We study the sensitivity of future medium baseline reactor antineutrino experiments on the neutrino mass hierarchy. By using the standard χ 2 analysis, we find that the sensitivity depends strongly on the baseline length L and the energy resolution (δE/E) 2 = a/ E/MeV 2 + b 2 , where a and b parameterize the statistical and systematic uncertainties, respectively. The optimal length is found to be L ∼ 40 − 55 km, where a slightly shorter L in the range is preferred for poorer energy resolution. The running time needed to determine the mass hierarchy also depends strongly on the energy resolution; for a 5 kton detector (with 12% weight fraction of free proton) placed at L ∼ 50 km away from a 20 GW th reactor, 3σ determination needs 14 years of running with a = 3% and b = 0.5%, which can be reduced to 5 years if a = 2% and b = 0.5%. On the other hand, the experiment can measure the mixing parameters accurately, achieving δ sin 2 2θ 12 ∼ 4 × 10 −3 , δ(m 2 2 − m 2 1 ) ∼ 0.03 × 10 −5 eV 2 , and δ|m 2 3 − m 2 1 | ∼ 0.007 × 10 −3 eV 2 , in 5 years, almost independently of the energy resolution for a < 3% and b < 1%. In order to compare our simple (∆χ 2 ) min results with those obtained by simulating many experiments, we develop an efficient method to estimate the uncertainty of (∆χ 2 ) min , and the probability for determining the right mass hierarchy by an experiment is presented as a function of the mean (∆χ 2 )