I wonder whether a large part of the problem is that, with large AAA budgets forcing developers to be risk-averse, there's no longer room for innovating upon the 'glorious failures'. By which I mean games that ultimately failed because they weren't all that great overall, but which could have been the first drafts for games that are truly special.
Example: Obsidian's Alpha Protocol. The game is weighed down by poor stealth feedback (ruining what is a fairly impressive system 'under the hood', with its focus on AI reacting to noise, and that the right gear is critical to moving with any degree of silence/stealth whatsoever), inconsistent combat quality, the use of 3rd person camera for a genre where FP would be far more preferable, and the silly separation of 'safe-house' hubs and the mission sites which hampers the non-combat interactivity.
BUT it still has, to date, the greatest choice+consequence interactivity of any game to date. And I don't mean the bits where you're told outright 'choose A or B'. In fact, those choices are part of the problem, as it encourages players to think that they're the only bits where your choices matter, when often the most important choices are subtle and well integrated into the gameplay, without any intrusive 'here's a big choice' signals.
If you could put up with the failings, it's the kind of game where you could play it a dozen times, read every FAQ around, and STILL miss major deviations in how the character arcs and main plot plays out. There are many ways of getting just about every major character on-side or off-side, many fights that you can only get if you combine special info with pissing someone off enough, or getting them to lower their guard enough, that they'll expose themselves instead of fleeing. One character, for example, can be convinced to help in the end-game by either discovering his connection to another dead character (needs discovery of certain info in the mission, plus particular choices with the other character while that character is still alive) OR by discovering enough intel on the geo-political situation that you can convince him that his calculations are off and he's about to create a hot war instead of a cold one, OR by combinations of discovery and intel to show him that a certain other character is untrustworthy. All of these options are easy to miss on most playthroughs.
You can befriend or backstab literally every character - you can join the lead bad guy, backstab him to take over his plot once he's outlived his usefulness, or (VERY difficult) get enough respect from his second in command that you can join with that guy and run a less heartless version of the bad guy's scheme.
You can get backup troops in the end-game either by befriending a certain ally, or by making him hate you so badly that he'll send troops after you (which aids you by making it a 3-way fight instead of you vs an army).
None of it is in the form of linear 'have X points in X stat' stuff either - it's all from your choices in-game.
And here's the big thing - it's ultimately based on systems, not just good writing. Obsidian explained it as this: instead of having a branching plot (starting with 1 thread and splitting into branches), they used a hexagonal plot system. Think of a square made up of 8 lines of 8 dots each. So each of the 8 start points will ultimately get to one of the 8 end points....but with an amazing array of potential paths that can be taken to get there (you could go straight to the other end, or zig-zag your way to the other side, or spiral to the centre and out again, etc)
...and so it's a mechanic that could have been replicated by other games that do a better job of the rest of the mechanics.
Dead State is another example. A game where the plot (and the survival/actions of other survivors and camps) moves forward with the passage of time (day passes whenever you rest at the shelter) instead of the player's actions. Again, wonderfully innovative mechanic that in a previous era would have had breathing space to be refined over multiple iterations, until we could a game that makes better use of the mechanic, and avoids the game's weaknesses.
Nowadays a game is seen as not worth drawing from if it isn't a hit in itself. Glorious failures are forgotten, depriving us of the opportunity for glorious successes.
Example: Obsidian's Alpha Protocol. The game is weighed down by poor stealth feedback (ruining what is a fairly impressive system 'under the hood', with its focus on AI reacting to noise, and that the right gear is critical to moving with any degree of silence/stealth whatsoever), inconsistent combat quality, the use of 3rd person camera for a genre where FP would be far more preferable, and the silly separation of 'safe-house' hubs and the mission sites which hampers the non-combat interactivity.
BUT it still has, to date, the greatest choice+consequence interactivity of any game to date. And I don't mean the bits where you're told outright 'choose A or B'. In fact, those choices are part of the problem, as it encourages players to think that they're the only bits where your choices matter, when often the most important choices are subtle and well integrated into the gameplay, without any intrusive 'here's a big choice' signals.
If you could put up with the failings, it's the kind of game where you could play it a dozen times, read every FAQ around, and STILL miss major deviations in how the character arcs and main plot plays out. There are many ways of getting just about every major character on-side or off-side, many fights that you can only get if you combine special info with pissing someone off enough, or getting them to lower their guard enough, that they'll expose themselves instead of fleeing. One character, for example, can be convinced to help in the end-game by either discovering his connection to another dead character (needs discovery of certain info in the mission, plus particular choices with the other character while that character is still alive) OR by discovering enough intel on the geo-political situation that you can convince him that his calculations are off and he's about to create a hot war instead of a cold one, OR by combinations of discovery and intel to show him that a certain other character is untrustworthy. All of these options are easy to miss on most playthroughs.
You can befriend or backstab literally every character - you can join the lead bad guy, backstab him to take over his plot once he's outlived his usefulness, or (VERY difficult) get enough respect from his second in command that you can join with that guy and run a less heartless version of the bad guy's scheme.
You can get backup troops in the end-game either by befriending a certain ally, or by making him hate you so badly that he'll send troops after you (which aids you by making it a 3-way fight instead of you vs an army).
None of it is in the form of linear 'have X points in X stat' stuff either - it's all from your choices in-game.
And here's the big thing - it's ultimately based on systems, not just good writing. Obsidian explained it as this: instead of having a branching plot (starting with 1 thread and splitting into branches), they used a hexagonal plot system. Think of a square made up of 8 lines of 8 dots each. So each of the 8 start points will ultimately get to one of the 8 end points....but with an amazing array of potential paths that can be taken to get there (you could go straight to the other end, or zig-zag your way to the other side, or spiral to the centre and out again, etc)
...and so it's a mechanic that could have been replicated by other games that do a better job of the rest of the mechanics.
Dead State is another example. A game where the plot (and the survival/actions of other survivors and camps) moves forward with the passage of time (day passes whenever you rest at the shelter) instead of the player's actions. Again, wonderfully innovative mechanic that in a previous era would have had breathing space to be refined over multiple iterations, until we could a game that makes better use of the mechanic, and avoids the game's weaknesses.
Nowadays a game is seen as not worth drawing from if it isn't a hit in itself. Glorious failures are forgotten, depriving us of the opportunity for glorious successes.