My firm just began evaluating MetaStock. We already have in place a data provider so our intial focus was ensuring that the data being imported into MS was accurate. We've noticed, especially with indices where volume can be quite high, that "The Downloader" program appears to be manipulating the volume when it's greater than 9,999,999. Numbers greater than this may be rounded up or down or just be off by a few digits. We haven't been able to identify any pattern. We've even gone as far as to create some test files verifying our findings but no solution.

As the technical support person I've been asked to investigate this further and I thought I'd post here to see if other users have experienced similar issues and have either identified the problem or is this a bug. I might add that all the other price fields import exactly as they should - this problem is just limited to volume data.

I did read where some users were havings issues with volume data but those posts appeared with respect to older versions. We're using v10.1.

Appreciate any comments

Iris

This issue is due to the data file format and general numerical representation of decimal numbers that MetaStock uses.

The data files are based upon the original CompuTrac format which uses a single precision floating point number to represent each of the following data points: open, high, low, close, volume, open interest.

One of the issues with floating point numbers in general is that they are in base 10 (decimal) and we are storing them on base 2 (binary) computers.

Without going into further details, single precision floating point numbers are only accurate to around  7-8 significant figures.

If you want to read up why, you look up topics in Google such as floating point, single precision, range, figures.

It can be significantly improved by using double precision floating point numbers but this would probably be a big change throughout MetaStock (data files, internal representation etc.) plus would also doulble the size of data on disk too.

Cheers,
Richard.

---

www.premiumdata.net

Richard Dale:

The data files are based upon the original CompuTrac format which uses a single precision floating point number...

Well there's a name I haven't heard for quite sometime. I thought the CompuTrac format was abandoned around Y2K. Quite honestly, I was so focused on the irregularity having something to do with the importing or a bug in the software when data was converted that "precision" hadn't even crossed my mind and likely wouldn't have at this early stage.

Thanks

Iris

Here's the history of what is now colloquially known as the "MetaStock file format" as far as I know it:

1.  The Computrac analysis package developed an open file format in the 1977 on the Apple II computer.  In 1987, Computrac released a Microsoft DOS-based version.  for securities data.  It was originally written in Microsoft Basic.  Indeed, this is the reason that there is a need to convert any data fields in the "Microsoft Basic floating point format" into the now standardised IEEE floating point number format. The Computrac software was licensed to Strategem Software, and the Computrac company was sold to Telerate (Bridge) since sold in part to Dow Jones and to Reuters.  I'm not sure which company(ies) today own/have rights to the code though.
2.  Equis used the "Computrac format" in all versions of MetaStock up to and including v6 (1996)
3.  MetaStock version 6.5 released in 1997 enhanced the file format by increasing the number of securities in a folder from 255 to 2000.  Also, composite securities were added at this time.
4.  In 1999, MetaStock version 6.52 was released that provided a Y2K-compatible date field inside the data files.
5.  In 2002, MetaStock version 8 was released that increased the number of securities-per-folder to 6000.

Cheers,
Richard.

---

www.premiumdata.net