There is limited research on evaluating an agent's reputation as a recommender. A key challenge is that a recommender's reputation is affected by both the recommender's trustworthiness and the recommender's expertise, including the recommender's trust knowledge of others and the reliability of the recommender's trust evaluation models. In this paper, we give an ordered depth-first search with threshold (ODFST) algorithm to find the optimal referral chain. We then develop a Hidden Markov Model (HMM) based approach to measure an agent's reputation as a recommender. This approach models chained recommendation events as an HMM. The features of the trust model are: (1) no explicit requirement of chained recommendation reputations; (2) flexible recommendation network with presence of loops; and (3) integration of learning speed into trust evaluation reliability. The experimental results showed the convergence and reliability of the proposed trust model.