Abstract: The keys of pre-registration for solving the reproducibility crisis are to admit bias, register bias and reproduce bias, as well as the construction of a whole data-chain from idea to publication. The system of pre-registration could control the following four bias in the research process: Firstly, registration of pilot data provide the data basis for preservation and disclosure of confirmation bias. Pilot data should be published, so as to improving data reproducibility of experiment design. Secondly, incremental discoveries should be unloaded to control hindsight bias, as well as provide database for reproduce hindsight findings. Thirdly, second peer-review should also evaluate the implementation degree of registered data-analysis methods, in order to control selective reporting bias such as floor effects, ceiling effects, p-value effect etc. Fourthly, pre-registration must publish null results to control publication bias.