Help File

As of November 2008, ebay.com and ebay.de no longer display the winner's login names.

Now, therefore, I have no means of correlating login names with code names.

I will leave untouched the data currently available but no further work can be done.



Since 2007 American and German eBay has displayed bidders names in code with only 
the winner of an auction having their full login name displayed; the losers have 
their name displayed in code.

German ebay (ebay.de) uses the same system but British ebay (ebay.co.uk) uses a more
secure system for which this sort of analysis is not possible.

This program gives a translation of some of the codes to login names using data from
ebay.com and ebay.de.

Basically I create a table of winners (with their login names and feedback count) and a table of 
losers (with their codenames and feedback count) and attempt to marry the two tables. The fact
that these two tables are in a state of flux means that certainty is not guaranteed!

The codes are given in alphabetical order.

It should be realised that eBay converts login names to a UNIQUE 5 character identifier.

However, only the first and last digits are displayed; the middle 3 characters are displayed as '*'.

Assuming that ebay is using only the characters a-z, 0-9, + a few specials in their display then 
there are only approximately 2,000 combinations possible. eBay has millions of bidders so that 
any displayed codename e.g. a***a will be associated with thousands of bidders.

Thus: "deniserose"  may be coded by ebay as "e123e" (but we will never know), and "arcaze" may be 
coded as "e456e" (and again we will never know).  Both of these will appear as "e***e". 
There will be several thousand other bidders each with the displayed codename e***e.

In these cases the feedback count (which is provided by eBay) may be used to distinguish 
between some of them.

So at this point I must insert a caveat.  With so much uncertainty and with feedback counts
changing comparatively rapidly - often on a day by day basis - when my program is only run
once a week, the output from this program should be treated with great caution.

The possibility for a partial solution to the problem is only present because of the comparatively 
small number of different bidders for slide rules.

To avoid excessive fairly useless data, only codenames which have a feedback greater than 13 (chosen
quite arbitarily) and can be associated with login names are displayed.

A line such as:

-***j	m-jlovell	135

is a clear indication that the codename of "m-jlovell" is quite likely to be "-***j".

The following types of line will also appear if the "All Known Code Names" option is selected.

A line such as:

g***b bwong1943 hideousgelfling 133 

is saying that one of these login names probably has the code name g***b

On further inspection of the table we may find:

i***l bwong1943 hideousgelfling 133 

That is the two login names bwong1943 and hideousgelfling are probably associated with the 
two code names g**b and i**l but at the moment I cannot distinguish between them because 
they both have a feedback of 133.

As their bidding develops and their feedback counts differ, identification becomes more
likely but not certain.

A line such as:

1***0 perniciousfrump unknown unknown 91 

is saying that perniciousfrump may have a code name of 1**0 but there are currently at least
two unknown bidders with the same feedback count.

The greater the feedback count, the greater the probability of success in correlating a
codename with a specific login name.

For a name to appear, the ebay bidder must have had at least one successful bid and one
unsuccessful bid in an auction for "slide rule", "sliderule" etc., since June 14th 2008.

If the optional search facility is chosen to look for code names, login names or feedback
counts then the Unique Code Names Only/All Known Code Names  radio button is ignored; a
search is made on the full table.

If only digits are inserted into the search string then it is assumed that you are looking
for a feedback count. This is treated differently to other searches in that an exact match
is not demanded of you.

I have found that I am interested in looking for a bidder with a feedback count I have just
observed on ebay. Now my table and its feedback counts may be up to a week behind a current 
auction. It is, therefore, quite possible that the feedback count in my table will be
slightly less than the one shown by ebay.

Therefore, when a feedback count 'n', say, is requested, I will also look for feedback counts
of n-1, n-2, and n-3.

The table will be updated weekly so that the feedback count should not be more
than one week out of date.