Randomness deficiency measures how atypical data is under a model: roughly, it is the gap between the outcome’s expected complexity and its actual complexity. Low deficiency suggests the data fits the model; high deficiency marks it as an awkward misfit. In algorithmic statistics, it helps distinguish structure from noise and assess whether a concise model offers a good explanation. section 8.2.1
A black swan
A black swan has high deficiency under the model “all swans are white,” but low deficiency under a model that allows black and white swans. section 8.2.1
