Matthew Cox recently wrote that, in the recently released beta version of Kerbal Space Program, he prefers Science Mode to Career Mode, specifically because “experimentation is so much of what makes KSP fun, and it’s far easier to do that in science mode where there are no real consequences when everything inevitably explodes for no discernible reason.”

I agree wholeheartedly about that specific example—Science Mode provides something of a concrete goal without sacrificing the game’s roots as a physics sandbox—but Cox also got me thinking about a more general principle in game design: Games are more interesting when I, the player, can fail completely at my main objective and keep right on playing. I’ve written about desperate situations in the original Legend of Zelda, and mining order out of chaos in Super Time Force Ultra, and what I called spectacles of attrition in King Arthur’s Gold.

Scott Juster and Jorge Albor, meanwhile, have said that Mario Kart 8 is the best Mario Kart partly because you can have fun while losing at it (which is a hard thing to pull off, from a design perspective—especially in Mario Kart, where luck plays an undeniable role). I buy that. But let’s go with another, even more recent AAA example, Middle-earth: Shadow of Mordor.

Shadow of Mordor is essentially two games at once, the first and more obvious one being a series nightmarishly rigid quests. Go here and do this, and be sure to do it within the designated mission area. Oops, you failed. MISSION FAIURE. Reason: You tried something novel and/or something interesting happened.

But as you’ve probably heard, there’s a second game semi-hidden in there, a fucking brilliant procedural open world masterpiece where any uruk who kills you gets promoted to Captain, and where Captains duel, feast, hunt, grandstand, bicker, grow powerful, and defend or betray their Warchiefs—where you can undermine and subvert Sauron’s forces or, better yet, make them into your forces, orchestrating a civil war among uruks and winning it by proxy.

The main thing that differentiates these two experiences, I would argue, is that the latter makes room for things to go wrong. Let’s say I’m trying to help one of my uruks win a duel, and I fail. My uruk gets swifty dispatched. The other dueling uruk (the one I don’t control) becomes more powerful, which could be bad for me, but the game doesn’t GAME OVER me, nor does it force me to start the section over. Rather, the game just keeps going, adapting to the new circumstances.

I can claim this other, apparently more powerful uruk as my own. Or I can seize the opportunity to kill him, or to interrogate him about one of the other Captains. There are lots of ways to salvage the situation—and even if things go truly wrong and I die, that simply means that my killer will become a Captain—a rival, a revenge target—and that all of the currently active feasts and hunts and such will resolve themselves without my intervention, leaving me with (possibly harsh) new difficulties and (probably exciting) new opportunities.

Basically, the game shines when it lets me ignore the main story and slip into another, stranger, messier one. This other story is built on the fly, and it embraces fuck-ups and unforeseen complications rather than finding them intolerable and insisting that I do it again and please have less fun this time.

This is the kernel of truth in Anna Anthropy’s half-serious adage that “AAA” refers to “a type of videogame that is only interesting when it breaks.” When everything goes according to plan in a big budget spectacle of a game, the play can get dull. Polished, and maybe even engaging, but never exactly exciting.

I’m confident that Kerbal Space Program will find the right balance between openness and hand-holding. Likewise with Prison Architect, that other slow-burning Early Access darling. But there are plenty of other games that could stand to let everything inevitably explode for no discernible reason. Within a sufficiently robust and flexible system, that chaos will only make things better.