Sorry to burst any bubbles but… Pathologic’s data actually shows tempering is likely either weighted or broken.
Using Pathologic’s results, both the G-test and chi square test of goodness of fit find the results are not consistent with expected results, meaning the distribution of tempering is likely NOT even. Sorry.
Caveat: it’s been a more than a few years since I took Business Statistics in college so I’m happy to be corrected. I tried to read and confirm which test to use and actually used two different methods. The calculated probability that Pathologic’s results [818, 738, 696, 748] are consistent with an even distribution [750, 750, 750, 750] is only .0172 (G-test) and .0165 (chi square). Anyone can use Excel to calculate the chi square using the chisq.test() function.