Swoopo Dataset

The dataset we used in our paper:

John Byers, Michael Mitzenmacher, Georgios Zervas,
"Information Asymmetries in Pay-Per-Bid Auctions: How Swoopo Makes Bank"
[PDF (Extended arXiv version)] [PDF (Shorter EC version)]

is publicly available from this page.

Download

IMPORTANT NOTE: If you downloaded the dataset on or before March 21, 2010 then the columns bid_user_secs_added and bid_butler_secs_added in the file traces.tsv were reported wrong. Please download a fresh copy below. Thanks to Jeff Nanney for pointing out this bug.

swoopo-dataset.tar.gz (27 MB)

Feedback

If you use our dataset we'd love to know about it. Please drop us a line and let us know if you found it useful and ended up using it in one of your projects.

README

These datasets were collected and used for the following paper:

John Byers, Michael Mitzenmacher & Georgios Zervas:
"Information Asymmetries in Pay-Per-Bid Auctions: How Swoopo Makes Bank".

For questions regarding this dataset please contact Georgios Zervas.

OUTCOMES DATASET (outcomes.tsv)
-------------------------------
The outcomes dataset is provided in TSV format and contains the following columns:

- auction_id: A unique numerical id for the auction

- product_id: A unique product id

- item: A text string describing the product

- desc: More information about the product

- retail: The stated retail value of the item, in dollars

- price: The price the auction reached, in dollars

- finalprice: The price charged to the winner in dollars

- bidincrement: The price increment of a bid, in cents

- bidfee: The cost incurred to make a bid, in cents

- winner: The winner's username

- placedbids: The number of paid bids placed by the winner

- freebids: The number of free bids place by the winner

- endtime_str: The auction's end time

- flg_click_only: A binary flag indicating a ``NailBiter'' auction

- flg_beginnerauction: A binary flag indicating a beginner auction

- flg_fixedprice: A binary flag indicating a fixed-price auction

- flg_endprice: A binary flag indicating a 100\%-off auction


TRACES DATASET (traces.tsv)
---------------------------
The traces dataset is provided in TSV format and contains the following columns:

- auction_id: the auction id

- bid_time: the date and time of the bid

- bid_ct: the value of the countdown clock at the time the bid was reported to
  us, in seconds

- bid_number: the number of the bid (1 for the first bid, 2 for the second one,
  and so on)

- bid_user: the username of the bidder

- bid_butler: 1 for BidButler bids, 0 otherwise

- bid_cp: the price of the item after the bid was placed

- bid_user_secs_added: the number of seconds added to the countdown clock as a
  result non-BidButler bids in this bid group

- bid_butler_secs_added: the number of seconds added to the countdown clock as
  a result BidButler bids in this bid group

- bid_infered: 0 for the final bid in a bid group, 1 otherwise (see note below)

- bid_group: bids that were reported as part of the same group will have the
  same group number

- bid_final: 1 for the winning (ie, last) bid, 0 otherwise

The other parameters of the auction can be looked up in outcomes.tsv using the
auction_id field.

Note on bid groups: Occassionaly, between successive probes, more than one bids
would have occured and would be reported together. We call this a bid group.
All bids in a bid group have been ascribed the same timestamp.