

Of course the Strategy Tester has some limitations that mean not all EAs can be tested using it . . . https://www.mql5.com/en/articles/1512 but if your EA can work within those limitations then you can get the same results.

The Ultimate way to go is Live Trading . . . . not Testing.

Excellent point, UK. Technically speaking, frontside, backside, the juice of it - taste the same.

I shall translate that as DATELINE, coz im getting tired of tests, too. Thanks for this, really appreciate the 'push'. This got to be over in due time.