[gumby]

Redistricting commissions should recruit go players.

There can be an amusing side-effect to aggressive gerrymandering. Let's say a state has a small majority of party A, but districts that are strongly partisan (let's say 80%) for A or B. You only need 50% to win your district -- that extra 30% is "wasted" and would be more valuable if those voters were pushed into one of B's districts.

Which is one of the things partisan redistricters do. Now they have more districts with a comfortable 55% margin and an assured lock on more districts, right? Umm, well, now they've made it easier for the other side to capture districts too...and it happens.

There's no magic way to avoid bizarre results (check out the wikipedia pages on the districting, and on voting methods, for some eye-opening problems), but something close to algorithmic probably ends up the least controversial and most comprehensible in the long run.

[Retric]

It's not just about party, an unusually disliked candidate can tank most districts. In the end it's a very complex optimization problem with benefits in the short term measured vs the risk of losing redistricting power in 10 years etc. Even a slight bias can let you shift resources to more competitive races etc.

PS: IMO it's really a scale problem. If districts give a ~3% advantage that's huge in aggregate, but races still feel competitive. However, we are currently in a situation where nobody bothers to run in many races and that does not promote democracy.

[mathfan2]

> something close to algorithmic probably ends up the least controversial and most comprehensible in the long run

One suggestion in this direction is the shortest split line algorithm: http://www.rangevoting.org/GerryExamples.html

(The OP also uses split lines, but they are optimized for maximum partisan effect rather than minimum length.)