In 2016, I generated maps for the shortest splitline algorithm based on the 2010 census. I posted them on the election science forum.
I figured it was worth updating it for the 2020 census. Conveniently, they barely changed the file format (or at least they also produced the legacy files). It only took me 2 evenings to update it.
The program generates tiles so that the maps can be viewed in a browser.
It uses the actual block boundaries, so the districts end up with the (correct) jagged edges.
These are the maps for the 2010 census.
These are the results for the 2020 census.
You can see slight differences, especially in states that ended up with different seat counts. Also, the may be new (or old) bugs due to the quick update.
The algorithm performs a post-processing step after the districts are decided.
Any census block that is on the boundary of a district is eligible to be moved. It tries random allocations of those blocks between districts. (New boundary blocks are not included).
This moves the districts towards equal population. Most end up with 1 or 0 differences.Nevada: 302 New_York: 28 California: 17 Florida: 8 Texas: 4 North_Carolina: 3 Illinois: 2 Maryland: 2 Massachusetts: 2 New_Jersey: 2 Ohio: 2 Virginia: 2 Washington: 2
I am not sure what happened with Nevada. There was probably a large block on the boundary. 3 of the 4 districts are the same population.