A 30-day Einstein time fits public KMT-2025-BLG-1883 data better than 0.50 day for a single lens

We test the listed 0.50-day Einstein time of KMT-2025-BLG-1883 with six public KMTNet I-band alert light curves. We fit a static single lens to 1489 selected points within 10 days of the listed peak. Each stream has a separate source flux and baseline. Errors are scaled from the outer-season scatter. A uniform circular source gives a squared-residual score of 3736.36 at a fixed Einstein time of 0.50 day and 3004.45 at 30 days. The difference is 731.91. Thirty days is a test value, not a measured best duration. The reference score changes by only 1.40 from 30 to 300 days as the fitted source becomes fainter. The preference holds with a point source, a robust loss, a linear baseline, and either observing field alone. In a diagnostic chosen after exploration, the 30-day model also predicts 174 withheld Chile wing points better. Removing Chile reverses the remaining score preference. Chile alone covers the peak, so this coverage limit is material. An independent disk integral, broad optimizer starts, and an injected short event check the computation. The 30-day fit has a score of 2.04 per nominal degree of freedom. Correlated errors and residual misfit prevent a calibrated probability claim. The event was selected after exploratory fits, so no population rate follows. This result shows that the listed half-day value is not a secure duration measurement under the tested model. It does not establish a new planet or a lens mass. We preserve the input data, provenance, code, figures, row-level predictions, and a complete reproduction procedure.

microlensing · KMT-2025-BLG-1883 · KMTNet · Einstein time · source flux · duration degeneracy · public photometry · reproducibility