Runners up: Mertacor (Aristotle University of Thessaloniki) and CrocodileAgent (University of Zagreb)
For the second year in a row, a newcomer has won the tournament. For the second year in a row, a Greek team has won the tournament (a Greek team also came in second this year). After ten years of Power TAC, we may not have enough data to draw big conclusions about team composition, origin, and 'freshness', but we are given to wonder – what might be the value of first-time participation? Do incumbents think too narrowly? Do newcomers see opportunities we don’t?
After all, in 2012, a winning team showed us a flaw in our own market design. Call it our very own Enron-moment. But the strategy then of only using arbitrage and not direct trading reminded us that good policy design is a crucial element of healthy markets.
We’re big fans of collaboration and interdisciplinary learning, so we’re looking forward to hearing from competitors on their strategy in agent design. So get going on those papers! In the meantime…
2020 PASS workshop
Are you a computational IS researcher? Consider participating in or presenting at our online workshop! Policy, Awareness, Sustainability and Systems: Next Generation Research Challenges for Computation IS Research in Energy, Mobility and the Sustainable Economy at the Energy Economics Institute, University of Cologne.
The workshop will take place on Wednesday, December 2 – 3, 3 PM – 6 PM (CET), virtually only. Interested presenters should submit an extended abstract (1 to 6 pages, excluding references) per e-mail to ketter(at)wiso.uni-koeln.de, including all names and e-mail addresses. Please submit your presentation to Wolf by November 30th.
The tournament is complete; see the results here.
In the release of Power TAC version 1.7.0 for the 2020 competition, customers became a bit skeptical about unreasonably high prices in regulation rates. Details are in Section 4.1.1 of the updated 2020 specification. In other aspects, the specification remained unchanged, so agents from the previous two years should have operated without problems. Game data from 2018 and 2019 was used for training. Some customer populations may have been enlarged or reconfigured, so overall demand and detailed demand profiles may have differed from earlier competitions. As in previous competitions, competitors were required to have two copies of their agents running and available to connect during the entire duration of the qualifying and final rounds.
What is Power TAC?
If you’re new to the Power TAC community, the tournament is a competitive simulation of future retail electric power markets, in which retail "brokers" buy and sell power in both wholesale and retail markets. The wholesale market is an abstraction of typical day-ahead markets in North America and Europe, and the retail market is a tariff market, in which customers are able to choose among tariff contract offerings from the competing brokers. Customers are models of household, business, electric vehicle and institutional users of electric power, as well as small-scale producers of power that own solar arrays or small wind turbines.
We suggest cross-discipline teams if possible, with strong developers and team members with experience or interest in economics and trading, machine learning, and/or energy domain knowledge.
The 2020 tournament finals will be held on November 2 – 13. If you’d like more information on participation, please contact us.
Scroll down to see tournament history, winning teams and partners.
The fundamental starting point for broker development strategy is to look at the behaviour and performance of other brokers from past tournaments. You can also download existing brokers and run them against your own agents. The version for the 2019 tournament will be 1.6.0 or later; if you are preparing for the 2019 competition you can also run a development snapshot.
After you put a team together, read the background on Power TAC in these three research papers:
Power TAC: A competitive economic simulation of the smart grid
A Multiagent Competitive Gaming Platform to Address Societal Challenges
and then follow the instructions for installing the Power TAC server and building a broker agent:
This project was created with the cooperation of the following affiliations:
Dorothy.Grandia@ewi.uni-koeln.de