If you are interested in the reason why this is the case, it comes down to small sample size. You figured that out to some degree, but I'll help you with the reasons why it was too small. First, if you are only looking at the first deck out, just set the pen to 1/6. Any hands you deal beyond that point, but then ignore, count towards your rounds simulated but not toward your results. Second, with something like hi lo, your true count "bins" will contain 5 or 6 different possible running counts that give the same TC, so you'll rack up hands for each bin 5 or 6 times faster than you will when you are tracking by running count. These two factors together mean you will need something like 25 or so times more hands than you would to get comparable data on a hi lo game that only deals 1 deck out of 6. The first of those two factors is in your control with sim settings, the second really is not.