Risk arbitrage, also called merger arbitrage, is an investment strategy that arbitrageur tries to lock the spread between offer price and post-announcement price of target company. If merger successes, the arbitrageur have opportunity to profit from the arbitrage spread. However, if the merger fails, the arbitrageur may incur a loss. This paper studies the performance of risk arbitrage portfolio and explore whether we can predict the outcomes of merger on announcement date. The empirical result indicates that average event return of arbitrage is 3.73% and excess event return is 2.74%. Arbitrage portfolio generates 1.5% abnormal return per month, and the volatility of arbitrage is less than market index. In addition, we find that risk arbitrage return in bear market is higher than that in bull market. We use three approaches including logistic regression, probit regression and exchange options to predict the probability of merger success. The two first models have total accuracies about 80%, which are higher than that of exchange options approach. The three models all have weak prediction performances on actually failed mergers. Besides, the relative sizes of target companies have negative effect on merger success. The ownerships of target companies hold by acquiring companies before announcement date and cash mergers can increase the probability of mergers success. The takeover premium has no significant effect on mergers success.