I think it's a cool idea, but your main paragraph about problems might be too important/pervasive to ignore, especially if your goal is to have an accurate measurement and to avoid trolling/epeen stroking. We all know once those numbers hit, people are going to look at the top and bottom and post accordingly.

I know from personal experience if you fight a few hard-fought battles and lose more clones than you can replenish in a day, then it's possible to lose a district despite winning essentially every battle. Or you can win a few, lose one, and then have so few clones left on a district it becomes basically impossible to defend. Both situations will lower your ELO and raise theirs despite the fact that you beat the opponent more often than you lost to them.

Another problem I see is how would you know who was attacking a district if they never take it? For example if you see Corp A under attack constantly, but they never lose the district, wouldn't it be impossible to calculate an ELO? Maybe Corp B, who is really good, is hitting them and they're successfully defending, which would increase Corp A's ELO a lot. On the other hand it could be their alt corp attacking and they're just practicing. How could you calculate the adjustments to their ELO?

I like the idea a lot and it would be nice to see implemented, but unfortunately I don't think you have sufficient access to information to make this accurate enough to be of any real worth.