[dgt20 (OP)]

Any suggestions to improve radioactivity dice experiment?

[alancalverd]

Use lots more dice (the curve fit depends on the square root of the number of dice, so you need to scale up from 100 to 10,000 to make a significant difference). Or fewer real nuclei (with less than 10,000 nuclei you can really see the underlying randomness of the process).

Better still, accept that the continuous decay equation in the textbook is the idealised fit to an infinite number of events measured with infinitesimal accuracy - real measurements are a lot rougher than the smooth curve we publish, and subject to errors due to coincidence, dead time, self-absorption and sample discretisation. The statistical interpolation methods used to get from experimental numbers to published halflives are all in the advanced textbooks but would be regarded as a bit of a distraction in a classroom demonstration. I have worked with physicists, chemists, electronic engineers and mathematicians who spend their entire working lives smoothing out the curve!