We've got this amazing supply of great and interesting RPG engines out now, which I think the broad internet community (just people getting together to DIY stuff) and crowdfunding (lots of these systems are kickstarted) sort of came together to support.
That said, I think that most systems authors need to start with a paragraph or two about "why this RPG engine is necessary, instead of making it work as a plugin to GURPS / Fate / d20 / Apocalypse world." There's a lot of good reasons to do this, but there's also a lot of cruddy systems with half-baked rules that are ripe for rules-lawyering and just bad implementation details (looking at you, FFG Star Wars). Writing a well-crafted, minimal-bug RPG engine is similar to writing a front end framework or something along those lines.
I mostly agree, as a long-time FUDGE fan. And recently it's been interesting to see how much AW has captured the imagination of the indie community. It just really seems to work, and we're seeing it pop-up in so many other places. I've been absolutely adoring Night Witches this year, for example.
I slightly disagree, only inasmuch as I like to see exploration and innovation. Sometimes you're right it really doesn't work, But setting specific mechanics are important, sometimes essential, if very difficult to get right. Call of Cthulhu never worked in it's d20 incarnation, imho. Sometimes, I think, it is better to try and fail, than to take the safer route.
Yeah, I'm also in favor of experimentation and a rich ecosystem of ideas. Worst case is one of those heavily-contextual rpg engines that only lives in one system, such as Rotted Capes. It's an okay enough system (but mathematically flawed in some ways), and a pretty neat idea (zombie and b-list superhero mashup), but it would be considerably better if it were built on a different game engine. Then the author could spend more time building out the world and storyline, instead of making lists of superpowers and stuff.
Even then, though, I often back stuff on Kickstarter at the PDF level knowing full well what I'll really be doing is mining the product for ideas, setting, and story rather than mechanics.
Then again, I'm probably just like one of those people who says, "why would you write code in anything but LISP?"
The first thing that needs to change is the core resolution mechanic. The dice system is just flat out silly, and looks like a transparent attempt to make money by selling custom dice. The die resolution system could easily and with very little change to probability of success be changed to something like a d6 die pool system with 5 and 6 as successes. I know the advantage / success split would need to be fixed, but that mechanic is confusing enough (empirically, my GM and several other players all of whom are experienced with RPGs but new to the system were all befuddled by what to do with a failure with two advantages). Just discard that and use a tested mechanic like extra successes - roll N dice based on skill, a 5 or 6 is a hit, you need N hits to succeed, based on difficulty, and overage successes increase the margin thereof. You can fiddle that to allow 1s to be extra failing etc, but that is a lot cleaner.
The confusing symbol math needed to decode a roll (there are what, six major symbols, some of which count as other symbols, others which cancel in pairs) is annoying enough that I imagine every dev who has played this game has written their own die roller.
While the story and fluff behind character generation is pretty good, the three talent tree thing feels again like a cheap attempt to appeal to video game semantics, but ends up introducing balance issues (it is extremely worthwhile to dip a tree to gain cheap skill updates and the low tree upgrades, given cost scaling).
Character generation also makes it way too easy to make a really cruddy character by accident. In this case, "cruddy" is defined as someone who is significantly less likely to succeed at resolution rolls in a variety of situations. In particular, because attributes play such a huge role in skills and are hard to increase outside of chargen, it is waaay better to invest initial points directly in those.
This is exacerbated by the fact that while it feels nice to add specialization to a skill, it does not actually increase the odds of success by all that much. You are a better pilot with three green dice than with a green and a yellow. This problem combines the funky and hard to understand statistics of the resolution system with permanent mistakes that can be made during character generation, where a novice player can spend a lot of their resources making a character who is 15% better of a pilot than anyone else, but is terrible at everything else, which usually results in unsatisfying gameplay.
There are a lot of other bugs in the system - the weapon / armor add-on system is ripe for exploitation if the game is combat-heavy and the GM is foolish enough to allow people to easily purchase custom stuff. There are so many moving parts and ways to minmax, it is hard not to. I mean, an under-barrel flame thrower is a reasonable option here. FFS, FFG.
Ultimately, I enjoy star wars RPGs, but that is despite the mechanics, rather than because of them.
That said, I think that most systems authors need to start with a paragraph or two about "why this RPG engine is necessary, instead of making it work as a plugin to GURPS / Fate / d20 / Apocalypse world." There's a lot of good reasons to do this, but there's also a lot of cruddy systems with half-baked rules that are ripe for rules-lawyering and just bad implementation details (looking at you, FFG Star Wars). Writing a well-crafted, minimal-bug RPG engine is similar to writing a front end framework or something along those lines.