[A note to the winners: Please email me your address so that I can send you your babka! - DavidNYC]
So, I'm lazy and took my sweet time in writing this up. In my defense, Martha Coakley happened, and I chose to distract myself by working on more datasets for Dave's App. I won't steal Dave's thunder, and I'll leave it to him to announce any news on his time.
Nevertheless, here are the results from the Great Swing State Project Redistricting Contest, Round 1.
Some specifics first. I received 18 entries that were judgable; if I was missing anything from you, I tried my best to reach out and get everything I needed to judge your entry.
Of the 18 entries, with David's generosity, I selected two winners: abgin, and MattTX2.
The first, I hope, is of no surprise - abgin's 28-0 map, I think, made everyone's jaw drop in its brilliance/hideousness/all-around-awesomeness. His/her entry was simply too good not to reward with babkaness!
For the other 17 entries, there was only one 26-2 plan, from silver spring. Three of you tried 28-0 plans, and the remaining 13 of you gave us 27-1 plans.
I tried to come up with some objective measures to help me consider them. Here are the two I considered:
1. To determine the amount of gerrymandering, I created a "county fragmentation index," which indicates the number of "unnecessary" splits.
2. I also created a "swing index," to measure quantitatively the improvements given to each district. I used a "safe line" concept (ala my New York State Senate district diary). Since reasonable people can disagree on what an appropriate safeline is, I evaluated each plan on safe lines at 1% increments from 50% Obama to 60% Obama.
Without further ado, I chose MattTX2 as the second winner. I thought he executed the screwing of Peter King in the most precise way, and as the first entrant, showed that this could be done (something I was skeptical of; I believed that the population simply wasn't in place for that to happen.) His map also produced the best objective improvements, improving the 27 Democratic districts by an aggregate of 5,600 basis points in the 57-60% ranges, the best among the 27-1 plans. The performance, notably, didn't drop off in the 54-56% range either. Matt's map, compactness wise, wasn't a horrendous gerrymander either; with 26 more county fragments than minimally necessary, this was solidly in the middle of the pack (the median, in fact.)
Here are improvement and county fragment scores for every entry.
I would be remiss in judging, though, if I didn't mention some other notable entries:
- Abgin obviously had the loosest defintion of "contiguity", but I think the award for loosest definition of "water contiguity" goes to andgarden.
- duffman gets the award for most compact map, creating just 11 more county fragments than necessary.
- bschak made the best attempts for population equality, acheiving a total deviation of only 3,902.
- AdmiralNaismith played population equality the loosest, racking up a total deviation of more than 113,000.
- Alibguy had the best 28-0 plan, yielding the best improvements from 52-60%.
In conclusion, thanks to everyone who participated, to Dave for developing this wonderful time-sink (I could never get GUIs down when programming), and to DavidNYC for giving me this opportunity to judge the contest, and for spreading the glory that is Green's chocolate Babka! Congratulations again to abgin and MattTX2!
Update: I should better explain what the swing index is: the swing index is, in total, a slightly-adjusted measurement of the Democratic percentage improvement. The index is the sum, across all Democratic districts, of swings toward Obama in basis points.
When I say slightly adjusted, for example, I want to punish someone for unpacking Louise Slaughter's district. So you have the Obama percentage before (Ob1), and the Obama percentage after (Ob2), and the "Safe line" percentage (SL).
So you have four situations:
1. Ob1 > SL, and Ob2 > SL: the district was above the safeline before, and the new district is above the safeline. The contribution to the swing index from the district is 0.
2. Ob1 < SL, and Ob2 > SL: the district used to be below the safeline, the new district was boosted above the safeline. The contribution to the swing index from the district is Ob2-Ob1, in basis points.
3. Ob1 > SL, and OB2 < SL: the old district was above the safeline, the new district was weakened. The contribution to the swing index from the district is SL-Ob2, in basis points. I did this so that you're not being punished for an astronomic drop for a district that was much much too safe before, like Louise Slaughters.
4. Ob1 < SL, and OB2 < SL: the old and new districts are both in marginal territory; the new district could be stronger or weaker than the old. The contribution to the swing index from the district is Ob2-Ob1 in basis points. This can be positive or negative.
So I totaled up the numbers from each district, yielding the swing index that you see.
Also, the ones highlighted in green are the highest two indexes for a given safe line.
Lastly, the CFI is the "County Fragment Index" - the number of excess county fragments created over what is minimally necessary for 28 districts with perfect population equality.