1: \begin{abstract}

2:   We study capital process behavior in the fair-coin and

3:   biased-coin games in the framework of the game-theoretic probability

4:   of Shafer and Vovk (2001).  We show that if Skeptic uses a Bayesian

5:   strategy with a beta prior, the capital process is lucidly 

6:   expressed in terms of the past average of Reality's moves.  From

7:   this it is proved that the Skeptic's Bayesian strategy weakly forces

8:   the strong law of large numbers (SLLN) with the convergence rate of

9:   $O(\sqrt{\log n/n})$ and if Reality violates SLLN then the

10:   exponential growth rate of the capital process is very accurately

11:   described in terms of the Kullback divergence between the average of

12:   Reality's moves when she violates SLLN and the average when she

13:   observes SLLN.  We also investigate optimality properties

14:   associated with Bayesian strategy.

15: %   including the choice of the prior

16: %   for the Skeptic.

17: \end{abstract}

18: