The theoretical probability P(H)

Which provides a better estimate of the theoretical probability P(H) for the unfair coin: an empirical probability using 30 flips or 1000 flips? Why do you think so?