How does adding identical values to all objects affect classification metrics?

Adding the same value to all objects (for example, shifting all features by a constant) usually does not affect classification metrics if the model and metrics are invariant to such transformations.

Why:

• Many algorithms (e.g., logistic regression, SVM with a linear kernel) and metrics (accuracy, recall, F1) depend on the relative differences between objects, not on the absolute feature values.
• Shifting all features by the same constant does not change the order or separability of classes.

However, if the model is sensitive to absolute values (e.g., decision trees or neural networks without normalization), adding the same value can affect training and, consequently, the metrics.

Overall, with proper normalization and standardization of data, adding a constant to all objects will not change the quality of classification.