20091230

Coop vs AI



An emerging trend in the world of strategy/tactical games of late is the focus on coop multiplayer vs AI, as opposed to the more traditional use of AI as a stand-in for an absentee 'real' opponent.

While not entirely new, this is an interesting development, and one of the many subtle ways in which the wargames (as a broad category/genre) are finally getting payback from the adventure/action/RPGs that have shamelessly 'borrowed' from strategy games forever, without giving back much love until recently.

***

While other genres have evolved spectacularly over time, thanks to genre fusion/bending and cross-pollination, strategy games have known few significant changes since the emergence of RTS:
at core, most strategy games tend to fall in either of two categories: computerized renditions of boardgame-styled designs, or mildly brainy action games sitting in a fancy simulated environment, with eyecandy added to taste.

The reason for that is simple: strategy and wargames are traditionally and by nature thought of as 'versus' games, yet current AI do thoroughly suck at emulating a 'real player' opponent.
Thus the tried and true solutions to make the AI look good boil down to either very solid game mechanics that even a seriously retarded bot can't screw up, or relatively simple game design (that an AI can handle, with cheats if needed) fleshed out by a rich enough simulated environment to overload the cognitive capacity of the human player and create the illusion of lifelikeness.

As a result, designing for single or multi player(s) has often been perceived as an either/or choice: a good single-player campaign usually will hinge on RPG-like scripting, crafty level design, extensive playtesting, with some storytelling and cutscenes thrown in to thicken the sauce, while the multiplayer PvP mode requires solid boardgame-like game mechanics to work — it's really tricky to do both at once and well enough in a single game.

***

Luckily, other game genres bring a comparatively lateral-thinking answer to this dilemma: cooperative multiplayer.
As hinted above, this is a counter-intuitive option for strategy and wargames, and although coop mode has been present in the multiplayer options of many a RTS in he past, it generally has been there for the sole sake of exhaustiveness, and rarely if ever been a defining feature of any srs bzs strategy game until recently…
Just to be clear, I'm talking about coop vs computer here, not team-based PvP, which is indeed cooperative too and can spice PvP quite a bit, but is still PvP at core and is not an essentially different play mode from FFA PvP, mechanics-wise.

Coop vs AI (CVA) comes from CRPGs (notably MMOGs) and action-adventure genres, which have been struggling for a long time to reconcile their single-player heritage with the commonly accepted wisdom (and now built-in expectation) that multiplayer adds value to a game. In those genres, Coop gameplay is a fairly obvious way to go about bringing more than one player on the same boat, and is also a welcome answer to the seemingly insurmountable problem of Artificial Stupidity of henchmen and other computer-driven teammates in RPGs and action-adventure titles.

***

Applied to strategy games, the eureka moment of going CVA involves putting the genre on its head: the common ancestors of all wargames are about two guys sitting on chairs across a table and duking it out over a board and tokens — at face value, making it about two guys helping each other to beat the crap out of a machine is more of a leap than a stretch — but in hindsight, CVA looks just as elegant and obvious applied to strategy games as to CRPG.

CVA works best when freed from the preconception that the AI's there to substitute an absentee human opponent, which AIs are notoriously bad at.
Dropping the Turing-test idiot ball and instead embracing a PvE-centric design approach allows to build a game that can be equally interesting in solo and multiplayer mode, as the game mechanics no longer need to cover for Artificial Stupidity. It's certainly extra work over just bolting more seats on a single-player game, but it's definitely the easier path (compared to faking human intelligence with scripts) to get something worth playing.

Coop doesn't mean the challenge, tension and sense of accomplishment have to be diminished, either : the human factor and drama fuel can be just as strong when players try to work together rather than intently butt heads, and a CVA design can still be spiced up with multiple coops competing against both the AI and each other, as long as the AI is clearly defined as the primary threat/target.

The brilliant (despite disputable looks) 2009 poster child for that approach is probably AI War — go check it out now if you haven't yet.

…and if you want to know more about the AI in this particular design, here's a good and thorough writeup series.

***




No comments: