Error processing negative value data (back-adjusted)
Author: lookingbackon
Creation Date: 5/26/2011 2:07 AM
profile picture

lookingbackon

#1
Hello,

Got a issue with a future symbol data with some negative value data due to back-adjusted for continuous contract purpos. The problem encountered are as below:

1. For such negative data (Daily), the true "price change" is adversely expressed in the price chart. e.g.: the true price change is up +2%, but the data window shows -2%

2. In case the symbol is clicked for strategy backtest, the Error report window always shows that the program is experiencing some problem and has to be closed. Then WLD closed.

Could you pls check what could be the issue? Or the WLD is not supposed to handle NEGATIVE value data?
Thanks
profile picture

Eugene

#2
Re: #2. This could be absolutely anything.

In Preferences, turn off any visualizers you might have enabled, leaving just the Performance and Equity tabs. Re-open the strategy for the settings to apply. Re-run the backtest. Does it help? If not, please help us reproduce the crash by following this checklist:

How to report a Problem?

P.S. I doubt that negative prices were ever considered seriously. (BTW, there are better back-adjustment methods that do not lead to neagtive prices e.g. ratio adjusted data).
profile picture

Cone

#3
If you need a confirmation, I confirm that Wealth-Lab is not compatible with negative price data series, i.e., symbol data. Some analysis will work, but some won't, so you should consider all results "garbage out".
profile picture

lookingbackon

#4
Looks like the negative price data doesn't make more sense for the backtest. So I have to fix the data issue first.
Anyway, thanks for the clarification and advice.
profile picture

akardar1

#5
Hi All - this was about 4 years ago, is this still the case?
(i.e. WL doesn't support negative prices that can be displayed in a back-adjusted data series).
I can add a workaround to the code (which would add a constant to an entire data series if there is a negative value present). Any other solution within WL ?



profile picture

Eugene

#6
Hi,
It is the case and will be because nothing has changed in the design. Why not consider finding a better source that back adjusts the data without producing negative prices (for example, ratio-adjusted data)?
profile picture

akardar1

#7
I was seeing errors with because I was using a OHLC series with negative data (backtested continuous).

I have adjusted the data with a constant value, which removes the negative data.
I have 2 CSV files (at this point I am just running a test to make sure the system works with the new data series).

file 1 (with negative values) = the chart loads correctly. I can apply a strategy to the chart. But I know its pointless due to negative values.
file 2 (adjusted data) = the chart loads with a weird time scale divider, (screenshot 1). When I apply a strategy I get an error message (screeshot2):

"Error in visualiser: RabPage: {Trade Graphs}, removing.
Index was out of range. Must be non-negative and less than the size of the collection.
Parameter name: index"

The only difference between the two files, is the fact that the second is adjusted to remove the negative values.
They are both CSV files. I am happy to raise a support ticket if needed with the files.



profile picture

akardar1

#8
Adding 2nd screenshot
profile picture

Eugene

#9
There is something visibly incorrect with the dates in your data on the right (the "corrected" one). Please inspect it manually, go line by line, making sure the data is valid, and review your ASCII DataSet settings. If a visualizer unloads itself like Trade Graphs, something is really wrong with the data. Also, if you notice any discrepancy on the By Period tab, you know it's true because this PV is sensitive to bad dates.

Since the problematic data does not contain negative values, I think it's become suboptimal to continue discussing it in this thread which deals with negative values in data.
This website uses cookies to improve your experience. We'll assume you're ok with that, but you can opt-out if you wish (Read more).