Preregistration: independent checks of WattMate's numbers on Apple silicon Macs

A preregistration of 3 tests of WattMate, a macOS menu bar app that shows how many watts and minutes of battery each app uses on Apple silicon MacBooks. The rules, thresholds and formulas were fixed on 29 October 2026, before any counted run. Results follow as the next version of this record by 15 December 2026. Every test runs on all 6 Macs: MacBook Pro 13″ M1, macOS 26.7.1 MacBook Pro 14″ M1 Pro, macOS 14.2.1 (the oldest macOS WattMate supports) MacBook Pro 14″ M1 Pro, macOS 27.0.1 MacBook Air 13″ M2, macOS 27.2 MacBook Pro 14″ M3 Max, macOS 27.0 MacBook Air 13″ M5, macOS 27.0.1 3 tests: A1. Does Apple’s built-in sensor at the power input agree with an independent USB-C meter (ChargerLAB POWER-Z KM003C)? Pass: within 2%. A FIBARO smart plug at the wall also measures the adapter’s efficiency, without a pass mark. A2. Known loads (1 busy core, 4 busy cores, a GPU kernel, an SSD write and read): does WattMate rank the right process first? Pass: 9 of 10 rounds. A3. Real apps (the ChatGPT app for macOS in front and hidden, video in Chrome, a C build): what share of the MacBook’s extra power does WattMate assign to the app? Published as a measurement, without a pass mark. The analysis scripts are committed by SHA-256 before each test’s first counted run. The author develops WattMate; this conflict of interest is stated in the text. Version 2 adds, under the rule stated in the text, the SHA-256 hashes of the A2 files and the sources of the A2 loads (loadgen.swift, gpu_burn.swift). The other A2 files are kept in a separate record with restricted access: 10.5281/zenodo.23765407. The rules are unchanged.