The University of Essex Computer Science home Page

Bargain Tournament 3.2
Announcement of Results, February 2004

Bargaining Project at the Centre for Computational Finance and Economic Agents (CCFEA)


This is a bargaining tournament based on a simple bargaining model, where a buyer bargains with a seller (see details in call for participation). The seller has a cost and the buyer has a utility. Each of them has a time constraint to sell/buy. 

This is the second tournament of this type. The first tournament, Tournament 3.1 was held in 2001-2002. In that tournament, the players have no information about their opponents. In Tournament 3.2, the players are given the opponent's range of cost and utilities (but not the exact values).

Thanks to everyone who participated in the Bargaining Tournament. I have received some very successful programs. The tournament was run in early February 2004.


I have received 20 student submissions of sellers and 9 buyers in 2003-04. I have added 3 sellers and 2 buyers from Tournament 3.1 (2001-02), and one seller by Mathias Kern (tutor and PhD student) for testing. Two buyers with buyers (Giddings_b.plg and srlawr_b.plg) generated negative scores. Seven programs generated errors; they have not been included in the tournament.

The best buyer:

Christopher Musgrave’s buyer ccmusg_b.plg is by far the best buyer in the tournament. Not only does it beat all the other buyers produced this year, it beats all the buyers produced in Tournament 3.1 in 2001-02. The nearest to Musgrave’s buyer, in terms of performance, is Robert Bragg’s buyer rbragg_b.plg.

The best seller:

The picture is less clear on the seller side. Robert Stacey’s seller written for the previous tournament in 2001-02 (rpstac_selle.plg) was the winner when negative scores were allowed. This is surprising as Stacey’s program did not use range of costs and utilities.

Seller mkern_s.plg by Mathias Kern (tutor and PhD student) was the winner when all players with negative scores were removed. Kern’s program did use information about costs and utilities. Stacey’s program scored heavily against players that accept losses.

With players that generated negative scores removed, competition amongst the CC283 students in 2003-04, lemeri_s.plg by Luis Enrique Merinero Herranz was the highest scoring seller, closely followed by ccmusg_s.plg by Christopher Musgrave (who also wrote the winning buyer).


Why does a strategy win? One can win by:

  1. taking advantage of its opponents;
  2. working together with its opponents in order to realize profits; or
  3. doing both.
Analysis shows that Christopher Musgrave's ccmusg_b.plg was successful in doing both (a) and (b). On average, it obtained 44% of the potential profits in each game. (Note that this does not mean that its opponents obtained 56% of the potential profits because sometimes there were no deals. On average, only 66% of all the potential profits were realized by the players together.) This is the highest score among all buyers. Together with its opponents, ccmusg_b.plg realized 74% of the potential profits on average. This is again the highest score except those buyers that were willing the make a loss.

Robert Stacey’s rpstac_selle.plg succeeded by scoring heavily on buyers who are willing to make a loss. Although it drives a hard bargain, it managed to realize 84% of the potential profits with its opponents. This is explained by the fact that it managed to compromise at the right moment. How does it fair with ccmusg_b.plg?  It scored only 27% of the potential profit, as opposed to 67% by ccmusg_b.plg. This is perfectly within our expectation as the latter uses more information than the former.

The success of Mathias Kern's mkern_s.plg and Luis Enrique Merinero Herranz's lemeri_s.plg will only be obvious when one removes from the spreadsheet the buyers that generated negative scores .  They both succeeded by scoring high against the buyers (28% -- they only differ behind decimals; one has to modify the spreadsheet to see this figure).

There is a lot more to be said about these results, but I'll stop here. Anyone interested should download the Excel spreadsheet for analysis of the results.

Summary of results:
Winner Buyer: ccmusg_b.plg by Christopher Musgrave
Runner up: rbragg_b.plg by Robert Bragg
Winner Seller: lemeri_s.plg by Luis Enrique Merinero Herranz
Runner up: ccmusg_s.plg by Christopher Musgrave
Download: Excel Spreadsheet

Christopher Musgrave and Luis Enrique Merinero Herranz will each receive a certificate and a bottle of wine from me. Many congratulations!


Professor Edward Tsang
CC283 Course Supervisor and Bargaining Tournament 3.2 Organizer
Department of Computer Science and
Centre for Computational Finance & Economic Agents
University of Essex
Tel: +44 1206 872774
Fax: +44 1206 872788

Hit Counter