EA tests on historical data seems defective - page 3

New comment
 
RaptorUK:
Be amazed . . YW.

Duh, you were right. Somehow I missed that LOL. Thanks.

But Here is another question for u. This goes back to my original question with which I started this chain.

Based on the article

I have downloaded the Alpari data which I am told is good quality.

I have run the MACD EA on the M5 version of the data...AND finally saw 90% modelling quality which I was so desperate to see.

Then I went back and ran the exact same test with 1 modification - I changed the period to M1 from M5.

The modelling quality went back to 25%. So I guess I am not sure I understand the following:

1) What exactly IS Modelling quality?

2) Why would period change change this metric?

I don't expect you to have answers to these BUT if you have any insight I'd love to hear it.

Thanks for the help thus far. Its made a big difference.

 

Article explains how modeling quality is obtain. Like most things around here, your have to sit and plunder what the cal means to you.

'Modelling quality' is calculated according to the following formula:
ModellingQuality = ((0.25*(StartGen-StartBar) +
0.5*(StartGenM1-StartGen) +
0.9*(HistoryTotal-StartGenM1)) / (HistoryTotal-StartBar))*100%;

From the article, seems to me that M1 is used as the measurement stick and it's value is 25%. So if you're using M1 data-testing then the Modeling Quality cannot exceed 25%. Also, there's another article posted by RaptorUK in his first response to this thread. I'll recommend reading that for additional info while you plunder.

 
marketpirate:
Can you guys please tell me this - if you have a M1 chart is there a way to scroll quickly through it. For example, suppose I am looking at the beginning year 2001, and I quickly want to see some portion of 2005.
You can also scroll directly using Fast Navigation line.
 
ubzen:

Article explains how modeling quality is obtain. Like most things around here, your have to sit and plunder what the cal means to you.

'Modelling quality' is calculated according to the following formula:
ModellingQuality = ((0.25*(StartGen-StartBar) +
0.5*(StartGenM1-StartGen) +
0.9*(HistoryTotal-StartGenM1)) / (HistoryTotal-StartBar))*100%;

From the article, seems to me that M1 is used as the measurement stick and it's value is 25%. So if you're using M1 data-testing then the Modeling Quality cannot exceed 25%. Also, there's another article posted by RaptorUK in his first response to this thread. I'll recommend reading that for additional info while you plunder.

Thanks, yea looks like i'll have to slog my way through your link also. As for the links RaptorUK posted, I read both. The 2nd (the how to for backtesting 90%) was very useful for learning the basics of backtesting. The first I quickly read through and didn't understand a lot, but that was a while back. Maybe now it will make more sense.
 
WHRoeder:
You can also scroll directly using Fast Navigation line.
Oh thanks. Good article.
123
New comment
Reason: