A conditional 2025 BAT test of 4U 0114+65 misses most sine pulses with half the mean rate as their semi-amplitude

We test pulse detection in public Swift/BAT data for 4U 0114+65. We use single-pointing data from 2015, 2021 and 2025. Each year has four independent pulse searches from 8800 to 9800 seconds. Daily offsets remove slow changes. Random signs on whole-day residual vectors define the primary noise law. The 2025 real statistic does not pass its 99% threshold (p=0.3293). At a sine semi-amplitude equal to half the observed 2025 mean rate, only 21.1% of 5000 trials pass this threshold. The conditional 95% Monte Carlo interval is 20.0–22.2%. The 90% detection level is 0.76 times the observed mean. Two-day and seven-day noise blocks give the same main result. The curves apply to sine pulses with fixed frequency and phase within each part of a year. They do not bound all physical pulse shapes or changes. Native satellite times lack Solar System and binary delay corrections. The package contains the exact data, code, noise draws, curves and figures.

4U 0114+65 · 2S 0114+650 · Swift BAT · X-ray pulsar · pulse detection · injection study · empirical noise