I use this project every year in my AP Statistics class to demystify the often tricky topic of sampling distributions! A great transition between probability and inference. Students analyze 2 bags of M&Ms, weighing the bags using scales I borrow from the science department and calculating the proportion of each color. They then build sampling distribution models around M&Ms published information about both average weights and color distributions, and see where their bags fall. I also have the student post their data to a class list - I do this digitally by making a Google Sheet but you can also do it manually - and then redo the analysis using our aggregate data. This allows students to see the power of a larger sample size.
I tell the students to come to class with two regular bags of M&Ms. However, you could adjust the project to be mini bags if you are purchasing them for your kids. I also reference this data again when I teach them Chi-Square, as the tables of color distributions are perfect for Chi-Square goodness of fit analysis. For my kids, this has been a 2 - 2.5 day (45 minutes periods) project.
I have had great success with this project at the AP level and even have used it with my regular Statistics classes. I hope you and your kids love it too!
Connect with me on Instagram @jamieHSmath and you can see pics of this project in action - and message me with any questions you have about it :)