Editorial Rampant Coyote on Dice Rolls and Luck

[Marsal]

You should elaborate further, Karmapowered. I find your posts severely lacking in detail.

Your combat example is great. It would work even better with some dice thrown in. What if you slip while attacking? What if your carefully placed blow just glances the armor? What if your blade breaks while deflecting the blow? Or what if you get lucky? Your highly skilled opponent steps on a rock while evading a wild swing, falls and hits his head on the stone floor.

Randomness accounts for the unexpected. Not everything is under our control, even if we sometimes think it is. You think there was no "luck" (or chance) involved in Bill Gates amassing his enormous wealth? In Hitler coming to power? In USA winning the battle of Midway? You can't plan for everything. You've got to adapt sometimes. And adapting to failure is fun. At least in games it is.

Also, Einstein was "wrong", just like Newton was "wrong". Standing on shoulders of giants, bro.

 

[EG]

At the heart of it, I play games to deal with unexpected situations rather than to manage resources in the most efficient way possible (and thereby win by default). o_O It shouldn't be one or the other.

 

[CappenVarra]

Karmapowered I'm not opposed to walls of text (as you can see in my own post), as long as you don't expect prompt responses. As I said before, Wall of Text is a spell I can use only once per day.

Since we seem to be in agreement about a number of things, I'll skip those parts and try to focus on the main points of contention.

I stand by the point that "fundamentally, PCs are the player's input and responsibility (and the more control players have over their characters, the better), while the game world is solely the game maker's responsibility" - it's just that you seem to assume that "control" equals "control over outcomes", while I use it in a narrower sense of "control over inputs". (And now we've reached the pre-orgasmic point of every codex discussion: arguing semantics. Yay.)

Basically, outcomes of actions are based on the combination of the player's input and the game world + ruleset. Yes, the player's actions are strongly correlated to their outcome, but I'm arguing that there are benefits of decoupling the two. The amount of such decoupling is arguable and depends on what the game wants to achieve (I believe the main reason AoD doesn't use randomization more because it's focused on the fabled C&C diversity, which requires a near-identity translation from player inputs [skill point allocation + dialogue choices] to outcomes).

But the imperfection of action execution (which Marsal already mentioned) is an important aspect that a game striving for any amount of "simulationism" should preserve. On one hand, you're increasing the level (resolution?) of detail in the system (with detailed hit locations and other similar ideas in your combat example, which I'll get back to), but you're simultaneously reducing it with:

I don't give a flying fuck about any "pseudo-realistic" chance of 5,5476476% that they might fail. Insignificant, hence pointless to have or keep.

Which brings me to this: I'd love to try a game that plays like your combat example. Even more if you added some randomness (as Marsal already said) It is certainly more interesting than the direction taken by current (pre-Kickstarter) quasi-RPG devolution.

However, without randomness (or other means of weak decoupling of player inputs from action outcomes), what it reminds me more of an adventure or action games than cRPGs. (Of course, "reminds me" is obviously a subjective function). But just like you consider a game in which the randomness is cranked up to 11 to belong to a completely different genre:

There is maybe one exception in the genre that is dear to some of us (me included) : rogue-likes. Why is not pointless in rogue-likes ? Because it's a genre in itself, that is played and enjoyed on its own merits, and especially because it doesn't tamper with randomness. There is no such thing as 50%, 30% or 10% randomness in them. It's brutal, it's harsh, it's 110%, it's *everywhere*. It's just not for every gamer.

I can't see why the logic doesn't also work in the other direction: when you reduce the randomness to 1, you also end up with a completely different genre, with completely different considerations and for a different player audience.

Furthermore, while purely mathematically you could say that anything that doesn't make an order of magnitude difference can be considered insignificant and, consequentially

I don't give a flying fuck about any "pseudo-realistic" chance of 5,5476476% that they might fail. Insignificant, hence pointless to have or keep.

In practice, that slight margin of fuzziness is exactly where the magic happens. While a chess pawn has 100% chance to capture any other piece located at its threatened field, cRPG characters never do (and I'm of the opinion that they never should). To bring too much order into organic matter is to kill it by freezing. To bring too much chaos into organic matter is to kill it with fire. But to keep it alive requires just the right, narrow band of acceptable temperature. Now, obviously, computer code is not "organic" and I'm skeptical It'll ever be, but it absolutely needs to fool the players with faked "organicness" if it's to result in an interesting RPG.

Also, I'd say that the reason Wizardry I (and Ultima I on the other hand) was such a breakthrough success that created the cRPG genre for all practical purposes is exactly that it had enough of that faked "organic" quality for the players of that age. It is not beside the point to say that we need somewhat different mechanisms of achieving the same result 30 years later (even if nostalgia keeps the Wizardry kind close to our hearts); after all, nobody maps their dungeons by hand anymore and spending 3 months figuring something out that you could google in 3 seconds is an... increasingly uncommon quality to be found in players. However, it would be foolish to claim that faked "organicness" is not necessary.

Which brings me to the next point: to me, it seems you are offloading too many of the required qualities of the game to AI. And while I'll be thrilled to see good AI in action, in practice I've never seen one that wasn't bypassable by the player. So previous experience makes me wary of anything that places sole responsibility in the hands of AI. Don't get me wrong, I'm not saying that you shouldn't spend the next 20 years writing an AI that makes your combat example possible; on the contrary, please do.

But I am saying that it sounds a bit like you're approaching a bunch of guys arguing about how to best design a car that suits their tastes and insisting that car design is dead and improvements can only be made with personal anti-gravitational belts. i.e. that's all well and fine, but: a) where do we get one, since nobody has ever seen such a thing? b) we're surrounded by idiots that insist rickshaws are the only possible option and cars were only ever used because of technical limitations c) even if we were living in a world of anti-gravitational belts, would there no place for car enthusiasts? d) wouldn't we keep calling cars "cars", instead of discontinuing all cars and changing the language so that "car" means "anti-grav belt"?

And yes, I have a randomness fetish. Hail Entropy!

 

[Carrion]

I didn't go through all of Karmapowered's posts, so maybe he or someone else already addressed some of this, but:

A non-random system can work perfectly well in single-character games, Gothic being a good example. You have full control of all of your actions, and every time you die it's your fault. Character skill matters too, and overall it's a system that works pretty well. When you try to apply it to a party-based game, however, it gets much harder. How do you determine whether a character manages to deflect a blow or not, especially if she's not under your direct control? Should the AI just play it out by itself, and if so, how would you do that so that it's somewhat unpredictable without being random? You'd still have randomization somewhere else, like in the reflexes of a character, or in her accuracy if she's using ranged weapons. The level of abstraction is just lower, which can be a good or a bad thing depending on a game, but which doesn't really remove die rolls in any way. Human beings aren't machines who can perfectly reproduce the same action over and over again, and a good AI should take this into account (by my experience the games that have the most randomization are actually the ones that try to simulate reality). You can be a master sniper, but if you fire ten shots at a target in a row, your every bullet will still hit a different spot. And what about turn-based games? The attacker is allowed to walk up to an immobile target and just bang away. Should the attacker always hit? How would parrying work? How would you make this system work without random elements? I'm out of ideas here.

If we really go all the way and eliminate randomness completely, we might end up with the system Sawyer mentioned when talking about the new XCOM, where you always hit your target, where your damage is fixed based on the character's skill and where things always play out exactly the same when using the same tactics. Every battle basically becomes a puzzle that you have to solve. To me Sawyer's idea sounded terrible because assessing risk and reward is one of the most fun things in a turn-based system. Should you try that power attack even with a higher risk of failure? When things get really tough, should you use your best spell immediately or cast a few debuffing spells on your enemies first to gain a greater chance of success, even if that might mean losing a party member or two? The most memorable things are usually those where somewhen thing unexpected or unlikely happens: getting that vorpal hit against a powerful opponent when you only have two hit points left, hitting your party members when shooting into melee (okay, this one is always to be expected, but it's still funny it happens), or failing an easy saving throw and getting transformed into a chicken in a battle that looked like it would pose no challenge at all. A part of being a good tactician is being able to react to all sorts of stuff that could happen during a battle, and that is the factor that makes fights exciting in the first place.

Besides, in party-based games randomization isn't really any kind of a problem to begin with because the amount of die rolls in combat is usually so big that the good and the bad rolls will more or less balance each other out even in the course of a single battle. You're unlikely to win or lose a fight just because of pure luck (well, save-or-die spells excluded), but the fight can still take unexpected turns and play out completely differently every time.

 

[Craig Stern]

I'm a little disappointed in Jay's article, honestly, particularly in his mischaracterization of the thesis of this article as being “randomness in RPGs, BOO!” I mean, if you're going to try to sum up an article's conclusion, a good place to look might be in the section marked "Conclusion," which might, for instance, read something like

There is a whole world of tools for building unpredictability. I would like to see turn-based combat systems–particularly those in RPGs–start focusing on a greater variety of these, and stop using randomized results as a crutch. Randomized results have their role, but I’d like to see them used more deliberately. Like special sauce on a burger, randomized results can add spice; just make sure the player still gets to taste the meat.

That's just a wee little bit more nuanced than “randomness in RPGs, BOO,” I think.

Anyway, randomness is not the same thing as unpredictability. Karmapowered seems to get it.

 

[DraQ]

The main advantage of randomness is noisying things up to ensure that gameplay doesn't rely on string of perfect, flawlessly executed, but extremely fragile plans as it allows those plans to be tested for robustness as well.

That's from gamist PoV, from simulationists PoV they are required because a lot of stuff is random or relies on stuff below simulation's threshold of granularity, but that's pretty obvious.

"pre baked" randomness

Won't help. Shifting between true randomness, pre-baked randomness or fixed but unknown outcomes can't eliminate advantage conferred by savescumming.

It can merely change the way how savescumming can be exploited - from trying to enforce favourable outcome to knowing what unfavourable outcomes to avoid or lessen impact of.

How is this different than if the skill thresholds are hardcoded?

That's not a valid excuse. You don't want to suck. sucking as much as existing solution, but in different way is just as bad.

You don't want savescumming to break your game, you limit savescumming either directly or by making it bite player in the ass.

Some minor fixed randomized adjustments to all fixed skill checks would definitely be a great addition.

Formally equivalent to just rolling for success.
Sorry.