Phaser is a software package for phasing macromolecular crystal structures using molecular replacement (MR) and experimental phasing methods, including single-wavelength anomalous dispersion (SAD). It employs maximum likelihood and multivariate statistics for improved phasing algorithms. The software is designed for high automation, allowing it to be called from Python or traditional CCP4 input. Phaser is a platform for developing and releasing improved phasing methods to the crystallographic community. Phaser's algorithms are based on maximum likelihood probability theory and multivariate statistics, offering better performance than traditional methods in distinguishing correct solutions from noise. For MR, it uses novel maximum likelihood algorithms for rotation and translation functions, while for SAD, it accounts for correlations between F+ and F- to provide better phases. The software also includes anisotropy correction, fast rotation and translation functions, and refinement target functions. The maximum likelihood functions in Phaser are derived from univariate structure-factor distributions, but multivariate statistics are used when correlations between structure factors are significant. The SAD likelihood function accounts for correlations between F+ and F- and includes terms for acentric and centric reflections. The software also incorporates ensembling, using structurally aligned models from the PDB to generate calculated structure factors. Phaser's normal-mode analysis uses eigenvalue decomposition of the Hessian matrix to identify the lowest frequency normal modes, which can help generate models closer to the target structure when there is a conformational change. The software is designed for high-throughput structural genomics, improving automation and algorithm quality to enhance structure solution efficiency. Phaser's algorithms are more sensitive to partial structure information and provide better discrimination of correct orientations and translations compared to traditional methods.Phaser is a software package for phasing macromolecular crystal structures using molecular replacement (MR) and experimental phasing methods, including single-wavelength anomalous dispersion (SAD). It employs maximum likelihood and multivariate statistics for improved phasing algorithms. The software is designed for high automation, allowing it to be called from Python or traditional CCP4 input. Phaser is a platform for developing and releasing improved phasing methods to the crystallographic community. Phaser's algorithms are based on maximum likelihood probability theory and multivariate statistics, offering better performance than traditional methods in distinguishing correct solutions from noise. For MR, it uses novel maximum likelihood algorithms for rotation and translation functions, while for SAD, it accounts for correlations between F+ and F- to provide better phases. The software also includes anisotropy correction, fast rotation and translation functions, and refinement target functions. The maximum likelihood functions in Phaser are derived from univariate structure-factor distributions, but multivariate statistics are used when correlations between structure factors are significant. The SAD likelihood function accounts for correlations between F+ and F- and includes terms for acentric and centric reflections. The software also incorporates ensembling, using structurally aligned models from the PDB to generate calculated structure factors. Phaser's normal-mode analysis uses eigenvalue decomposition of the Hessian matrix to identify the lowest frequency normal modes, which can help generate models closer to the target structure when there is a conformational change. The software is designed for high-throughput structural genomics, improving automation and algorithm quality to enhance structure solution efficiency. Phaser's algorithms are more sensitive to partial structure information and provide better discrimination of correct orientations and translations compared to traditional methods.