In macromolecular X-ray crystallography, refinement R values measure the agreement between observed and calculated data. However, Rmerge values, which assess data quality by comparing multiple measurements of a reflection, are not suitable for determining the high-resolution limit. The authors introduce CC*, a statistic that estimates the correlation between observed data and the true signal, providing a reliable guide for data quality and model refinement. CC* allows direct comparison of model and data quality on the same scale, revealing when data quality limits model improvement. Accurate protein structures guide drug development and nanotechnology. Despite decades of progress, the selection of resolution cutoffs remains controversial. The authors propose CC* as a statistical guide for determining the resolution limit that optimizes model quality. In X-ray crystallography, measured data are reflection intensities, which yield structure factor amplitudes. The standard R value measures agreement between observed and calculated data, while Rmerge measures the spread of multiple intensity measurements. Rmerge must be adjusted by a factor of sqrt(n/(n-1)) to be independent of multiplicity. High-resolution cutoffs are typically set when Rmerge or Rmeas exceeds ~0.6 or when the signal-to-noise ratio drops below ~2.0. However, this approach is too conservative, as shown by an example dataset (EXP) with weaker intensities than the original data. Refinements using EXP data showed that adding more data improved the model. CC* was derived as a correlation coefficient between two halves of the data, showing strong correlation at low resolution and weak correlation at high resolution. CC* provides a reliable measure of data quality and model refinement, allowing direct comparison of model and data quality on the same scale. The authors conclude that CC* is a robust, statistically informative quantity useful for defining the high-resolution cutoff in crystallography.In macromolecular X-ray crystallography, refinement R values measure the agreement between observed and calculated data. However, Rmerge values, which assess data quality by comparing multiple measurements of a reflection, are not suitable for determining the high-resolution limit. The authors introduce CC*, a statistic that estimates the correlation between observed data and the true signal, providing a reliable guide for data quality and model refinement. CC* allows direct comparison of model and data quality on the same scale, revealing when data quality limits model improvement. Accurate protein structures guide drug development and nanotechnology. Despite decades of progress, the selection of resolution cutoffs remains controversial. The authors propose CC* as a statistical guide for determining the resolution limit that optimizes model quality. In X-ray crystallography, measured data are reflection intensities, which yield structure factor amplitudes. The standard R value measures agreement between observed and calculated data, while Rmerge measures the spread of multiple intensity measurements. Rmerge must be adjusted by a factor of sqrt(n/(n-1)) to be independent of multiplicity. High-resolution cutoffs are typically set when Rmerge or Rmeas exceeds ~0.6 or when the signal-to-noise ratio drops below ~2.0. However, this approach is too conservative, as shown by an example dataset (EXP) with weaker intensities than the original data. Refinements using EXP data showed that adding more data improved the model. CC* was derived as a correlation coefficient between two halves of the data, showing strong correlation at low resolution and weak correlation at high resolution. CC* provides a reliable measure of data quality and model refinement, allowing direct comparison of model and data quality on the same scale. The authors conclude that CC* is a robust, statistically informative quantity useful for defining the high-resolution cutoff in crystallography.