@mariosv: apart from the fact that your example is also good with errors to prove similarity between the formulas in question? …
would you mind updating your sample in a way that it doesn’t give the impression that both random formulas are erroneous? … your totals and subtotals in B29:U32 and X29:AQ32 reference to questionable areas, and thus the results seem to be mostly positive (V32 vs. V30 and AR32 vs. AR30), which they are not really …
(if was ‘on the way’ filing a harsh report about insufficient rand routines in calc, spotted the fail in the last second … ufff)
with adapted formulas i still see a slight overweight of ~2:1 in favor of positive results in V32 and AR32 (30 tests each with full-recalc), but that would be another issue, it’s definitely better than ‘black always wins’ with your formulas,