The expected deviation curves shown in the figures are larger than the standard deviation for a random distribution. This keeps them from encumbering the space within which the cumulative mean varies. The figures show a sequence of means for a series of recalculations of the binomial samples that represent the counting of peas or plants that Mendel carried out. In all the figures except f), it would be possible to show mean counts that were nearly exact by stopping the count early. In a) would your subconscious tell you to stop at 15 or 50 samples, or would you persist to a large sample count, and move from the exact value for credibility? In b), after 100 samples the mean persists at near the exact value. How many samples would you count? You face the same dilemma in c), d), and e). It will be hard for you to over rule your subconscious and try for an error sufficient to make it appear that you are not fudging the results. Figure f) does make an honest man out of you with no effort on your part. Given the character of actual cumulative mean curves I conclude that the chi square test of the data of Mendel is simply too unsophisticated to be relevant.