Recall that each of these 10 standard deviations was based on just 10 samples drawn from the full population, so significant fluctuations should be expected. BUT, the standard deviation which you calculated for all 100 samples of 10 flips is expected to estimate the population standard deviation much more reliably. Likewise, the mean of heads across all 100 samples (of 10 flips) should tend to approach 5 more reliably than any single sample. Comment on related trends observed in your data.