This paper considers a cell with many caching helpers, such as femto-base stations with memories for caching files. We formulate average bit error rate (BER) as a function of caching placement in a Rayleigh flat fading channel. In order to minimize the average BER, we propose an algorithm which finds a caching placement in greedy way and prove that the caching placement found by the proposed greedy algorithm is optimal. We also show that the proposed greedy algorithm has complexity independent of file library size. It is shown that the optimal caching placement is not just caching all different files; however, for a given environment, the optimal caching placement is to balance between channel diversity gain achieved by caching the same file in multiple helpers and caching diversity gain achieved by caching different files in helpers to enlarge a set of cached-file.