Different Results Each Time - Individual Optimization

AFL Programming
1

I'm hoping that someone can show me where I might be in error. I'm suspect it is something simple.

Using the following code, I am getting different results each time I perform a Individual Optimization. If I perform a Single Thread Optimization, the results are consistent each time.

Short = Cover = 0;

PRM_P01 = Optimize ("P01", 10, 1, 30, 1);
PRM_P02 = Optimize ("P02", 10, 1, 30, 1);

MA1 = MA (C, PRM_P01);
MA2 = MA (C, PRM_P02);

Buy = Cross (MA1, MA2);
Sell = Cross (MA2, MA1);

Optimization heatmap showing changes with each Individual Optimization:

2019-07-15_08-19-24

  • using Amibroker 6.30.5.
  • The database is set to local and "wait for back fill" is un-selected.
  • there are no static variables or anything other code
  • Only one symbol being tested
2

This "heat map" does not come from AmiBroker. Chances are that you have problem with the 3rd party "heat map" itself. Also check input data as described here.

http://www.amibroker.com/kb/2014/11/27/how-to-synchronize-backtesting-setup-on-different-computers/

3

thanks @Tomasz

The heatmap is indeed from a 3rd party, and when I noticed the issue, one of the first things that I looked at was that perhaps it was reading the data incorrectly.

I have checked however, and I can confirm that I am getting different CSV's directly from AmiBroker each time I run an Individual Optimization. I get identical CSV's when I run a regular Optimization.

I tried running the checksum code, as received the same result each time I ran the expiration. I also tried on a brand new installation of Ami with a new database, and still received the same results.

Is there anything else that I can look at or would anyone be willing to run the above code in an Individual Optimization and report back if different CSV's are generated with each run?

Here is a screenshot of a comparison between two of the files:

2019-07-16_07-31-36

4 
Short = Cover = 0;

PRM_P01 = Optimize ("P01", 10, 1, 30, 1);
PRM_P02 = Optimize ("P02", 10, 1, 30, 1);

MA1 = MA (C, PRM_P01);
MA2 = MA (C, PRM_P02);

Buy = Cross (MA1, MA2);
Sell = Cross (MA2, MA1);

Really struggling with this. Would anyone mind running an Individual Optimization (multi-threaded) on the above code and confirming/denying that the system is outputting results that are different each time?
Basic settings are:

1 sec database
database in (local data) mode
1 min bars
2 weeks of data
@ES#C is the symbol

5

How did you solve this later? I found the results from individual optimization are indeed different from the direct optimization without multi-threading

6

I strongly suggest using data NOT coming from real-time data stream when doing optimization. If you feed analysis from constantly changing data set (as it is the case for real-time data stream) you are going to get different results.
Since @ES#C was mentioned I guess that the data source was IQFEED (real-time data stream).