I love Andrew Stadel's Estimation 180 collection and sequence, for all the reasons why lots of people have been praising it. Estimation is an undervalued skill! Kids are terrible with units of measurement! It's super-accessible across multiple grade-levels! It's quick! It's fun! Etc!
One of the features is that Stadel always first asks for a guess that is too low and a guess that is too high, before asking for a the final estimate. I initially liked this because of how it increases accessibility for students and also trains them, in the long-term, to get more specific and accurate with their estimates and reasoning. Now I found a new reason why I like this practice: preparing kids for confidence intervals. Confidence intervals are just a more systematic way of making estimations, and really what the confidence interval is saying is "here is my range of guesses that are neither too high or too low." I imagine that when having kids share to high/low guesses for Estimation 180, there will sometimes be guesses that are actually correct, especially as students get better at their estimation and try to get "just a little bit" off in their too high/low values. That, in a lot of ways, is like a confidence interval! A 95% CI is saying that it is possible--5% of the time--that our interval doesn't capture the true population statistic (mean, proportion, etc). The too high/low guesses for whatever population statistic fall in those outer 2.5% tails. They're possibly correct, but it's highly unlikely. You'd be really surprised a value from those tails turned out to be the true population statistic.
Somewhere in here there has to be a lesson/activity where we collect estimations from a bunch of people and find out that the mean estimation is actually pretty close to the actual value. This is true for people guessing about the number of jelly beans in a jar, and so on. Can we use estimations to set up a confidence interval for the real number of jelly beans in a jar? Is that legitimate statistics?
