Jump to content
  • entries
  • comments
  • views

Balance Part 2, Randomness



On randomness and encounter Balance


Warning : Most of these things are either my opinion or gathered from a variety of sources, which I can quote on request.


The age old problem : "Randomness isn't actually Fair"

Or worded differently: Most humans (and thus players) suck at probability.

Or less politely worded (wheither correct or not): The PC is a cheating Bastard (and that's OK).


Most RPG's, heck, most games, hold within themselves a form of randomness. 

But when is this desired, and when is it just bad game design hiding as (Fake) difficulty?

I go back to my favorite standby when I talk about randomness : Magic the gathering 

This not an RPG, I get it, but it illustrates my point, so bear with me.

Magic has a heavy randomness element in your opening hand of 7 cards. Did I get the right spell land combination ? 
An acceptable hand is between 2-5 lands. 
In a deck with 40 % lands this results in an 85 % Chance to get this mix, with a 50 % chance to get the optimal 3-4 lands. (Math not shown, but can be supplied on request)

On top of that, a player is allowed to mulligan , drawing a new hand with one less card. He may do this any number of times.

Even for a 6 card hand, which has a 40 % win chance on average according to the pros, the odds of a playable hand (2-4 lands) is till 74 %. 
So a player should only be affected by real flood/screw once every (100%-85)*(100%-74%)= <4% of the time, and have to go down to 5 Cards in hand.(Which technically still leaves you at a 20-28% chance to win according to the pros)

Now, mana screw (not drawing enough lands) and mana flood (drawing too much land), or our opponent getting a lucky draw are still the number one complaint about magic from beginning players, yet the designers state it to be one of magic's greatest assets.


Why ?


Part of it is mechanical : New players don't shuffle enough, leading to clumps in the deck (because at the end of a game , 
most land cards are in the same place, so when you go to pick them up, they form a clump in the deck , which should be shuffled away), they draw a bad hand,
they get mad, don't shuffle enough again , and one again draw a smaller hand from the still excessively clumped deck.


There is also the big impact of that opening 7. Nothing is less fun then outright losing the game because of factors beyond your control. So your mind doesn't tend to remember the games where you drew just enough land and spells. It remembers the couple of times you got screwed.


It also allows the worse player to sometimes win.
Why is this necessary ? Because nobody likes losing their first 50 games (I'm looking at you chess!).


Our feeling of luck also isn't symetrical. If we draw the right card at the right moment, it was "skill" that lead us to this moment, but if our opponent does it, it's just luck that made him win the game.


We also suck at estimating and interpreting probability. Suppose we have only 3 cards that could have saved us from a loss in our 30 card deck, most players will genuinely feel like they still have pretty good odds of winning, and blame randomness when they lose when they only had a 10 % chance to make it).

Another thing that inhibits us/our player in our understanding of  randomness is the distinction between random and fair. If we flip tail three times in a road, we feel like the next coin flip should be heads, but our chance to toss either is still an unchanged 50%.
Yes, throwing TTT is unlikely (1/8 chance), but starting from TTT, TTTH is as likely as TTTT. Now, on a long enough timescale the amount of T's over H's will trend towards 50 %, but nothing guarantees that the next one will be H or T.
The opposite is the "Hot Streak": I'm having good luck tonight, so I should continue to have good luck. "Ooh, Red is on a hot streak , better not bet on black."


The final problem I will discuss is goalposting. The first couple of times we use a certain skill/card/die we make a judgement on it. It worked the first 3 times I used it, so it must be good. No matter that it hasn't worked since, the players impression will remain that of a good skill.
This can of course backfire a bit. Suppose that a player uses a 50 % accuracy skill, and by coincidence hits the first 3 times he uses it. When later on it starts performing more according to it's accuracy, the player will suddenly feel the skill or the random number generator is out to get them.



The most succesfull MtG players in the world know all of these to be a factor and account for it.Dealing with randomness is a player skill. They realise the randomness in the game but know that they have so much control over it. They know and have heard all of these "bad beat" stories a 1000 times, and they know that in most the cases luck had little to do with it.
They know the things that a player can do to reduce randomness, and apply them (Play more copies of cards you want to draw often, mulligan correctly, make sure you have some cheap spells in case your deck decides to not give you a lot of lands, play some mana sinks in case you draw too many, play more/less land,...)
They believe in the numbers and less in gut feeling (bluffing aside). It also helps that games are best 2 out of 3, but I don't really think that would apply to an RPG 
(Somebody please make this!. Can you imagine having to fight the boss in a best 2 out of 3 match ?Fighting games can do it, so why not RPG's?)

Now how do we apply this to an RPG ?

1. Give the player agency in how to approach Randomness

This is the main difference I believe in bad randomness and good randomness. Magic for example gives players a big measure of control in how to approach randomness. 

Now, how do we do this in an RPG ? I believe this can be achieved by allowing players to choose how much randomness they want. 

For example, have some high damage low accuracy weapons, but also some weaker 100 % hit ones. Have some skills hit Randomly, but certainly not the majority. Advertise these options.
Another small thing to add: if you build a lot of skills/weapons, ... that nobody Loves/Hates, but everybody kind of likes, then you have achieved blandness, not balance. Something for everyone but not everything for someone. 
Don't be afraid of angering some your players with something, be afraid of boring all of them.

A nice example for this is the casino dungeon boss in Persona 5. 
She has a roulette/random chance based minigame during one of her boss forms, but the player always gets the choice of the safe bet or the high risk bet, and are given a round to prepare for the outcome. 
So there is randomness, but also choice in how much you want, and a chance to prepare for the result. She doesn't just randomly leech 2/3 of your party's HP.

How I did this in my game ?
For melee characters : Seeing as I avoid missing in combat, unless an effect is specifically causing it, I play with damage range and crit chance/Damage. So one weapon might doe between 90-110% of its damage, while the other might do between 50-150% Damage. Or maybe a clown punch that either does 0 damage or a devastating critical.
For Casters: the lightning element is almost completely random in its targetting (It will still only target enemies), and has an enormous spread of dealing 20-200% of its damage, with a critical damage of +200% (vs the normal +50%). If you want that , great, if you don't, then don't pick lightning.


2. Have randomness set the stage, but don't let it end it.


Yes, the cards you draw in magic are random, but everything after that is up to you. Do you mulligan or not, do you play aggressively or defensively,... 

How to apply this to an RPG ?

Back attacks/Ambushes/... However you want to call them. They can be fun once in a while, but they should never place me in a situation where I am dead without any input from me, if I am at full health. No encounter should be able to kill me completely without my input. 
If ultima slime can take away 55 % of my entire teams HP in one ultima, then under no circumstance should he be able to use that move twice before I have a go at him, even if it is to use an escape item. Anything less is insulting my autonomy as a player.
I'm fine with being half dead after an ambush, giving me a shitty starting position, but I should still be able to fight/escape my way out of it, no matter how much damage the enemies are doing.

This I applied by making it sure that no enemy troop can actually kill the players in one turn from full health. Not by powering them down , but by simple tweaks in the AI.
For example, suppose I do have my 2 ultima slimes in one battle, then one is going to cast ultima, and the other casts something like shell or haste, or uses a single target move like flare (FF  examples used, but you get the gist).
This still leaves me in a shit position, but there is still a fight, and more importantly, choices.

3.The mulligan.


Having randomness based mechanics not go our way leads to us wanting a way for us to do this over. And thus the save scum was born.
Oops my lockpick failed? Reload.
I didn't get the item I wanted from this boss/chest? Reload.
The boss started with spamming its supermove twice and I don't feel like dealing with that ? Reload.
I failed the minigame? Reload.

How you could solve this:

Remove mechanical randomness outside of combat, or heavily restrict saving. In this day and age restricting saving is likely to backfire, as most people react negatively to that.
Just have your lockpick skill be all or nothing (Lv3 Lockpick opens lv 3 locks). Anything else leads to the player reloading. Or maybe just have the lockpick break, but the lock still opening with a random chance. The player got a small penalty, but probably won't feel the urge to reload. Have them fail forwards, do not brickwall them.

If there is a big reward for doing well at the minigame, do not gate it after a single attempt, either allow a retry, or even better have some kind of point system so the player can "save up" for the item. 
Nothing like a good item being stuck between a (semi-)random minigame hey guys ? (Shakes fist angrily at chocobo minigame for Tidus' ultimate weapon).

Now in combat, how do we factor this in ? Maybe the player can bounce back from a devastating roll of the AI dice using that mega potion he had been saving ? (On the hoarding of powerfull items, and the "I might need this later effect" a lot could also be written.), but if the item needed to bounce back is actually too valuable, we once again have our player reaching for that reset/reload button.
Persona 5 had a rather nice solution in having one use escape items that where both pretty expensive/annoying to craft, but had no other use then to escape, giving the player a couple of fail-safes, next to an escape skill that could actually be upgraded to work better/in more situations(when surrounded for example). FFX solved this with having a character with a 100% succes flee skill. As long as he lived, you could always get away.
This also leads to my other point : escape has to work reliably enough or should not be there. "But won't players use that to skip some of the more difficult encounters?" Then you have a problem with the encounter not being fun/rewarding enough, not with the escape function. Using escape already has a penalty: no xp, no gold and the nagging feeling that you failed as a player.

But personally I find these patches on an underlying problem:  your game balance is either out of wack, your enemy AI is terrible, or your players do not have enough tools to deal with varying situations.
Small aside : Your AI is not there to kill the player, it is there to provide an engaging challenge, they are there to lose in a fun way. Yes, it might be logical for every enemy to attack the character with the lowest defence and HP, quickly reducing it to a pulp, but that gets old fast and leads to weird things. 
No one wants to be in the situation where raising the tanks defence does nothing because everyone just keeps attacking the mage anyway.(Cover and guard systems not included).
The game designer(Game leader in TableTop RPG's) is not there to kill the player, but to entertain/challenge him. He can always kill the player, but a true challenge ? that takes Finesse and skill. (A little nugget from my days as a D&D dungeon master, which coincidentally has a lot in common with being a game designer).

Warning : Controversy ahead:

In short : each time your player reaches for the reload button bacause of something he had no control over you have failed a tiny bit as a game designer.

Now, to coat this in some context: Nobody is saying every combat has to be easy. Just that in each combat I should be able to win/esape with the tools given to me, no matter what the RNG says.


4. Players suck at probability.


Now, there are actually studies that show this. We both over and under estimate probabilities.

We feel that everything above roughly 75 % should always happen (if it's us), and anything below 25 % can never happen (if it's the opponent), except when it's an enemy, then it seems like the 75 % skill hits way too often. It is why skills with "75 % accuracy" that actually have 75 % accuracy feel wrong. 
We feel like either the game is lying to us , or cheating us. Once again , this is player perception and is in very broad strokes.

How to solve this ? There is no simple and ethical solution to this. Human nature is what it is, and fighting that in a game is an uphill battle.
We could lie to our players ( saying that the accuracy of our skill is 75 %, but actually giving it a 90 % accuracy in the hands of the player and a 60 % accuracy in the hands of the enemy.
This might work, untill your player looks at the code. This also has the problem of confirming your player bias, and thus deepening the problem.

Another problem is once again the percieved dependance of the events. If we just missed twice, we feel as if the next one shouldn't miss, even when the accuracy is 50 %, and 3 misses in a row happen quite often.
Once again, do we cheat, and give the player a hidden accuracy boost after a (string of) misses? Once again , this might work until your players find out. Once they do , 2 probabilities : they exploit it, or they get out the pitchforks and torches.

There is however a third option, that is rooted in the players penalty avoiding nature. As humans we dislike penalties more then we like rewards. 
It's the wow rested Xp problem. In early playtesting of WOW, they experimented with a "tired" mechanic, where you would get -50 % xp if you played too much. Players where livid.
How did they solve this ? They changed Normal to be called "rested" and gave a 100 % xp boost, while  the new normal (the old tired) gave regular XP. They then cut the gained xp in half, so you got the old 100 % xp when rested, but only 50 % when "Normal". 
So they actually changed nothing, but they worded it as a reward, and not a penalty. That made all the difference.
Another facet is the the same reason why you would rather buy an artificially overpriced "80$" pair of yeans at 40$ then a "40$" pair at 40$ (the JCpenny effect, look it up!), even if the pants are both only actually worth 40$, we feel like we got rewarded, when actually we got exactly what we where going to get anyway.


So in short:Random rewards not penalties, make it feel like the player got something extra, even if the end result is the same.

A miss feels worse then a crit feels good.


An in RPG example:


Say we have Lucky strike, a move that hits 50 % of the time, but for triple damage. Obviously, this is mathematically speaking a good deal, you get on average 50 % more damage, but still, a large fraction of your player base will avoid this skill.
Now maybe, let's look at lucky strike 2.0: It always hits, but has a 50 % Chance of doing double damage. This skill just feels better, and I bet a lot more players will pick this one, even if the chance was only 33 % or 25 %.(Which they will stil unconsiously feel like it should happen 50 % of the time, but you can't win them all).

Another small suggestion could be to not have misses, or to have a miss deal some small amount of damage. Rename miss into weak hit and rename hit into solid hit.


Anyway, there is  still a lot to be said about luck, and how players handle it, but I'm already going long as it is. I hope I gave everyone ome ideas for their own games.


If you feel like I am wrong and an utter idiot, feel free to comment. As always, discussion is a door to new insights, and as some would have said it:" An evening in which everyone agrees, is a dull evening"


Recommended Comments

Myself, I like roguelikes. And if you don't know that much about roguelikes, they are basically randomness incarnate. They shift around the game based on the whims of our lord RNGesus. Not just little things but map layouts and such are randomly generated every time you play. Their preferred solution to save scumming is usually to just flatly disallow the player from freely saving and loading whenever they wish. You die, oops, either start over from the beginning or at least lose a ton of progress. Yet since the experience is at least slightly different every time you play it isn't as boring as it would be if you had to play the same thing over and over.


However I that how they use randomness is a bit different to what you are talking about. Oh sure, you often roll dice for things like if your attack hits and how much damage it does, but it's so much more focused on long term consequences then it is short term ones. Not that their won't be cases where RNGesus will screw you over in some unfair way, but for the most part I think the best roguelikes simply don't have that much instant death from unlucky rolls. There is a reason 'Yet Another Stupid Death' is a common saying with roguelike fans, because most death in Roguelikes are either the result of not managing resources very well, or overconfidence.


That's the thing about randomness. In lots of RPGs randomness is almost pointless, because it doesn't really have any impact. If you can just reload all the time, or if it's something you can easily repeat until you get lucky, it's just kind of filler. Or it has too much impact all at once and you have things that just require a roll to succeed or fail to progress. But the more you randomize in slight ways, and the more the player is able to deal with any given result, the more they will build on each other in the long term and build a unique experience.


Another way to look at it is: What's the difference between being given 500 lockpicks that only have a 1 in 500 chance of working or only having a 1 in 500 chance of finding a lockpick, but it always works? The player is more likely to be frustrated and save scum with the former when they find a locked door and will have to waste time trying all 500 picks. The later on the other hand means the player knows right away if they have a tool needed to open a door and it's harder to save scum.

Edited by Kayzee

Share this comment

Link to comment

Aah, the roguelike. I agree with you there, the randomness is part of the fun there.


But a good roguelike uses the randomness to set the stage, not to determine the result, as you say.


On the lockpick issue : why have breakable lockpicks at all ? Just say that this is a lockpicking lv 5 or higher lock. Have it depend on the skills of the characters. You didn't buy the lockpick skill ? No treasure for you! Breakable lockpicks map to one of 2 problems : either they are too bountiful/cheap and might as well be infinite, or they are so rare as to never be used (the megalixir conundrum), or people start saving before opening a chest seeing what's inside and reload if they don't like what's inside.

Share this comment

Link to comment

Hmmm... I don't know if I made one of my major points very well. A good Roguelike does more with randomness then just set the stage I think. It would be silly to say randomness has no real impact on the result of roguelikes. It wouldn't be a very good roguelike if there isn't a fair bit of luck involved if you ask me. It's just that the result in a lot of roguelikes is influenced by a lot of small dice rolls and little decisions that smooth each other out over time. Can you get a run of incredibly bad luck where you never get a break and get frustratingly worn down? It does happen in a lot of roguelikes, more then many people would like. But most often I find that good and bad luck tends to balance each other out.


The whole point of randomness in roguelikes is to prevent you from winning by pure rote memorization and actually force you to think about how to deal with problems. Running into a spot of incredibly bad luck does happen, but usually this is balanced by coming across some amazing item or something, some tool that can help you deal with bad luck. It often balances out to make any given dice roll not a big issue because, well, the game is rolling dice all the time for everything. It's the bell curve in action: When you roll more dice, the combined result tends to be closer to the average.


Also, it makes a lot more sense to have breakable lockpicks in a roguelike then it does in a normal RPG. Because it's not about having a roadblock for the player, it's about giving you an option for how to deal with a particular problem that you may come across. But a good roguelike will have multiple ways to deal with the problem besides that, and the best will have a tool be a solution for more then one problem depending on how the player applies it. A classic example is throwing potions at things. Nethack in particular is great at this, so many of the objects you come across may have an obvious use, but many more have hidden uses. Even useless looking items like blindfolds and towels might be useful in some situations. The 'megalixir conundrum' is also much less of an issue when you are often face to face with a situation where you have to either pull out all your trump cards or lose everything, which is a big reason permadeath is a thing in roguelikes.


I think the reason why there is such a big difference between how a roguelike and a normal RPG uses randomness is that for roguelikes randomness is a core gameplay element that everything is carefully designed around. In a normal RPG not so much. Any given mechanic may have a randomized element, but few really reinforce or have much to do with each other. I mean, what is the point, for example, of random encounters? If you can just walk around till you fight one, why not have them show up as events and save everyone the trouble? What about monsters randomly dropping stuff? Most of the drops are either trash the player will never wanna bother with or super ultra rare items that will just be farmed. Why not just have them drop stuff all the time or after x amount of kills?



Edited by Kayzee

Share this comment

Link to comment

The main difference I see between rogue likes and standard RPG is the length of the concequence.


Most RPG's are designed to be played all the way through in one go, so a random screwing over (last lock-pick breaking, having to use that super item) has a way longer consequence than in in a rogue-like, and to avoid that, we will often just reload ( which is what I am trying to prevent here). 


You fail in a rogue-like ? Well, better luck next time. If you fail in a regular RPG and have to burn through a valuable item, or yyou didn't get that super item from the locked chest, well, your save file is now forever a little worse because you don't have that super item.



Share this comment

Link to comment

You know, it seems kind of odd to put it like that. I mean, being able to save and load means the player can avoid the consequences of their bad luck or bad decisions. It seems to me like there is little reason to not hoard items and reload when something goes wrong if you are able to. One of the reason some RPGs do the whole 'save point' thing is to make sure there is a time investment between safe areas and can't just scum every encounter or something. If there are no meaningful consequences for failure random elements seem more like annoyances that waste your time. 


Plus, many roguelikes are designed so you have to beat them in one run, and they can sometimes get pretty long. I mean it can definitely be frustrating when you die far into the game and have to start over and I think I prefer it when any given session doesn't take more then an hour or two, but there are definitely roguelikes where you are going to be spending a while on a run. Look at Tales of Maj'eyal for example, it's crazy long. Though you can play it in a mode where you have multiple lives or even unlimited lives, so eh.

Edited by Kayzee

Share this comment

Link to comment

What I am advocating is the elimination of the need to save scum. If a challenge can be beaten by save scumming alone, it is not a challenge and should be cut.  If I can only beat the boss if he does not use his super move, then why is that super move there ?


Certainly if I fail as a player, sure punish me, but if the game decides to go Back attack => ultima => ultima =>I'm dead! then no, I should not suffer permanent consequences for that.


I have the same problem with stealing from bosses if it is random. The only thing this does is make me hit steal for an indeterminate number of rounds, and reload the boss if I did not steal the ultima sword by the time I beat him.


A point against save points : If your game is playable on mobile, you have to account that the player might only be able to play for 10-15 minutes.


Some of these things might be fine in roguelikes, nut not every RPG has to be that.

Share this comment

Link to comment

Yeah, I am not saying all RPGs should be roguelikes or have limited saves something, I am just saying that if it's possible to save scum away the bad rolls you get, well, wouldn't most player's first instinct be to do it? I mean, what is really preventing them from doing that? It's a strategy with very little downside besides it taking a bit of time. You can take this to an extreme of course. I think some designers love to encourage that kind of thinking because it makes the game 'last longer'. Every time a player needs to replay something or repeat an action it's extra time spent playing in some people's eyes. But it's really just a Skinner box in the end. Press the button again, you might get the cheese this time! :P


But my question is this: If the randomness in a game is not meant to just waste the player's time, and I expect for you it is not, what is it for? Why have random elements at all? In most roguelikes the randomness serves a function to the gameplay, to force the player to think on their feet and make them unable to rely on rote memorization. In a card game like poker the randomness serves a function to the gameplay, because the whole game revolves around figuring out your odds and reading the players, knowing when to bluff or call a bluff. But it seems to me there is a bit of a mishmash for a lot of typical RPGs, especially more linear story-based ones where you are expected to be able to freely save and load anytime. It seems to me like that makes the random elements kinda pointless.


It's not like you have to have randomness to make good RPG gameplay. You can just use flat formulas for most everything, and just balance it in a way that doesn't involve chance at all. I like randomness as a mechanic, but only when it actually has a point. I just wonder if many games just kinda throw it in without thinking about the larger picture of the game ya know?



Share this comment

Link to comment

It seems we are in agreement. Big randomness has no place in most story driven RPG's, because each time we reoload, we kind of break the story a bit. 

Small randomness, sure, but the player always needs to feel as if he has a fighting chance besides reloading. That's why I am a fan of medium term consequences. He has a bad encounter and has to spend more MP then usual ? That's a bit sad, but as MP regenerates slowly in my game, he can just be a bit cheap on MP the next 2-3 combats and be back where he was. He has consequence, but he isn't royally screwed. 


What I am trying to say is that the difference between the all-nuts situation and the total screwover shouldn't be as big. Getting a high potion or ether from a chest isn't a reload offence, but getting a potion or a megalixir, that might be reload worthy.


It surely makes balancing a nightmare. Do I balance this fight assuming the player did or did not steal that ultima sword from the last boss? Bt that is a whole other bucket full of problems.

Share this comment

Link to comment

The question for me is if you can still have randomness be meaningful like that though. I know that playing with odds is a classic gameplay mechanic, but I can't help but think that many RPGs might be better if they were less like poker and more like chess. A fully deterministic RPG would be interesting I think. I rather see RPGs punish bad decisions, not bad luck. It makes more sense to me to use saved games for trying different strategies rather then doing the same thing and hopping it works this time.

Share this comment

Link to comment

There can still be some randomness in small ways: will this move do 100 or 120 damage? Or in enemy behavior: will he cast ice or fire next turn ? And on who ? Randomness is a spice, not a main ingredient.

Share this comment

Link to comment

Well to be fair, wouldn't that depend on the dish you are making? There is nothing wrong with a little sprinkling of randomness, just like there is nothing wrong with spicy food. But it's hard to argue the spice isn't a main ingredient in curry, and it also seems weird to add spicy curry powder to, say, ice cream. Even if I love my spicy food, putting it into everything just doesn't work I think. But there are many foods where a little spice can be a nice extra like you said, like red pepper on pizza. On the other hand, I am perfectly fine eating pizza without red pepper. I tend to prefer garlic salt.  Does that also count in a way? I donno, this metaphor has run away from me and now I am hungry.


Anyway, I am not trying to say that you should make your game without randomness in it, it's your game after all. I just thought it was a fun little thought experiment to think about. Like what would a RPG without any randomness at all really be like? How would you balance it? How would you make sure it's still fun? I don't think in the end it would be all that different to be honest. I think the Paper Mario games might be kinda like that? Not sure.


Heck if you wanted to you could try and cheat and use some kind of seeded psudo-randomness in such a way that, yes it's effectively random what rolls you get, but the rolls have all be made for you already and you are stuck with them. Even reloading will only get you the same results. I think the Fire Emblem series does that maybe? I donno. There are often ways to cheese the system if you are not carful, but I think it's a nice way of solving the save scumming problem.

Share this comment

Link to comment