The 2020 offset rule has lower mean absolute error than the origin rule for 15 later PSR J0537−6910 waits from published timing fits

We compare two fixed 2020 rules for the time to the next glitch of PSR J0537−6910. The test uses 15 later completed NICER waits from public timing tables. It uses the original size for glitch 8 and later published sizes for glitches 9 through 22. The offset rule, 23 + 5.08x days, has a mean absolute error of 19.33 days. The origin rule, 6.4x days, has an error of 25.31 days, where x is the step in microhertz. The reduction is 5.97 days, or 23.6%. We vary the epochs within the tested published ranges. For every allowed assignment, a conservative bound puts the mean error reduction between 3.462 and 10.271 days. The bound remains positive when each size also varies by its quoted one-sigma error. The offset band contains 8 of 15 printed outcomes; the origin band contains 9. The offset band has a lower mean interval score. Five input checks retain the error ranking. The study releases source-cell records, event scores, code, figures, and an offline reproduction method. The sizes come mostly from later timing fits. This is a retrospective test, not a record of forecasts issued before each event. It does not establish a physical cause for the glitches or calibrated future coverage.

PSR J0537-6910 · pulsar glitches · NICER · prediction errors · timing data · interval score · reproducible benchmark