Disparity in materiel is always good. Games with lots of high end entities with high end equipment running around tends to get old fast.
I just don't think taking past player performance into account is going to work out well. It's hard to measure (what if you're good at specific things, such as fights vs fighter-heavy fleets? what if you're bad at piloting, but your strategic decisions make you do well in large battles?).
Indeed it is hard to measure, but there is an implicit problem when you claim that auto-resolve is going to behave like how battles are fought. It's often the case that it will not resolve in the way that battles are fought, or in the way that the player will resolve the battle if he or she fought it manually. Unless the program can use the ships in a similar meaningful way that the player uses them, it's not going to behave just as battles are fought. For example, take mount and blade for instance, I often use auto-resolve and while it does give me favorable results in most battles, there are some where there should be no loses at all for example a 500 man patry fighting a 3 person party.
You're also creating an environment which punishes experimentation. Say you want to try out a new loadout - but you know that if it doesn't work well, you'll be punished for it in auto-resolve from here on out, until you erase that somehow (presumably, by "grinding" some wins to level it back up). That's powerful disincentive. Never mind that you'd need to have some visibility into the inner workings of that - i.e., what the system thinks of you.
I think this will only be true if battles are far and in-between large amounts of time. One measly battle, no matter how big, is still one battle in a huge collection of them. Unless the player deliberately keeps doing it, there's really going to be no significant mark on the overall player percentage. Besides, whether or not it encourages the player to fight his or her battles or experiment really depends on the skill of the player, for someone who is incredibly poor at the game or for someone who isn't very consistent, auto-resolve might prove to be so much more successful that they might never fight a manual battle themselves. If the player advantage quantifier for the calculator disintegrates over time, the player will have to fight manually to keep it, better yet, maybe it's a pool that the player can use to weigh in on insignificant battles to ensure a more one sided battle as if the player played it.