Chade on 18/7/2010 at 22:41
Quote Posted by Papy
With your example, you basically start with a fixed difficulty level, and then adjust that difficulty to make the player who is better than that fixed level win with a less impressive margin, and you make the player who is worse than that fixed level lose in a less shameful way. But in the end, it is a fixed difficulty level ... (Actually, your example makes the AI unbeatable, the "75" should be lower, but I get your point)
Well, in a more realistic game, this would, of course, only be the "target" speed that the AI was attempting to go at. You'd still have all the standard mechanics where either you or the AI could make mistakes.
But yes, the "75" figure should also change over time or it isn't really adjusting to the player in the long term ... but maybe only to go up, or only change between races, or something like that. Either way, the player can't cheat.
(And yes, the AI is unbeatable, but I felt it was easier to explain that you could tie with the AI if you didn't cheat, or lose when you did cheat ... rather then beating the AI by some margin when you didn't cheat, or beat the AI by some lower margin if you cheated by some sufficiently small amount.)
Quote Posted by Papy
First, any dynamic difficulty level MUST have a delay before changing the difficultly level, either up or down. Also, those changes must be progressive and the delay must be large enough to make sure the player can't see it.
Second, all dynamic difficulty system start with a fixed difficulty which is independent of the player. That means the adaptive difficulty will HAVE TO become worse or better than the player in order to compensate for previous misjudgment from the dynamic system. Adjustment toward the player is not sufficient.
So now... what if the player try to hide his own skill in order to fool the dynamic difficulty system and suddenly raise his level of play near the end? Because of the necessary delay of reaction from the adaptive system, don't you think it will be an easy win for the player?
For practical reasons, I agree that a dynamic difficulty system will lag the player.
I don't agree that the difficulty system should compensate the player for previous misjudgements, and my skimming of the literature suggests that nobody else believes this either. And if you don't compensate the player for previous misjudgements, then the player can't necessarily cheat the system.
Finally, you have this huge implicit assumption that doesn't hold in many genres other then sport games ... that there is some externally imposed finish which the player reaches regardless of how the game plays out in the meantime, and all that matters is that the player is winning when that externally imposed ending occurs. In most games, this is not the case. The player accumulates rewards for good play right throughout the game, and often cannot complete the game until he/she is performing at some level.
Thirith on 19/7/2010 at 09:20
Quote Posted by Papy
For me, the question is : how do you make a game with adaptive difficulty rewarding? How to you push the player to do his best so he can end up satisfied with the game rather than bored?
As far as I'm concerned, I'd be okay with a system where you choose a level but the game adapts somewhat to how you play it; i.e. if Easy is at level 1, Medium at level 3 and Hard at level 5, and the game can dynamically go +/- 0.5, so you get a chance to 'grow into' the difficulty you've chosen, but if the difficulty is still off you have to change nevertheless. I guess you could even program the game to reduce how much the difficulty is adapted to the player over time.
Koki on 19/7/2010 at 10:53
Yes, it could also connect to your facebook account and read your "current mood" variable and adjust the difficulty based on that. Oh he's frustrated, better tone that enemies down somewhat!
But guess what? Just like dynamic difficulty, it's a retarded idea that requires huge amount of effort from the developer for little to no final gain.
Thirith on 19/7/2010 at 15:34
Quote Posted by Koki
But guess what? Just like dynamic difficulty, it's a retarded idea that requires huge amount of effort from the developer for little to no final gain.
Perhaps. Perhaps not. I just answered the question. And I'm afraid that just because you say an idea is 'retarded' doesn't automatically make it so.
DDL on 19/7/2010 at 16:03
Plus I'm not entirely sure how papy's idea of "playing carelessly to get an easy ride" would work in..say, a shooter. I mean, if you play carelessly, you'll simply die a lot. Yes, then the game might indeed give you an easier time, but you've still got to sit through dying and reloading a whole bunch of times. Sure, people might not always bring their 'a-game' to "generic corridor shooter X", but I suspect on balance your average person probably doesn't deliberately try to fuck up.
Dynamic difficulty is a great idea if implemented well: you generally want players to feel like they've just managed to succeed, so if you can spot points where the player in question will almost always just fail, and tweak back the difficultly ever so slightly, they'll still get that feeling of achievement.
It's just rarely implemented well.
Papy on 20/7/2010 at 02:01
Quote Posted by Chade
but maybe only to go up, or only change between races, or something like that. Either way, the player can't cheat.
What you are describing is pretty much what a lot of racing games do : you start with the easiest race, and you slowly go up the ladder each time you win. There were even some where the player would be relegated to a lower division if he was losing. That's a meritocracy difficulty system, not adaptive difficulty system.
Quote Posted by Chade
I felt it was easier to explain that you could tie with the AI if you didn't cheat, or lose when you did cheat ... rather then beating the AI by some margin when you didn't cheat, or beat the AI by some lower margin if you cheated by some sufficiently small amount.
I won't dispute the fact that the programmer can decide the outcome to be anything he wants with a "good" adaptive difficulty system. That's not a desirable goal for me..
Quote Posted by Chade
I don't agree that the difficulty system should compensate the player for previous misjudgements
No, I didn't mean compensate error from the player, but compensate misjudgment made by the adaptive difficulty system about the player's skill. That's the inevitable flaw that will allow the player to abuse the system and lead him to do think about a form of meta-gameplay.
Quote Posted by Chade
The player accumulates rewards for good play right throughout the game, and often cannot complete the game until he/she is performing at some level.
Good play means a more difficult game. That's punishment, not reward. Again, look at Oblivion to understand that a reward is NOT just a cool looking graphics or a fancy name.
As for not being able to complete the game until the player is performing at some level... Don't you think this would be a major failure from the adaptive difficulty system?
Quote Posted by DDL
Plus I'm not entirely sure how papy's idea of "playing carelessly to get an easy ride" would work in..say, a shooter. I mean, if you play carelessly, you'll simply die a lot. Yes, then the game might indeed give you an easier time, but you've still got to sit through dying and reloading a whole bunch of times.
That's pretty much how people play right now. The problem is everyone has a different threshold for what is the acceptable number of death/reload. For me, one death is a failure. For other, quick saving and reloading every 2 minutes is perfectly normal gameplay. That means using death as a way to judge the player skill is pretty much the best way to displease a lot of people.
Quote Posted by DDL
Sure, people might not always bring their 'a-game' to "generic corridor shooter X", but I suspect on balance your average person probably doesn't
deliberately try to fuck up.
If you tell the players that if they waste a lot of ammos then the game will give them more to make sure they always have enough... Do you think they will be as careful?
Quote Posted by DDL
Dynamic difficulty is a great idea if implemented well: you generally want players to feel like they've
just managed to succeed, so if you can spot points where the player in question will almost always
just fail, and tweak back the difficultly ever so slightly, they'll still get that feeling of achievement.
It's just rarely implemented well.
I agree, dynamic difficulty is great in theory. Can you give a single example of a game where it was implemented well?
For short games that you don't play much, this could work. For a longer game, the player will, sooner or later, realize that something is wrong. After a while, he'll just feel like in a theme park and lose interest in the game.
Koki on 20/7/2010 at 05:41
Quote Posted by Thirith
Perhaps. Perhaps not. I just answered the question. And I'm afraid that just because you say an idea is 'retarded' doesn't automatically make it so.
Ain't that a shame.
How about this then: For dynamic difficulty to work, the player must be completely unaware that it exists. Why? Because as soon as he knows it does, why try? Why bother? Why come up with elaborate strategies or do my best at dodging bullets? It's not required to progress in the game anymore. If I just die a lot, the game will make it easier for me and I'll progress anyway. All challenge is gone. Games without challenge are banal, shit, and boring. Oblivion was best example of that because it had RPG elements which amplified the effect tenfold. Why explore that ancient ruin? Why do that quest? The whole world was getting stronger as your character was*, so there was no point to do any of the normal RPGish stuff - gain levels and find phat loot. A bear could kill you as easy on level 1 as on level 30. Fight with last boss was as hard on level 1 as it was on level 30. Loot was scaled to your character as well, so there was no reward for navigating particularly hard dungeon(not that there were particularly hard dungeons anymore because everything scaled to your level), as at level 1 you'll find Sharped Stick of Minor Rash anyway, and not Great Greatsword of Great Pestilence.
Hell, I even have a personal example of dynamic difficulty failing from back when I was trying Left 4 Dead(The "AI Director" is a basically fancy name for dynamic difficulty). I reached one of the "press this button/yank this lever and hordest of Infected will spawn" moments. Problem was, right in the middle of the area there was a Witch. Not wanting to be raped by her and Infected at the same time, I decided to take her out before proceeding. Well, this didn't go so well and I was poisoned, something I noticed only after fighting off the Infected. I had no medkits and neither did the bots at this point; I knew it's now a race against time to the Safe Locker. But, sure enough, after not even a minute later in some random warehouse I find a single medkit lying inconspiciously on the shelf. Thanks AI Director! For a moment there it actually felt like a survival horror! Now we can go back to Serious Sam at Night.
So yeah you could tell the marketing to not mention anywhere that the game has dynamic difficulty. But how far would that take you? If I attack enemy base and get killed three times by five guys in a room, then on fourth attempt there's only three guys - well, okay, it's a bug or they ran away somewhere. But if that happens again, I'll be all waaaitaminute. Or someone will tell me before playing, or someone will dig it out of the files.
So you can say, well make it so there is a minimal difficulty below which the game cannot go. Okay, but then how is that any different from simply picking Easy at the New Game screen? Similarly, you need a limit from the top(or clusterfuck of Homeworld 2 proportions happens), and how is that different from picking Hard at the same screen. Depending on the game you might even give players an option to tone down difficulty during the game, Modern Warfare 2 did this.
And then you just go through the game and set it, spawn three enemies at Easy, four at Normal, five at Hard in Room 6. All done! Or leave the enemies and items alone and make simple cfg file which multiplies damage taken by 0.5 at Easy and 1.5 at Hard. Even less work! All better, easier and simpler than creating elaborate algorithms analyzing and calculating the player's "skill" in the background.
* - the Oblivion also fucked it up by scaling purely by character level. If you, say, put all the points into non-combat skills like mercantile, sneak and lockpick, you were completely fucked.
DDL on 20/7/2010 at 11:57
Hey! You can write long posts! :D
I guess my issue here is that the idea of simply "deliberately playing like a cretin because it doesn't fucking matter anymore" never occurs to me. Most people, I would hope, play games where difficulty is a factor to match their skill against some arbitrary opposition. I certainly do. You don't suddenly think "HAY WATE: DINAMIC DIFFICULTY LOLZ" and deliberately get yourself shot in the face fifteen times in a row so the next level has no bad guys. You still try and shoot those fucking bad guys.
Dynamic difficulty can work provided people don't try to deliberately game the system, and if people try to game the system, what the hell is wrong with those people? Those people are weird. Hah, well no: sweeping generalisations aside, I would suspect those people do not represent a majority of gamers. If you take the take to critically examine the core game mechanics to identify the maximally laziest way to complete the game, regardless of how silly it may seem, then you're not (to my mind) really playing games, as much as you're metagaming game systems.
The trick is to retain punishment, but minimise LOLZ U FAIL punishment. In your ammo example, it would be to provide MORE ammo if a ton has been expended, but still 'less than ideal amounts'. If you're careful about ammo expenditure, you can safely be assured of having near to full ammo most of the time. If you blow almost every round you have each encounter, you can be reasonably sure you'll find some ammo, but not enough to get close to full.
What difficulty do you play L4D on? Because despite your railings against the 'dynamic difficulty' in that, it does have difficulty settings. And at the highest level it simply amounts to
(if players==alive)
rapePlayers();
But again, it tries to skirt that limit of actually punishing the shit out of you. If you're doing really well, you'll probably have really tough fights, but that's ok: you're good. If you're doing really badly, you'll probably get horribly raped, but it will at least make you think you may still have a chance.
(plus you were lucky fucking francis didn't grab the medkit and use it before you got the chance)
And finally, remember, oblivion is not an example of dynamic difficulty. It's an example of static difficulty, pegged to character level. It is not more or less difficult depending on skill, skill is irrelevant: it is exactly as difficult at all levels. Dynamic difficulty in that setting would be "ohshi- the player keeps running into giant fucking daedra at level 2...let's spawn fewer giant fucking daedra". Given that you won't even see them until you're high enough to deal with them, that's moot. And not dynamic difficulty. In fact, in your example of someone getting to a high level having invested mostly in lockpick and speech skills, dynamic difficulty would be a positive boon.
In summary, I can understand your arguments, but find it unlikely that many people share the enthusiasm for 'tricking' dynamic difficulty. And yes, I will admit that beating something really tough in a game through repeated attempts, knowing that the game will not make it any easier on you, is ultimately more satisfying than having the difficulty adjusted down behind the scenes, but I will also freely admit that I would rather have succeeded after fewer attempts, and have a few extra hours of my life back.
On balance, my preference is definitely for static, but adjustable within play, difficulty, but I am not willing to simply write off dynamic difficulty as an inherently flawed idea.
Chade on 20/7/2010 at 23:59
Quote Posted by Papy
What you are describing is pretty much what a lot of racing games do : you start with the easiest race, and you slowly go up the ladder each time you win. There were even some where the player would be relegated to a lower division if he was losing. That's a meritocracy difficulty system, not adaptive difficulty system
...
If you tell the players that if they waste a lot of ammos then the game will give them more to make sure they always have enough... Do you think they will be as careful?
Haha, well
actually ...
In many ways, these points only go to show what sort of pointless distinctions get made between advanced adaptive AI, and game mechanics that have the exact same effect but aren't attached with the same labels.
The traditional "linear progression of races"
is a form of adaptive difficulty. It's just one that is so obvious that people don't notice it (Salvor Hardin quote goes here). And it illustrates one of the errors that you and Koki keep making: not recognising that challenge and reward are usually packaged together. People don't keep playing the same race again and again because it lets them win again and again ... they get bored and move on to bigger and better things. Usually as fast as they can.
Additionally, almost every single game I know of has an adaptive ammo system. The amount of ammo you can collect gets reduced to zero once you accumulate a sufficient amount, encouraging the player to play less efficiently so he can pick up ammo again. It's called an ammo cap, and it stops a player who finds the "ammo quantity" difficulty setting too easy from beating the "ammo quantity" game too easily. And again, it's so obvious that people don't notice it.
Of course, you can't package every adaptive algorithm into an easily explained gameplay mechanic. This is when you do something that you hope is subtle and won't be noticed, which is what you are objecting too. But let's be clear that basically every game ever made has mechanics to respond differently to the player depending on how well they perform, so that the player is always presented with an appropriate level of challenge. That is not an issue. The question is whether or not you can do it with mechanics that aren't integrated into the game's fiction.
Quote Posted by Papy
No, I didn't mean compensate error from the player, but compensate misjudgment made by the adaptive difficulty system about the player's skill. That's the inevitable flaw that will allow the player to abuse the system and lead him to do think about a form of meta-gameplay.
I knew what you meant, and you don't have to compensate the player for the difficulty systems earlier misjudgements. The objective is generally to improve the player's playing experience, not have a game that swings wildly from too easy to too hard just because that "balances out".
Quote Posted by Papy
I agree, dynamic difficulty is great in theory. Can you give a single example of a game where it was implemented well?
Well, I rarely see people talking about Max Payne's dynamic difficulty when they discuss the game. They might know it exists, but I don't know anyone who found it distracted them from playing the game.
Papy on 21/7/2010 at 02:22
Quote Posted by Chade
The traditional "linear progression of races"
is a form of adaptive difficulty.
No, it's not. With the "linear progression", if the player is not good enough, he will never be able to do the more difficult races. With "adaptive difficulty", the game will lower the overall difficulty to make sure the player will be able to participate in those "more difficult" races. Do you understand the difference?
In the same way, I'd really like to understand how you can come to the conclusion that an ammo cap is a form of adaptive difficulty.
Again, can you define what is an adaptive difficulty system for you in very specific terms? Because we obviously are not talking about the same thing.
Quote Posted by Chade
And it illustrates one of the errors that you and Koki keep making: not recognising that challenge and reward are usually packaged together.
Oh, but I do recognize that challenge and reward are usually packaged together. Did you already forget who I am and how I play games? But, from my point of view, you fail to understand what a "reward" is, and so you fail to see why a system which adapt to the player well enough to eliminate the need of an "easy, medium, hard" system, also make a lot of those possible rewards lose their value.
Quote Posted by Chade
But let's be clear that basically every game ever made has mechanics to respond differently to the player depending on how well they perform, so that the player is always presented with an appropriate level of challenge.
I'm sorry, but I thought the discussion was about adaptive difficulty, not about the general concept of having different difficulty within the same game.
Ok. Back to adaptive difficulty. I know the objective is to improve the player's playing experience. You don't have to convince me that this is a noble idea. My problem is that I believe this can't be done well. I believe an adaptive difficulty system has to be tailored to some particular player in order to work correctly. Different players will experience the same thing completely differently. Different players like different things. No matter what point of data you use, you will never find one which are common to everyone. You have to choose who is your average model, the same way the "medium setting" is tailored to an average model. In the end, adaptive difficulty is for me just another layer of complexity, but it doesn't really solve anything.
