They don’t.

It’s almost as if you didn’t check!

It doesn’t seem like 1837 to me…

But that’s WolframAlpha. Let’s check with another reliable source, such as CODATA, an international organization that collects and stores scientific data.

(The CODATA page itself is visually distressing. This is why you won’t get it directly from CODATA!)

Huh — it looks like WolframAlpha might have got it right!

This indicates that the proton-masse ratio is somewhere in the area

mpme=1836.1526739+-0.0000002

A 69% certainty (I have rounded their errors a little), and lies within this region

mpme=1836.1526739+-0.0000006

Up to 99.9% Confidence

But…it could still be 1837, yeah? If all the measurements are wrong?

To determine if the ratio is exact 1837, you need to calculate how many standard deviations it is from the measured value. A rough estimate is Ns

=4×106

The majority of calculators I found couldn’t calculate a probability higher than Ns=10

I can only offer the best estimate of the probability that this ratio is 1837, based upon the CODATA values. It is located in the region spandata-mathml=’ 1837 10 – – – >P183710_

That’s….pretty small!

It isn’t 1837 …., I believe.

The actual mass ratio is (unsurprisingly) a non-integer, and the current accepted value of the ratio is given by:

mpme=1836.1526739+-0.0000002

Not 1837.

It’s close to 1836 for a start.

It is therefore difficult to explain why something isn’t true. This is due to the simple fact that it isn’t true.