As it stands now, casualties are calculated as a percentage of units sent or units defending. This means that the more units you send, all things considered, the more you lose (and viceversa). This means that there is very little incentive to oversending offense, except so that your elites are out and they don't get killed defending.

At the same time, attacking is a fairly straightforward process. You pick a target, you gather intel, you sent ≈104.4% of what the defense is and you attack. Sometimes there are spells or ops involved beforehand; sometimes you send your armies for a shorter time to catch up with another army, or for longer, so you can sleep. There is some strategy involved, in picking targets and whatnot, but for the most part, battles are meant to take land. If you want to reduce someone's army, you pretty much have to chain them – sometimes you get lucky, you catch them with armies home, and you kill some elites, but this is rarely devastating, and if anything its effects are mostly due to armies getting locked in due to overpopulation.

My proposal, thus, is very simple: oversending offense kills more enemy defenders. While I think a linear formula is simpler (you send 50% more and you kill 50% more) it would be better if it were slightly skewed (say, you send 50% more and you kill 60% more) – this way, there's one more element of strategy to ponder when attacking:
do I oversend troops so that I can kill more of the enemy's defense? Is that offense better allotted elsewhere, or staying at home and defending?

And also:

Are you better off sending four attacks and getting more land, or sending one big army and killing more troops?

This would also incentivize trying to catch people with armies home instead of just sending armies out the minute they get in. The benefit of being able to kill lots of elites with a well-timed shot would probably be tempting, wouldn't it?
It wouldn't affect gains, of course. And as it stands now, this would also mean more offensive casualties, which I think should be reduced somewhat in this scenario (say, killing 7% of the offense you would need to successfully attack instead of 7% of your total army, or some average in between) – but this last part is debatable, and probably something to tinker with in order to balance this proposal.