Magic Numbers

[Josh_Kablack]

What's the context, here? A 60% miss chance for a given action means something completely different when you're talking about having one action vs having five, or whether you're talking about all or nothing actions or ones that have secondary effects even if you miss, or whether or not the miss chance involves enemies expending resources or not, and so on and so forth. You can't just pull out some "magic" number outside of any meaningful context.

The thing is, the "hit/miss ratio" the OP brought up isn't a terribly important concept on its own. For example, if you fire 12 shots in a turn and expect half to miss, that's very different from if you fire one shot in a turn and have a 50% chance of doing nothing and a 50% chance of doing full damage. In the latter case, a player can reasonably expect to go multiple turns without accomplishing anything whatsoever on a rather frequent basis and the gameplay is more swingy.

In short, I think the OP is asking the wrong question.

This. Also recall that the to-hit rolls of the enemies are actually the activation rolls of the player's defenses. If a player's defenses fail to reduce or negate the effects of incoming attacks for several rounds in a row, the player is as apt to ask what the point of having the defenses is as when their attacks fail to accomplish anything for several rounds and they justifiably ask what the point of taking actions is.

-Username17

I'll take the meta out one further.

It also depends on how much actual real time at the table actions take and how many of them there are in total and how they are sequenced around the table.

If you get a pile of actions for a few minutes but then everyone else gets a pile of actions, there's a pretty big go-round before your input matters again, and in such a case it's essential that your input always have a meaningful result and never end in total failure or futility.

[ishy]

The question asked by the op is what the various percentages are for people to stay interested. And an example I can give there are 3e flask rogues or pf alchemists. With touch attacks and usually some good attack bonus, all attacks usually hit on a 2. So when people play those classes/builds the attack roll is not interesting to them at all.

This. Also recall that the to-hit rolls of the enemies are actually the activation rolls of the player's defenses. If a player's defenses fail to reduce or negate the effects of incoming attacks for several rounds in a row, the player is as apt to ask what the point of having the defenses is as when their attacks fail to accomplish anything for several rounds and they justifiably ask what the point of taking actions is.

-Username17

This thing right here is why many videogames don't use the listed percentages anymore, but instead change the % based on if you were hit on the previous attacks. So that if you have a 95% chance to hit, you can't miss 4 times in a row. Because with enough people playing it will happen eventually and people complain about that sort of stuff a lot.

[Foxwarrior]

There are games that lie to you about the probabilities of things out of fear of offending people?

That's so depraved.

[Avoraciopoctules]

Have you ever played Battle for Wesnoth? It doesn't lie about success probabilities, and the amount of whining in the fanbase can be truly stupefying. There are mini-essays in the FAQ preemtively explaining why you are wrong if you think the miss chances are off.

[RadiantPhoenix]

There are games that lie to you about the probabilities of things out of fear of offending people?

That's so depraved.

One can do it honestly.

State something like, "The listed probability of your attack hitting is the base probability. If you miss (2 / P(Hit)) times in a row, your next attack will be an automatic hit, and if you hit (2/(1 - P(hit))) times in a row, your next attack will automatically miss. This is only true when your P(hit) is greater than zero and less than one."

[Foxwarrior]

Doing it honestly is good, although it leads to strange combos between quick attacks and strong attacks.

Avoraciopoctules: People whining about probabilities only to find that they were wrong means that they were ignorant before, and now are less so. Considering how often people complain about the majority's stupidity, that must be a good thing.

[violence in the media]

Have you ever played Battle for Wesnoth? It doesn't lie about success probabilities, and the amount of whining in the fanbase can be truly stupefying. There are mini-essays in the FAQ preemtively explaining why you are wrong if you think the miss chances are off.

I've played it and it's an addicting game. Though aside from the people arguing that there's something wrong with the programming of the RNG, most of the complaints are about the frustration of having 4 70% attacks miss and then getting walloped by 3-4 30% attacks in return, or when 4 dudes with multiple 50%+ attacks fail to kill 1 regular guy. Given the number of turns and attacks that happen in a given campaign, it's almost inevitable that events like that will happen seemingly "all the time".

I find myself thinking of Elennsar frequently when I play the game.

edit: Blah, blah, blah...Confirmation Bias...blah, blah, blah...no one complains about strings of unreasonably good fortune.

[angelfromanotherpin]

There is something wrong with the RNG in Battle for Wesnoth. I've actually mass-savescummed to check for probability bias. The most striking example was one in which out of 100 uses of 3x70% attacks, The average number of hits was <1. But that was only the most flagrant instance, there were many others which bucked the odds a lot, even under significant reiteration.

[tussock]

Likely you're not testing what you think you're testing, as it'll be saving the state of the RNG when you save the game. The only thing your save-scumming is testing is how much you can change the state of the RNG between loading and taking your attack, which isn't much.

[codeGlaze]

Likely you're not testing what you think you're testing, as it'll be saving the state of the RNG when you save the game. The only thing your save-scumming is testing is how much you can change the state of the RNG between loading and taking your attack, which isn't much.

You can't save the state of a RNG, as long as you're talking about true randomization.

That'd be illogical.

[Caedrus]

You can't save the state of a RNG, as long as you're talking about true randomization.

That'd be illogical.

I'm not sure what you think "true randomization" is supposed to mean in the context of computer games.

Now, I don't know anything specifically about Wesnoth's RNG. However, I'm pretty sure you can "save the state." For an overly simple (and likely unused method) example, say you generate 1000 random numbers when you start a game, then use them in order. It's functionally just as random as if you rolled it on the spot, but if you saved the game then continued, you'd always get the same result.

All you'd need to do to produce a save-scumming-resistant system would be to produce and store a random seed and call count or some similar thing.

But this seems like it's getting off topic.

[Murtak]

You can not define a best or even decent hit percentage (or save percentage for that matter) since that fluctuates wildly, as has been discussed. One thing stays constant though: Every action should be meaningful. That is, any successful action needs to visibly change the state of the game. Also, at least for player characters, successful actions need to be more common than unsuccessful ones.

So if you are fighting a hundred orcs and you attack them, just killing an orc with every action is not good enough. Seriously. 100 Orcs or 99 orcs? An imperceptible difference. No, if you are mowing down hordes of enemies you need to at least have a chance of taking out half a dozen orcs in one action. If you are fighting a dozen orcs, taking one of them out of the fight is fine, and need not even be guaranteed. And if you are in a duel with the orc chief it is fine if half of your actions do nothing, or only minor things like forcing him to use an active defense or fatigue points or whatever. And if you and your friends are fighting a single dragon or titan or sea serpent then just trading your actions for it's actions with neither of you doing anything is a huge success, as it leaves your friends free to win the fight.

But in all of these examples your actions count. The difference between you taking your action and not taking it must be visible. You have some amount of leeway for playing with success percentages and outcomes - say, a 50% chance of killing an orc outright vs a 75% chance of severely wounding an orc - but the important thing is to ensure that not having your character present and hacking orcs into pieces will be noticed by the party.

Incidentally that is a big part of what makes 4E bad. It is not even that their balancing is bad - I doubt they could top 2E, no matter what they tried. But the stupid amount of tiny bullshit bonuses and penalties is actually insulting. If you can't tell fights where you remembered to bless the party from fights where you did not the ability might as well not exist. And the same goes for all abilities, including attacks and saves.

[ModelCitizen]

All you'd need to do to produce a save-scumming-resistant system would be to produce and store a random seed and call count or some similar thing.

But this seems like it's getting off topic.

Fuck the topic, this tangent is way more interesting.

You could also save the state of the PRNG by getting the current seed and storing it in a save file. However, this wouldn't prevent save-scumming. If you get the same PRNG seed every time you load a save then you get the same sequence of random numbers every time, but the game can't actually force you to use the first number in the sequence for the check you care about. Instead of failing a check, reloading and immediately retrying the check until you succeed, you fail a check, reload, take a superfluous action that requires an RNG call, and then retry the check that matters. If that doesn't work you reload and take two actions that require RNG calls and then retry. If that doesn't work you take three actions, then four actions, and so on until you succeed.

To put it in Fallout 3 terms, you can save-scum Fallout 3 by reloading until you pass a Speech check. In a hypothetical Fallout 3 that saved the PRNG seed, you could still save-scum it by failing a Speech check, reloading, and firing your pistol into the air a variable number of times. This makes save-scumming simultaneously more attractive (because instead of relying on factors you can't see or control you're finding the optimal sequence of actions by trial and error) and more irritating (because it's metagamey and counterintuitive). Basically it's fucking awful which is why games don't do it.

[codeGlaze]

All you'd need to do to produce a save-scumming-resistant system would be to produce and store a random seed and call count or some similar thing.

But this seems like it's getting off topic.

To put it in Fallout 3 terms, you can save-scum Fallout 3 by reloading until you pass a Speech check. In a hypothetical Fallout 3 that saved the PRNG seed, you could still save-scum it by failing a Speech check, reloading, and firing your pistol into the air a variable number of times. This makes save-scumming simultaneously more attractive (because instead of relying on factors you can't see or control you're finding the optimal sequence of actions by trial and error) and more irritating (because it's metagamey and counterintuitive). Basically it's fucking awful which is why games don't do it.

S'what I did in XCOM.

[ModelCitizen]

Chrono Cross on emulator for me.

[tussock]

OK, so just to clarify, a computer's RNG is a thing that turns one number into another number, usually involving dividing by 13 somewhere along the way. The trick is every random number it produces is also the seed for the next random number, and it just walks it's way through it's own little subset of probability space from there until it finally loops (which will ideally take enough years so you can't possibly notice).

You can totally make a new seed each time the program starts up, but you certainly don't do that every time you call the RNG. Many programs just save the "game logic" seed separately and carry on where you left off. A huge advantage of this is in playtesting and bugfixing where you can duplicate errors by following steps from a save game.

[nockermensch]

OK, so just to clarify, a computer's RNG is a thing that turns one number into another number, usually involving dividing by 13 somewhere along the way. The trick is every random number it produces is also the seed for the next random number, and it just walks it's way through it's own little subset of probability space from there until it finally loops (which will ideally take enough years so you can't possibly notice).

You can totally make a new seed each time the program starts up, but you certainly don't do that every time you call the RNG. Many programs just save the "game logic" seed separately and carry on where you left off. A huge advantage of this is in playtesting and bugfixing where you can duplicate errors by following steps from a save game.

This is true, but also there also programs that implement some way to circumvent/ignore this characteristic of PRNGs. I don't know what are the current methods for that, but I'd multiply the seed for the current millisecond counter, or some other input beyond the players' control.

[Desdan_Mervolam]

In Torchlight 2 you autohit every attack, and your to-hit is only to determine if your hit crits (Does extra damage) or fumbles (Does significantly reduced damage).

Perhaps a system where there's a minimum damage done whether you "hit" or not might be worth looking into?

[Prak]

On the basic question, of opinion, I would think something like a 55/45 Hit/Miss rate for PCs, and a 45/55 Hit/Miss rate for NPCs. So that PCs are just ever so slightly more likely to succeed and NPCs are ever so slightly less likely to succeed.

[codeGlaze]

In Torchlight 2 you autohit every attack, and your to-hit is only to determine if your hit crits (Does extra damage) or fumbles (Does significantly reduced damage).

Perhaps a system where there's a minimum damage done whether you "hit" or not might be worth looking into?

I was actually just thinking about this last night.
What if hits were done more akin to skill DCs?
Have a (relatively) low "you must be this tall to ride" number. Anything over that number could allow characters to do different tiers of damage, or damage+status effect. While rolling under would have minor to 'major' (a 'confirmed' crit fail) consequences, too.

Damage could be a static list for each individual or at least a list of how many d4s (or something) to roll within a given 'pass' range.

This is me just thinking 'out loud'.

[tussock]

but I'd multiply the seed for the current millisecond counter, or some other input beyond the players' control.

You need massively more entropy than the RNG, otherwise you just keep dropping into an even smaller subset of the probability space than the RNG provides. Getting one really good seed and then leaving it to the RNG should work best, as long as your loop is always sufficiently large.

Picking an arbitrary set of bridge hands for competition is always interesting, you want a loop that can generate something like 2^96 different numbers in every possible order, which is an extremely large amount of entropy that RNGs aren't written to hang onto even if you could provide it to them, so every program cheats in some way or another.



@auto damage. Ew. It doesn't feel any better for the players, and makes all the outcomes more certain. The illusion of challenge is a delicate one, and may not survive that sort of thing.

[codeGlaze]

@auto damage. Ew. It doesn't feel any better for the players, and makes all the outcomes more certain. The illusion of challenge is a delicate one, and may not survive that sort of thing.

It'd reflect 'rolling well' better.
I can obviously see arguments against that very point, but it does make me wonder if house-ruling crits to be auto-max damage would be fun; at least with the group I run with. (Only about half optimize; and only half again do it well >_>)

[norms29]

To put it in Fallout 3 terms, you can save-scum Fallout 3 by reloading until you pass a Speech check. In a hypothetical Fallout 3 that saved the PRNG seed, you could still save-scum it by failing a Speech check, reloading, and firing your pistol into the air a variable number of times. This makes save-scumming simultaneously more attractive (because instead of relying on factors you can't see or control you're finding the optimal sequence of actions by trial and error) and more irritating (because it's metagamey and counterintuitive). Basically it's fucking awful which is why games don't do it.

I feel I should point out that some games DO.

Actually, the only one I can name is Civ 3, which, I shit you not, has a check box when you start new game where you can decide whether or not to pregenerate the random seed at the start of the turn or wait until it needs a random number.

[Caedrus]

So if semi-related but more interesting tangents are okay, I might as well turn it back to pen and paper RPGs with a controversial tangentially related topic. I already seem to have gotten across the point that you don't actually care about hit %s on their own... but more an issue of how much of an impact player decisions have and how often.

A lot of talk on RPG message boards talks about dice, and treats them as a sacred cow... including, to some degree, TGD. One might even say the dice are the biggest, fattest sacred cows of them all. But how valuable is the RNG really?

Many games, including videogames and more recent "designer board games" go out of their way to avoid RNGs (sometimes even treating them as taboo), or place alternative methods of randomization (often ones that involve hedging the odds or playing mind games, such as many card systems. Think the difference between slot machines and poker). And yet to hear some of the talk on here, you'd really believe that one of the things at the core of what gamers valued was seeing the way the dice fell.

But what if we went beyond the idea of dice-based RNGs? What if we had systems where you had to worry more about the next piece in the pattern, adapting correctly to new and dynamic situations, guessing the enemy's next move, or, in cases where there is randomization, making it something that's a bit more under your control (such as being able to draw a hand of cards and then decide how they're used), such as the way Hero Academy lends variety to its matches? Mind, these things aren't necessarily mutually exclusive with RNGs, but they have the potential to replace the role they fulfill, and doing so can lead to a game with deeper and more concretely rewarding strategic play. When one sits down to play, say, Culdcept Saga (I should probably pick a better example than a relatively obscure videogame series but whatever), what's more rewarding? Rolling the right numbers to land on the right space? Or the back and forth bluffing involved in battles where players try to figure out which item card (if any) will be played by their opponents and how to react to that? Bluffing the enemy into wasting his defenses or making him think he's calling your bluff then using your own items? Or using powers from your carefully constructed deck to forego or manipulate the results of rolls with idols, spells, and terrain effects?

Would you rather lop a foe's head off by rolling a 20 on a vorpal sword, or put your enemy in a hopeless situation via careful investment of randomly provided resources (such as you do in Hero Academy, which could just as easily be a board game)? Both situations involve randomness, but in the former it's deciding your ultimate failure or success rather significantly after all the strategic maneuvering is already done (and thus you might always approach the situation rather similarly and then succeed or fail based on the roll of the dice), while in the other it's serving more to force you to adopt new strategic approaches from the outset. There are of course elements of that in the former as well... whether you hit or missed an attack forces you to change up your plan for the next turn. But hopefully you understand what I'm trying to say.

Any thoughts here? What do you guys think of moving focus away from traditional RNG resolution, and how might you do so? The possibilities for things like mind games and anticipations of the other guy's strategy are made less obvious by the lack of a true adversarial format.

[Username17]

RPGs are by definition more open ended than a board game or card game can be. This means that actions can have fuzzy inputs. While in a board game you can determine "exact" inputs, in an RPG you can't. And that gives you two options:

• Approximate the fuzzy inputs and then generate a random number.
• Have an endless argument about how slick pavement is or whatever the fuck every time you have a borderline result.

Even Munchhausen has an RNG to resolve borderline cases generated by fuzzy inputs and fundamental disagreements.

-Username17