Explain the smaller estimated rating-noise σ0 in Atomic chess versus standard chess

Determine the causes of the observed smaller estimated measurement-error standard deviation (σ0=33) for Glicko‑2 rating differences in Atomic chess compared to standard chess (σ0≈58) under the SIMEX calibration, evaluating whether the discrepancy reflects variant-specific rating processes, temporal windows, rating-system changes, or other data-generation differences.

Background

Using SIMEX, the authors estimate σ0 for each variant. Atomic chess yields σ0=33—again considerably below the ≈58 estimated for standard chess.

They note uncertainty as to why Atomic’s σ0 is smaller and suggest that differences in rating computation or sampling windows may be implicated.