1: \begin{abstract}

2:   We propose a sequential optimizing betting strategy in the

3:   multi-dimensional bounded forecasting game in the framework of

4:   game-theoretic probability of Shafer and Vovk (2001). By studying

5:   the asymptotic behavior of its capital process, we prove a generalization

6:   of the strong law of large numbers, where the convergence rate of the sample

7:   mean vector depends on the growth rate of

8:   the quadratic variation process.  The

9:   growth rate of the quadratic variation process may be slower than

10:   the number of rounds or may even be zero.

11: % We analyze the structure of sequential optimizing betting strategy

12: %   in multi-dimensional bounded forecasting games and deduce asymptotic

13: %   behaviour of its capital process and the covergence rate of strong law of

14: %   large numbers, including the case that the quadratic variation process

15: %   grows slower than the number of rounds.  

16:   We also introduce an 

17:   information criterion for selecting efficient betting

18:   items. These results are then applied to multiple asset trading strategies

19:   in discrete-time and 

20:   continuous-time games. In the case of continuous-time game 

21:   we present a measure of the jaggedness of a

22:   vector-valued continuous process.

23: % , a mutiple type of Girsanov's   theorem. 

24:   Our results are examined by several numerical examples.

25: \end{abstract}

26: