Something I’ve noticed in Dragon Age 2 that I didn’t in the original is that rogues are way more MAD than any other class… that is, they’re Multiple Attribute Dependent. Warriors want Strength for their attacks, Constitution for more hit points, and, as a far tertiary, Willpower for more energy. Mages want Magic for their attacks, Willpower for more mana, and, as a far tertiary, Constitution for more hit points. Rogues want Dexterity for their attacks, Cunning for their damage and unique skill (opening locks), Willpower for more energy, and Constitution for more hit points. Building either of the first two classes means keeping the first two scores high (with which one is higher dependent on build) and sometimes tossing a point into the tertiary score. Building a Rogue means a constant fight to keep Dexterity high enough to wield the best weapons, Cunning high enough to open locks at that stage of the game, Willpower high enough to have enough energy to use your skills, and Constitution high enough to not die all the time.
This is a classic problem in game design, and D&D in particular gets a lot of flack for classes that are highly MAD. Classic examples in 3rd edition are the Paladin and Monk, which both have only two of four ability scores that aren’t in some way essential to various class features. So Paladins tend to be stuck as clumsy idiots, and Monks as boorish idiots, by virtue of what stats they have to dump to keep the others high. Meanwhile, other fighting classes don’t really require high mental scores at all, and can generally deprioritize at least one of the physicals, making them way easier to make effective with comparable ability scores.
The problem has a pretty simple source: once you’ve created a meaningful array of attributes and abilities, you’re essentially written into a corner by simulation logic, despite what it does to game balance. If you have a Constitution attribute, it should be involved in any system where a healthy guy would logically do better than a sickly one. If you have a Dexterity attribute, everything done better by a quick and agile guy should use it. If you tie any of them into important derived scores, you’ll have to work very hard to justify why different classes derive that score a different way.
One way to avoid the problem is the way used in Changeling: the Dreaming, Fading Suns, and Mage: the Awakening: have all the abilities within one power group use a different (almost random) combination of scores that are at least superficially related to that particular trick. This effectively makes every character MAD, limiting them to either being universally mediocre at their special powers or good at some tricks while terrible at others within the same specialty. This does make the system a bit more balanced (unless a player finds a powerful combo of tricks using a small pool of abilities), but makes your system less elegant: everything is a special case that players will have to look up to remember anyway.
The other way is to continue the trend that both tabletop and video games have been pursuing for the last decade or more: package all of your game rules into the final expression of your system (the fun), and allow players to decide for themselves what that means about their character in context. That is, for example, do you really need to use a magic or intelligence attribute as a prereq for sorcerous abilities, or can you just let the player buy the abilities and roleplay being a powerful savant/towering intellectual? Keeping players from cherry picking can just as easily be accomplished by putting good abilities several steps down a feat progression as by tying powers to different attributes.
I’m a simulationist at heart, and I enjoy the idea of having numbers on the sheet saying, “my guy is this strong, this smart, this fast, etc.” But in a lot of modern games (particularly video games) those numbers aren’t really meaningful from a system point of view: all the really important verbs (the ones that your players will want to use most often) are hidden behind special abilities. We’ve been gradually moving from an era where players could try most anything they could think of and then roll against their attributes to one in which interesting options are prepackaged as abilities to give players ideas and an instant niche (but that, de facto, prevent a player from doing that thing without the ability). Kewl powerz have gone from being rare seasoning to being the dominant flavor of many game engines.
So if you’re making a game where special abilities make up the majority of the things your players might want to do (particularly a video game where thinking outside of the pre-scripted options isn’t even possible), consider very strongly if you still have any reason to include attributes and abilities. You may find that you’re holding onto them purely out of nostalgia, and that your game will be far easier to balance without them.