Search engine marketing provided by search engines enable companies to promote their products to internet users based on their queries is now a major online advertising channel. In most search-based advertising services, advertisers could have dozens of keywords for the same product or service, and in most instances, the reward of each keyword is unknown. Identifying the most profitable keywords becomes challenge for advertisers. In this research, we develop a new model named POKER based on multi-armed bandit problem to help solving this keyword choosing problem, we compare POKER with three currently frequently-used keyword selecting strategies by simulation. It turns out that POKER's aggregate performance overmatches other strategies.