Comparative Analysis of Multi-Agents Competing for Trading Agents Competition

摘要

We present a comparative analysis for several multi-agents participating in Trading Agents Competition, Classic. The game is first partitioned into separate modules, for which distinct strategies may be developed. The strategies used are taken into consideration both individually and in relation with other agents, but also the game medium. Conclusions regarding possible improvements, better strategies and potential weaknesses are driven from each agent analysis. Alternatives like static market algorithms vs. dynamic market algorithms are considered in detail and advantages and disadvantages are discussed. Also a discussion of TAC market and possibilities for the stochastic system approximation with a deterministic one is realized.