Human behavior depends on evaluating uncertainty in decisions—a metacognitive ability. This ability can be quantified in various ways. Typical measures are meta-sensitivity (meta-d’) and metacognitive efficiency (M-Ratio). These measures are derived from a Signal Detection Theory (SDT) framework and make (truncated) normal noise assumptions. They were, together with 15 other measures, recently assessed by Rahnev (2025, PB&R) using real world data (rather than just simulations like previous studies). Results on validity, precision, susceptibility to nuisance variables, and reliability led Rahnev to conclude that M-Ratio outperformed the other 16 measures. Nevertheless, it did not fulfill all desiderata. In fact, M-Ratio shows a slight tendency to overcorrect for a nuisance variable and, as do most other measures, exhibits poor reliability. Thus, there is room for improvement. A recently proposed, conceptually intriguing family of measures that were not included in this assessment are measures based on Information Theory: meta-I, RMI, meta-U, and others (Meyen, 2025, OpenMind). These measures do not assume specific noise distributions allowing them to be applied in a wider variety of contexts and tasks. But how do these information-theoretic measures compare against the established measures such as M-ratio in real data? We will close this gap by comparing the established SDT-based measures with these newly proposed information-theoretic measures. We will do so by using Rahnev’s assessment battery and demonstrate on which particular data these measures disagree. This will give guidance for future research on when to employ which of these measures.