this is still a bit rough - please consider it a beta release.
In the last lesson we saw how 'real' shills differ from imaginary shills, and how to identify an 'imaginary' shill.
So, now it's time to gather the bid histories we'll be needing.
Make a new directory to dump all the files you're going to need in.
Head back over to http://pages.ebay.com/search/items/search_seller.html and search for every item the seller has sold in the last month. Make sure you tick "Yes" for all completed items.
Now a bit of finger exercise - work through the list right-clicking every auction that has closed (indicated with an asterisk,) and choose 'Open in New Window'. Save the auction ('File > Save as' or [CTRL]'F' + 'A'), highlight and copy ([CTRL]'C') the auction number, right-click on 'bid history' and 'Save Target As' each of these, substituting the auction number for the random filename supplied by eBay's server (typically something unhelpful like "CABW174XS") with [CTRL'V'). This way you'll know which one goes with each auction.
When you've finished this (might be 10 auctions, might be 100 - you really do want them all if possible, or as many as are possible. If it's a dauntingly large quantity you might save time by not bothering to save the bid lists from buy-it-now or auctions with only one bid as these aren't vital,) go along to http://pages.ebay.com/search/items/search_bidder.html and repeat the process for all the suspected shill bidders associated with this seller's auctions. Again, save all the auctions and their associated bid histories that have closed (indicated with an asterisk.) All the files can go into the same directory as Windows 'search' will easily help you filter the information you need.
HANDY TIP: if you want to save your fingers and/or there are a ridiculous number of auctions to gather, you can use web 'spidering' software like the excellent shareware WebZipto grab all these automatically from the 'Search seller/buyer' pages - a godsend for investigating particularly prolific sellers with hundreds of items to track!
|
I shall base this section on an example drawn from my own recent investigations, so these figures are genuinely representative and easily to hand. They stem from my second campaign against the 'Michigan gang' (June-03). The bids and comments are accurate and comprise the totality of the sample data recorded for analysis of these two known shills.
They may seem rather obvious when tabulated, but that's often the way. Shill-bidders are in it for the money - they're almost without exception too greedy to be cunning, and a short time spent digging up the data on them will expose them. They rely on the naivity of fellow eBayers to line their pockets, but are too lazy to make much effort to hide - figuring that if they're suspended they can always come back under a different member-name.
And by the time they're getting regularly kicked off before their auctions can close it's too late for them to hide. They've revealed so much incriminating data that you can uncover and bust them with ever-increasing ease :-)
As a matter of interest, goldengirl04 scored 19, time-to-listen scored 15 and and vlaminck1922 scored 14 on the Significance Calculator presented in Part 1. Of course, they were all the same person, so call it an average of 17ish. If you've got this far chasing a suspect who scored much lower than that and who isn't as obvious once the data's tabulated, give up - you're probably chasing a shadow based on an erroneous hunch. It is sadly all too easy to convince yourself of a suspect's guilt, which is why it is so useful to lay out all the facts in a manner designed to reveal the inevitable patterns and connections which will always betray the true shill. Knowing how to recognise a blind alley is probably the most useful skill you can hone when shill-busting!
Open up the directory you've saved the files in, and a text editor (or spreadsheet if you prefer,) to tabulate some relevant data.
One by one open up the bid history files (identifiable as being named with just the auction number.)
Bear in mind that you can only see the maximum bid entered where the bidder has lost the auction. (If they win, the amount shown is only one bid-step over the next highest bidder!)
So record the following information in table form in your spreadsheet or text editor, and looking for pattterns as you go, record any interesting observations in the 'Comments' column, and especially make a note of any auctions sniped - ie: any auctions bid on within, say 60 seconds of the End of Auction (though if they're habitually manually sniping a couple of minutes before the end, this might also indicate it's not a shill, so you may wish to alter this figure):
time-to-listen Member since: Tuesday, Dec 03, 2002 Location: United States
Bids indexed via eBay search-bidder form:
| SELLER | CATEGORY | AUCTION # | BIDS | COMMENTS | |
| birchlawn | banknote | 3026431214 | $775.88 | ||
| ronswrite | watch | 2637135149 | $9.88 | ||
| downingtowncoins | banknote | 3026394520 | $826.88 | ||
| chuck4737 | doll | 2930881185 | $51.50 | (high-bidder, reserve not met) | |
| jazzyjeffrey | painting | 3522976324 | $127.88 | (goldengirl04 3x bids subsequently) | |
| vlaminck1922 | coin | 2636639672 | $412.00 | (only after goldengirl04 made 3x bids) | |
| dollco2 | doll | 2931919014 | $76.88/$101.98 | ||
| visakay | sketch | 3524298418 | $110.00 | (exception 2 the rule?) | |
| double_r_coins | banknote | 3029301717 | $319.00 | (buy-it-now) | |
| bestbeasleys | doll | 2931861645 | $101.88 | ||
| currencyseller | banknote | 3028761640 | $555.88 | ||
| brightstar47 | coin | 3029191150 | $197.98 | (exception 2 the rule?) | |
| oozentoo@aol.com | coin | 3029123065 | $1,055.99/$1,688.88 | ! | |
| bigdad_1@ix.netcom.com | banknote | 3029316929 | $328.88 | ||
| bigdad_1@ix.netcom.com | banknote | 3029318157 | $388.88 | ||
| birchlawn | banknote | 3030083392 | $327.88/$515.88 | ||
| bovona | painting | 3527698096 | $110.88 | ||
| aleko | sculpture | 2837418751 | $36.88 | ||
| fineartnet | painting | 3527776246 | $400.00 | (exception 2 the rule?) | |
| vlaminck1922 | painting | 3528010853 | $336.00 | (only after goldengirl04 bid $98.88!) | |
| zermelo | coin | 3030696648 | $475.88 | ||
| bluesg@aol.com | coin | 3031450345 | $25.00 | (but after chcarew bid $17.88 & $21.88!) |
time-to-listen: unfinished auctions bid on (all subject to administrative cancellation):
| SELLER | CATEGORY | AUCTION # | BIDS | COMMENTS | |
| stevenskf | statue | 2540363495 | $250.00 | subsequently mark@psci.net bid $1300.88 | |
| greatsouthernantiques | coin | 3032470672 | $9.95 | ||
| treashunt@aol.com | banknote | 3031447592 | $9.99 | subsequently wscheid bid $88.88 | |
| coins@astampcoin.com | banknote | 3032628273 | |||
| firstpick | painting | 3532205477 |
goldengirl04Member since: Sunday, Feb 02, 2003 Location: United States
Bids indexed via eBay search-bidder form:
| SELLER | CATEGORY | AUCTION # | BIDS | COMMENTS | |
| spoiledrotten26 | coin | 3028000004 | $426.88 | ||
| franticize | coin | 2636639672 | $248.88/$333.33/404.87 | ||
| odona | painting | 3524269176 | $49.95 | (only bidder) | |
| denbear2 | sketch | 3526088365 | $28.50 | (won) | |
| vlaminck1922 | painting | 3528010853 | $98.88/$482.11 | ||
| jazzyjeffrey | painting | 3522976324 | $128.99/$130.38/$130.38 | (won) (also bid on by time-to-listen) | |
| time-to-listen | coin | 3032162178 | $279.00 | ||
| vlaminck1922 | coin | 2636639672 | $248.88/$333.33/$404.87 | (also bid on by time-to-listen) |
and via feedback left for her (all won, therefore no max bid shown):
| SELLER | CATEGORY | AUCTION # | BIDS | COMMENTS | |
| time-to-listen | coin? | [private] | ? | (first feedback - highly complimentary!) | |
| cambrittcoin | coin | 3017527070 | $51.00 | ||
| alajose | coin | 3021665235 | $102.50 | ||
| alajose | coin | 3021662955 | $61.00 | ||
| adrydock@aol.com | painting | 3514782614 | $99.95 | (only bidder) | |
| trmcho | painting | 3519460855 | $36.00 | (bid twice) |
Closer examination of the auctions offered by time-to-listen and vlaminck1922 (see example auctions) revealed similarities between these auctions in particular and these sellers in general (and that the New England shillings offered in auctions 3012940692 & 3032162178 were the same coin,) which led me to believe that they were not just shilling for eachother, but were the same person:
|
And you also may notice that there's a pattern emerging which anyone can see fairly plainly now the data's tabulated, and ploughing through the pages one by one, the fact that (excluding auctions won or bought with 'buy-it-now',) one or other shill ALWAYS placed one, and only one (usually the first,) bid ending in 88c, and if subsequently bidding, varied the amount while usually 'doubling up' the cents (33,99). And in the cases of auctions #3522976324 and #2636639672, the pattern was actually spread across both bidders. ie: goldengirl04 put in the 88c bid, and time-to-listen followed up with "00". Such pattern 'spreads' (in addition to happening to share identical patterns,) are a surefire giveaway!
| NB: This is not unusual - everyone has their own little quirks - some people round bids off to "01" to try to 'pip' bidders bidding whole $s. Others may double $s and cents ($51.51, $14.14, $30.30, etc.) The commonest is to bid in round dollars, but some bidders will round eg: under $10 to the nearest dime and everything over $100 to the nearest $10. I favour "03" and "53" to pip the bidders who choose "02" and "52", who in turn are trying to pip the "01" bidders...
The only way to avoid such patterns emerging in your bidding would be to use a random number generator to calculate your bids and choose which auctions to bid on - not a strategy to be recommended! finding patterns not just shared by bidders, but split across their bidding is certainly a useful stick to beat them with! |
wscheid bidding activity: 3026505225 banknote 32.08/38.88 3027052718 banknote 128.88 3026509003 banknote 25.08 (seaweb also bid 102.88) 3027055405 banknote 25.08 3027038091 banknote 21.88 3026988938 banknote 21.08 3026517739 banknote 25.88 3027067207 banknote 22.88 3026521118 banknote 58.88 3026526429 banknote 26.08 3027163075 banknote 16.08/20.88/25.88 3027163515 banknote 30.88 (efiasconaro@aol.com also bid 48.88) 3027164407 banknote 26.08 (efiasconaro@aol.com also bid 37.88) 3027165280 banknote 26.08/28.46 (won) 3027165616 banknote 28.88 3027166490 banknote 15.08 etc... Seems to only ever bids on banknotes (over 300 in the last 30 days) Member since Thursday, Sep 30, 1999 Location: United States Pretty clearly the pattern is consistently a roughly 50:50 choice of "08" or "88" cents, and thus differs. No feedback interaction with any of the known shills
chcarew bidding activity: 3027882808 coin 112.88/118.88/122.88/127.50 (won) 3027563668 coin 121.88/128.88/135.88 3028197872 coin 5.50 (won) 3028198331 coin 14.88/16.88/19.88 3028198905 coin 88.88/91.88 3028227644 coin 16.88/18.88/20.88/22.50 (won) 3028238755 coin 14.88 3028769445 coin 8.88/9.99 3028405967 coin 29.00 (won) 3028409119 coin 5.79 (won) 3028411605 coin 9.88 3029214035 coin 88.88 3028416939 coin 10.88 3028896076 coin 19.99 (won - only bidder) 3028653517 coin 19.88/21.88 3028653926 coin 41.88/43.88/45.88 3028686059 coin 44.88 etc... Seems to only bid on coins (65 in the last 30 days) Member since: Monday, Sep 09, 2002 Location: United States The pattern differs consistently in that almost every bid ends in 88 No feedback interaction with any of the known shills
No auctions were sniped. Inspection of feedback profiles and ID History at http://cgi3.ebay.com/aw-cgi/eBayISAPI.dll?MemberSearchShow showed that:
|
In fact because this was the second investigation of the time-to-listen puppet account (which had been suspended for shill-bidding 2 months previously following a previous couple of shill-bidder reports by yours truly!) I was able to confirm that:
|
It transpires (to the best of by knowledge,) that goldengirl04 was an imaginary shill of time-to-listen, who may have been a 'real' shill via both accounts for the more newly arrived (but clearly criminally connected) vlaminck1922. Close examination of the grammatical style and spelling errors in common or not so (as well as comparison of email header data and other advanced techniques,) as well as the dates they joined/location/feedback comparisons/etc. support this assessment.
In practical terms, predictions of maximum bids ending in "88" cents in auctions not yet ended, or where the bidder won the auction, and the identities of bidders in several private auctions certainly seem to have swiftly convinced eBay - we can't see such things, but they can with a couple of mouse-clicks!
time-to-listen was suspended for her second time. Bang! 2 strikes and GONE!
the newly arrived vlaminck1922 acccount was suspended for its first time.
All their auctions (with current high-bids totalling over $1000!) and bids were cancelled.
goldengirl04 wasn't suspended - I guess because she's never auctioned anything (eBay always seems reluctant to suspend members who only ever bid,) but no matter - she is now a busted flush. I shall be watching her very carefully indeed now! As will I vlaminck1922 if that one pops up again, and as I was indeed already watching time-to-listen after the first suspension.
And while compiling this lesson I discovered a gem of an auction dating from my first campaign - and how I wish I'd seen it at the time! It would most likely have completed the hat-trick and got goldengirl04 suspended as well!
How dumb do you have to be to be? time-to-listen was suspended for shill-bidding via goldengirl04 (that was when there were only the two of them - before vlaminck1922 appeared,) and three hours after her bid for this painting was subject to an "Administrative Cancellation", she reentered it using the goldengirl04 shill that got her suspended in the first place!!! I guess she really, really wanted it :-)
As I mentioned previously, sadly eBay always seem very reluctant to suspend shills who are not engaged in selling as well as bidding, but this might have done the trick!
Maybe it still would, but as she's a busted-flush, I'll leave her in place as a monument to the sheer stupidity and greed of the Michigan gang. (And if she ever starts bidding again, I'll be onto them again in no time! :-)
The next lesson (shortly to be released,) will complete this Introductory level course in Shill-busting with 'How to write reports to eBay that get results', and covers all the information you need for dealing effectively with 90%+ of cases you're likely to encounter.
Later Intermediate and Advanced level courses in statistical analysis and advanced data-gathering techniques will follow when I have time to complete them, and may prove useful for aspiring 'extreme shill-busters'.
One intermediate technique which has many uses (and one you might like to start now if you are interested in continuing this course to the more advanced levels,) including better detection of shill-bidding operations, is to archive as much as possible in the fields you're interested in. I archive dozens of auctions a day - everything remotely interesting to me (just save the auctions and bid histories as described above - if the auctions haven't finished yet, prefix each with a "!" and every week or so go through these files to reload and save the final 'auction ended' data.) Thus I can search for bidding activities stretching back over a year, rather than the 1 month searchable and 3 months archived on eBay's servers.
For easy archiving of whole eBay categories, check out the shareware WebZip. Run it once a week in your favourite categories and you'll have a monster of an archive in no time, easily searchable with Windows Search! (Just make sure you've got acres of harddisk space! :)
|
For many advanced statistical techniques covered in later lessons (chi-square test, Poisson's distribution analysis, Fisher's exact probability test, etc.) minimum sample sizes often exceeding the number of auctions extractable from the eBay search-bidder page are required. Otherwise you'll have to wait gathering data until you have enough to go on (and it's also invaluable for analysing competitor bidders in your own field of interest, which throws up hitherto undetected shill-activity surprisingly often!)
| PREVIOUS LESSON | NEXT LESSON |