Bug Reporting

[Julia CJ]

Hi Robocod,

The value shown for the stop is wrong. Please check.

This issue was already fixed. The fix of the issue will be included in upcoming release.

[Julia CJ]

Hi Robocod,

Could you please specify the combinations of streams you used for testing?

[robocod]

Hi Robocod,

Could you please specify the combinations of streams you used for testing?

For example, EUR/USD, GBP/USD, USD/JPY, etc. m5 chart, correlation of last 20 bars.

You might find this screenshot interesting. It's an indicator that I developed. On the top is a version using the mathex.correl(), on the bottom is my own implementation.

The correlation is the triangular matrix in the upper right.

I also dumped the raw data to Excel to test this.

[Julia CJ]

Hi Robocod,

I have not found the Excel file with a data.
Please send this attachment again.

[robocod]

Hi Robocod,

I have not found the Excel file with a data.
Please send this attachment again.

I didn't attach the data. I just said that in my research I tested my hypothesis by using Excel. If you need the data I will have do all this work again. Alternatively you can just get your own data.

Take 20 values from EUR/USD, and 20 values from GBP/USD. Use the mathex.correl function on it and see the result. If I have time I will write a short indicator to demonstrate it.

[robocod]

I've made a small test indicator that shows the effect. the indicator loads for data sources: EUR/USD, GBP/USD, USD/JPY and AUD/USD. It uses the time-frame of the chart to which it is applied. It then computes the correlation of each instrument pair and outputs the results to the Event trace.

You can see in the example trace below that some values are 0. This is highly unusual for a correlation (0 is a valid value, but the fact that you get 2 or 3 results that are all exactly 0.00000 is just not likely at all).

I found that I seemed to get more zero's with small time-frame, e.g. m1, m5. With bigger time-frame it seemed better. I also got more zero's if I replaced the USD/JPY with something else (e.g. EUR/GBP). This further lends to the hypothesis that there is some kind of numerical under-flow happening (since JPY based pairs have numerically higher values, e.g. 116.00 rather than most forex which are smaller numbers).

Code: Select all

USD/JPY   TEST(USD/JPY)   Trace: 'i = 3, j = 4, correl = 0.472084'
USD/JPY   TEST(USD/JPY)   Trace: 'i = 2, j = 4, correl = 0'
USD/JPY   TEST(USD/JPY)   Trace: 'i = 2, j = 3, correl = 0.143555'
USD/JPY   TEST(USD/JPY)   Trace: 'i = 1, j = 4, correl = 0'
USD/JPY   TEST(USD/JPY)   Trace: 'i = 1, j = 3, correl = -0.452865'
USD/JPY   TEST(USD/JPY)   Trace: 'i = 1, j = 2, correl = 0'.

test.lua

(4.29 KiB) Downloaded 586 times

[Julia CJ]

Hi Robocod,

Thank you for the detailed description. The error was caused by the existing algorithm with high threshold. To fix the issue, we have changed the threshold value. The fix will be included in the upcoming release.

[robocod]

I have some more errors that seem to come from the new TradingStation version when run with Windows 10. I don't really have any clue what it causing this, the error message is not helpful, and I don't use Windows 10 myself, so I can't reproduce. Can I send you details via email?

[Julia CJ]

Hi Robocod,

Please send me the description of issues and possible snapshots at this e-mail: ycherepanova@gehtsoft.com.

Thank you.

[mechanicjon]

Hello,

Latest update of TS won't configure Hotmail account. I tried original settings plus every other combination. I was able to setup Gmail account first try. Also when entering the email config. program none of the common email providers are listed like before. IE: Hotmail, Gmail, yahoo etc.

Thanks
Jon