The Bahtinov mask doesn't lie. But if you must rely on chromatic aberration for focusing, here are the physics:
The index of refraction of a real material is wavelength dependent. So the photons going through it are slowed down by different amounts depending on their color (blue more than red), resulting in their beams being bent by different angles (think of white light split into its constituent colors through a prism). This makes the focus point of a lens wavelength dependent. For white light through a lens onto an RGB sensor, it looks like this:
Since the sensor intersects the light cones on a single plane, the "focus" point is where the center wavelength—green—is brought to a point (technically, minimum spot size, defined by the diffraction limit); the blue light cone has gone past its focus point and expanded, while the red light cone has yet to reach its focus point.
For incoming light from a white star, it looks like this on the sensor:
Whether or not this is detectable depends on several factors: the brightness of the target relative to its surroundings (high contrast is worse), the focal ratio of the lens (faster focal ratios have more CA), the size of the pixels relative to the size of the CA ring (from circles of confusion—e.g., a full-frame camera's larger pixels will be larger than the CA, whereas an APS-C sensor will show CA from the same lens), and the heroics of exotic glass type and lens combinations implemented by the manufacturer to reduce CA (basically, none, 2-color/achromatic, or 3 color/apochromatic).
If you're averse to using a Bahtinov mask, as an alternative to CA focusing I'd recommend stopping down to f/4 (or higher) and using the aperture preview button in 10x Live View on a nearby bright star to focus. That way, you can try to make the smallest point possible vice a magenta halo from a potentially non-white star. Try using some reading glasses (+1 to +2)—in low light, I've found this to help me see the tiny dot more clearly up close on the rear screen of my T3i.
Edited by BQ Octantis, 26 February 2020 - 03:49 AM.