Boson sampling can effectively simulate the sampling problem based on quantum effects, which provides a new path to verify the exponential acceleration of quantum computation. In reality, it is difficult to prepare large-scale photons with indistinguishable properties, which greatly suppresses the computational complexity advantage of boson sampling. In order to study the influence of photon partial distinguishability and photon losses on the probabilities of boson sampling, the boson sampling process based on the Reck model with 6 photons and 10 modes is simulated. The experimental data indicate that with the decrease of the indistinguishability and the increase of photon loss probabilities, the error between the actual frequency of output combinations and the ideal one increases gradually, however the error variation speed decreases. When the partially distinguishable photon number is changed from 0 to 1, the error variation range for boson sampling output combinations is the largest. The accuracy for boson sampling output combinations gradually decreases with the further increase of partially distinguishable photon number, but the reduction speed decreases. Our results demonstrate that boson sampling based on the Reck model is sensitive to photon indistinguishability, partially distinguishable photon numbers and photon losses in optical networks.