Explain the higher estimated rating-noise σ0 in Antichess versus standard chess

Determine the causes of the observed higher estimated measurement-error standard deviation (σ0=67) for Glicko‑2 rating differences in Antichess compared to standard chess (σ0≈58) under the SIMEX calibration, investigating whether variant-specific rating behavior, dataset composition, or rating-system changes account for the increase.

Background

Under the same SIMEX framework, Antichess produces σ0=67, exceeding the ≈58 estimated for standard chess.

The authors explicitly note that the reason for Antichess’s higher σ0 is unclear, highlighting a need to understand variant-specific rating dynamics or methodological differences.