January 2017
This map shows the high estimate for the size of each crowd based on this spreadsheet which is being collected by Prof Jeremy Pressman and others.
Code/Tech Notes
After downloading the spreadsheet as a csv(ok, after unsuccessfully trying to use google's sheet "feeds" to load it dynamically as json), I used jupyter notebook for data processing (see wmcrowds.ipynb) and then mapbox-gl-js to actually create the map.