User blog comment:BunsenH/Entbrat Stats Phase 1, Report 2/@comment-27661991-20170716153502/@comment-24577221-20170716225008

[Urgh. One of the things I hate about the Wikia system is that if one takes a while to write up a good long comment or change, the wiki just burbles about a timeout and forgets everything one has written.]

Um. Thanks, but -- no offense intended -- I'd rather stick with "minions" that I can supervise personally. The difference in platform is also a point. I think it's not likely that the different versions of the game really have different mechanics, but we can't be certain. Also... for most purposes, it's annoying that Big Fish Games isn't maintaining MSM; a lot of nice features are missing. But for our research, it does simplify matters, since we don't have to worry about Breeding Bonanzas or anything else altering the breeding probabilities any more.

Please encourage your niece to continue her research. As you've seen, I'm not just trying to find out what the probabilities are; I'm using this stuff as an educational tool. A lot of people don't have a good grasp of how probabilities and statistics work, and the best way of solidly understanding this is to work with it oneself. Mind you, I'm being a bit hypocritical here. One of my middle-school science classes spent a full week on flipping pennies to see how this stuff worked. But I complained to my mother after the first day that I already knew all of that stuff, she talked with my teacher, he quizzed me, and I spent the rest of the week in his office reading a biology textbook.

But all of the stuff about expectations, and how small sample sizes can give different results from large sample sets, is important. When people hear about a survey or poll they often ignore the part about how the numbers are, say, "within 4%, 19 times out of 20". It's really important to remember that that means that the result won't be in that range 1 time in 20. I'm not surprised that her results are somewhat different from ours. You can take a look at the curves I pulled out of Abedshark's Daily Roll project, and see how rough the early ones are. Even the last one, with 279 data points, isn't a really good bell-shaped curve. And the first two sets of numbers in our Ghazt breeding, each with about 10,000 trials, seemed to show that a torch's effect was only about ¾% per torch. It wasn't until we got a broader range of data, with nearly 50,000 trials, that we could see that the real effect is just about exactly 1% per torch.