I've heard people defending the Armored Core series on the grounds that it's merely challenging, complex, and unforgiving. The thing is, that's not the nub of the problem. Call of Duty 4 is pretty unforgiving; you play it wrong and you are not proceeding one foot from your pathetic little bolthole behind a wooden fence in some godforsaken desert wasteland. Dark Cloud 2 is complex; the one time I stopped playing for a while and came back a few months later, I was so bewildered by the state of all the photos and weapons and Spheda and so forth that I had to start over from scratch. As for challenging, well, I can crank up any edition of Civilization to maximum (or set any RTS to "toddler") and watch it crush my pitiful bases with ease. Combine those three qualities and you'd have one heck of a hardcore game... but I don't think that's enough to make Armored Core.
The issue is simply that AC is not going to tell you how to play it, and I don't mean that in the positive sense of "tell you how to win" but more in the negative sense of "tell you what all those buttons do." And I honestly do not think that's good game design in any way. I don't mind at all the fact that you can choose between a dozen different leg joints and each one will affect the mech's speed, power, weight, dexterity, intelligence, and wisdom, and then you can install a better CPU and hack it to further modify those parameters. Seriously, that's super great. But the interface for managing all that stuff is disastrous. It's a gigantic spreadsheet, essentially, with no tutorial or instructions, no sense of scale for what a particular number signifies vs. another number, and the help function merely tells you which button to press to exit the spreadsheet. Then once in-game, it's not immediately clear how all those parameters are affecting the mech's performance, because combat is extremely fast-paced with virtually no onscreen feedback about your robot's performance and thus there's little opportunity to notice that your turning speed has changed by three percent before some offscreen enemy has blown your legs clean off and rendered the issue moot.
I'm reminded a bit of an old PS2 game called Warship Gunner, where you had the opportunity to assemble your own battleship and then take it into a fight. As a fan of Star Blazers since childhood that idea filled me with delight. Unfortunately, the interface for building your battleship was grotesquely clumsy, and there was no way to test out your creation short of starting the next mission and hoping for the best. (I ended up making a monstrous gun platform that was capable of about two knots of speed. It pulverized the enemy positions from halfway across the map without ever taking any damage and then took a half an hour just to drive to the mission exit point.) In AC4 you can at least test your mech out in-between missions, but the interface gives me a serious sense of deja vu otherwise.
Complex is cool! I want complex -- I'm a little full up on games that don't think I can handle a few variables! But if you're going to be complex there are some corollaries you must accept. You have to spend extra effort on your user interface. You have to be clear on what happens when the player tweaks variables, as opposed to simply being a black box. You have to be open and generous about allowing super-fast iteration. And the variables have to have a deep and fundamental effect on the game. If you don't follow up on at least most of these qualities, you're going to make a game that drives potential players away, not because it's hard, but simply because it is being deliberately obtuse.
Sigh, I dunno. I really want to like this game. I may seek out some FAQs to hopefully explain how I should actually play it (this hugely increased my enjoyment of Resident Evil 4, that's for sure.) But at present these are my thoughts on it.
I may also try to accept my own challenge and think about how I'd design an interface for a supercomplex build-your-own-mecha game. That could be a fun exercise.