An extension of play against the random past strategy. Choosing the right experts on IBM forecasts

Abstract

In sequential plays with two players, the players have the opportunity to use information on opponents’ past moves in selecting a move for the current stage. Strategies for Player II are considered in our study, in particular, play against the random past (PRP) strategy. In this paper, PRP strategy will be reviewed and discussed. Hannan consistency of PRP strategy in term of regret (difference in average loss and an envelope loss) on k-extended Bayes envelope risk problem in matching binary bits game will be shown. The simulation of two-experts selection problem on real experts’ forecasting data of IBM share earnings confirms the consistency of PRP strategy.

Keywords:

• play against the random past
• sequential prediction
• Bayes envelope risk
• expert selection
• game theory

Acknowledgements

I appreciate Prof. Dennis Gilliland's help and support all along. Wharton Research Data Services (WRDS) was used in preparing the application simulation in this paper. This service and the data available thereon constitute valuable intellectual property and trade secrets of WRDS and/or its third-party suppliers.