Dynamic difficulty adjustment

[Phil Steinmeyer]

I was considering the idea of dynamic difficulty adjustment for my current game.

Basically, I would adjust the difficulty of a level (which, for my game, generally means how much time/turns the player gets to complete a level) based on how much time/turns they had left over when they finished the prior level. If a player is blowing through levels using only half their allotted time/turns, I'll reduce time/turns for the next level (not by 50%, but perhaps by 15% or so).

The downside is that an observant player would notice this - as replaying parts would yield different amounts of time/turns for the same level, based on how they'd been doing.

Thoughts?

[Nexic]

A difficulty system like this would seem to be a good idea, but like you say it may become obvious to a lot people. If you could find a way to do it invisibly (by changing statistics that the user cannot see) I think it would work much better. For example:

Player tries hard to complete level and fails.
Tries again and wins, but then realises the game just made it easier for them.
Player gets no sense of achievement and might be frustrated by this.

If they couldn't see the adjustment they would feel achievement, which is good :)

[papillon]

Hate dynamic adjustment. Want to be able to choose my own difficulty setting. Sometimes, I want things to be easy. Sometimes, I want them to be hard. Don't want game to cheat - at least, not very much. "Last Chance" style cheating - delaying the end just a few seconds past when you thought the end was going to be, so you can JUST scrape by - is okay.

[Hamumu]

For me, I think it's only a good thing in the sense of making it easier if I lose. There were some bits in Jak & Daxter where I died 10 times in a row, and actually had the very clear thought of "why don't they just make the platforms slower after 3 deaths or something?" You know someone is frustrated if they repeatedly lose on the same level, so it's definitely a good plan to do something about it.

Aside from that specific 'cheating' on the behalf of a clearly failing player, I think it's best to stick with choosable difficulty, so Papillon can choose whether to play an easy or hard game today!

[Musenik]

Hate dynamic adjustment. Want to be able to choose my own difficulty setting.

Respectfully, you are not the market. Dynamic leveling is a troublesome tool, but I think it can work. One way I've investigated is to simply back down the list of levels while calling it the same level. (only works if levels aren't hard-coded.)

That said, my personal distaste for dynamic leveling comes from the awesome 'close combat' series by Atomic games. I remember repulsing the Allies from the beaches time and again. Wave after wave of increasingly stronger attacks all failed before my awesome, but dwindling defense. I was proud of my victories, only to be told to "withdraw for strategic reasons". What no reward for kicking ass? All I needed was a few reinforcements and I would have KIA'd the enemy all the way back to London!

[undersan]

As papillon and Hamumu touched on, there are really two variables here:
- The player's skill at your game
- The player's tolerance (desire?) for frustration

To figure out the second one, you could ask the player to choose a difficulty level at the start. Or, for casual games, maybe you can just assume that the player's tolerance for frustration will be "low".

In either case, you're establishing a "target" level of frustration that your dynamic difficulty system can aim for.

[James C. Smith]

I like the idea of offering several levels of difficulty and then helping the player decide.

Pop-up at end of level when player barley finished in time when they “lost” the level for the second time “You seem to be having trouble. Would you like to switch to an easier setting now?”

[Diragor]

I'm reminded of one case where I think this was done pretty well - Crash Bandicoot and the Wrath of Cortex (and I think the other Crash games, too). There's this mask thing you can pick up throughout the game that basically gives you a free "hit" - something hits you and the mask goes away rather than killing you. After three (or so, I don't recall exactly) failures in the same section of a level it'll start you off with that mask to make it easier. It's like a free powerup as a temporary boost to get you through a specific spot rather than altering the game to make it easier overall. It didn't help in all situations (like platform jumping sections) but it was a nice help sometimes.

[MrCranky]

I'm with James on this one - allow the user to choose.

Not only does adjusting automatically rob the player of any sense of achievement, it can also destroy any sense of feedback learning. If a player doesn't know why sometimes they do well, and sometimes they do badly, then they can't isolate good strategies from bad.

If the player is completely unaware that the difficulty is changing, then maybe its acceptable; but if they ever catch you doing it, it'll rob them of any faith they have in the game. I'd say only consider doing it if you're absolutely sure that your system is perfect. I can't imagine anything worse than a game that oscillates from too easy to too hard, frustrating the player when its difficult, and robbing them of any feeling of achievement when its too easy.

If they choose the difficulty level themselves, they can avoid frustration when it comes up, and still feel rewarded. Of course, this works best when you can change difficulty mid-way through the game; if you can only choose at the start of the game, they can still feel frustrated (as they need to restart at a lower difficulty level to get past the tough bit).

[Black Hydra]

For about 90% of games DDA is always better overall. Sure some hardcore gamers might not like that they can't have as much control, but these people aren't are market. I'm sure some of these same people would like to adjust the screen resolution and color depth and have access to debug commands. Most people don't.

If it is obvious then it won't work. As long as the variables controlling it are not discrete or visible to the player then it most players shouldn't notice it at all.

Selecting a difficulty mode is a good idea, but I dislike this idea. Because until players play until a point at which they probably have already seen most of the content, they really don't know what difficulty is best for themselves.

Using a system like DiabloII where the later difficulties had to be "unlocked" by completing previous ones seems like a legitimate measure to expand content.

Using a EASY, NORMAL and HARD mode as an alternative to DDA is, in most cases, laziness. Now, you might not be able to find time to squeeze DDA into your project, but don't tell me it is because people like having EASY, NORMAL and HARD modes better. Perhaps in addition to DDA but it is no replacement of it...

[mot]

Here is an interesting article on the subject:

Game Matters: "Auto-dynamic difficulty"
http://dukenukem.typepad.com/game_matters/2004/01/autoadjusting_g.html

[Black Hydra]

@mot - yes, I read that article earlier (I'm too lazy to look for links...). It's nice to see my opinions backed up by someone who knows what they are talking about.

[Anthony Flack]

don't tell me it is because people like having EASY, NORMAL and HARD modes better. Perhaps in addition to DDA but it is no replacement of it...

I like having EASY, NORMAL and HARD modes better. I hate DDA. Hate it.

You might say, "ah, but you're not the market", but actually, I am. I'm going with my instinct and doing things that I like. Is that the best strategy for maximising profit? Maybe, maybe not. But I'm not trying to maximise profit, I'm trying to make a good game. And it's not laziness.

[Anthony Flack]

Something else to consider - if DDA is such a good thing, why does everybody stress how important it is that the player never ever realises you're doing it?

[Black Hydra]

People say they hate DDA but in reality I'm sure most people would prefer a game that used it if they didn't know.

There are three ways to describe a game too hard, too easy and challenging. When the game is challenging it is fun. Little areas of ease or great difficulty can add to the feeling of a game, but trying to maintain that level of optimal playing difficulty should be desired.

Using EASY, NORMAL and HARD modes is forcing your player to fit inside your predefined difficulty settings. DDA forces your game to adapt to your player, not the other way around.

Now, I'm not saying EASY, NORMAL and HARD modes are bad, but I hate people seeing this as a form of play balancing. If you used them in addition to DDA or had them as unlockable features then I would think they were a great addition.

To use your own game, Platypus as an example, Ant:

A friend of mine (a very good game player) played the demo and decided to start it off on normal mode. He played through the first level but he lost before he got to the first boss. Sure he could have switched it down to EASY mode and start over, but he just quit there and left. I started on EASY mode and had no trouble beating the boss.

Now, Platypus isn't the best example because you can continue onwards without ever having to retrace (assuming you're fine losing all your points), but had you lost causing you to have to do the entire demo over, I'm sure many people might just leave it at that.

I believe that DDA can really increase your margin of players who will find the challenge level right for a game. Of course, if you tell them it has DDA they probably will have mixed feelings because DDA works subconciously. If you ask a person if they won't buy a game if it has so-so graphics, they probably say they don't care that much, but it obviously has a sub-concious impact on play.

You bring up some good points though, so I'm sure there are alternatives to DDA, but I just don't see the three preset difficulty mode as a very suitable one...

[arcadetown]

The Black Knight (http://www.arcadetown.com/blackknight/gameinfo.asp) does DDA and it is by far my most popular game.

The amount a player must score to beat the current level is based on the best scores made during previous levels, plus a little extra. If a newbie plays it, the scoring requirement rachets up slowly allowing them to enjoy a few levels and get into it. While for advanced players it rachets up a lot faster, still giving them a good challenge.

[arcadetown]

I like having EASY, NORMAL and HARD modes better. I hate DDA. Hate it.Anthony, based on the later levels in Platypus that Mike didn't adjust difficulty in to much, I'm guessing you're preference is more like HARD, HARDER, and HARDEST. :rolleyes:

[Anthony Flack]

Sorry, Platypus is a dreadful example of play balancing. (although I would say my preference is more like PATHETIC SO STOP YER WHINGING, CHALLENGING, and GOOD GOD YOU'RE A MAD ONE).

I'm sure there are many other, better ways to handle this sort of thing - offering to lower the difficulty might be a good one. But part of what I like about EASY, MEDIUM and HARD is that they are very concrete things that you can test your skill against, and accurately guage your progress. Whereas DDA gives you a game that you just can't pin down. And then there's that whole business of getting better at the game not giving better results... plus the fact that DDA makes all kinds of assumptions about how hard I want it to be.

Some people don't want the game to get harder if they're doing really well. And as for me, if I've just died 10 times in the same spot, and it doesn't feel unfair, then that's when I'll be going "well all RIGHT, here we go now" - the last thing I want is for the game to auto-wimp on me.

[soniCron]

There's no reason not to use both. If you use the player difficulty setting to influence the DDA (or auto-dynamic difficulty, "ADD"), then you have the best of both worlds. In that manner, the difficulty will always reflect the player's chosen difficulty setting accurately, rather than the developer's arbitrary judgment of what's easy and what's hard, which may significantly vary from person to person.

The only problem I can see with that is the stigma attached to a player selected difficulty. Often players will chose "Easy," not because they really want it easy, but because they don't want it hard. If the combination of player selected difficulty and dynamically adjusted difficulty is used, the player's selected difficulty will be accurately used during play - something the player may not expect, and may ruin the game.

To that I say, postpone the player's difficulty decision until a little way into the game, at which point, prompt the player to choose between the difficulties. In addition, instruct the player at this point that s/he may change the difficulty in the options menu at any time - something that's obfuscated from a new player, but an experienced player would keenly access. In this manner, the player will have the option of choosing their desired difficulty after having played the game for some time, and if they ever start a new game they know where to go immediately to choose the difficulty, or should they choose to do so during the course of normal play, they may.

[Rainer Deyke]

DDA leads to the meta-game of "play as badly as possible without quite losing". Which may be interesting meta-game, but if that's the game you want your players to play, why not make it explicit?

Basically I think that DDA is a horrible idea. There are much better ways for games to adjust to the player's playing skill. For example, why not give the player multiple goals to shoot for? Beginners reach the first goal and are satisfied. Experts aim for the highest goal. Both get a game as challenging as they want, all without punishing the player for being too good.

[Erik Thon]

Hi,

Here's what I think: DDA is a horrid thing to use in games where skill and replayability is in focus, like side-scrolling shooters and arcade-style games in general. People play these games for the challenge itself and to work up a level of skill high enough to take them through and that's where the fun is at. Use DDA in any significant degree and it kills the appeal those games have, hidden or not. Gamers quickly notice it when they never seem to get much better, while their kid sister gets almost to the same point they're at by bashing buttons at random.

[Black Hydra]

I suppose my opinion stems from a different paradigm about gameplay itself.

I think some people have a sort of adversarial idea of Player VS Game. The goal of the game is for the player to have fun, not as a test. The only reason you challenge the player at all is because challenges are fun.

The fact is, players are not going to give your game "second chances" if they aren't immediately drawn in. In retail you might be able to get away with this because they will have already bought the game so might be more willing to put up with some initial frustration. But especially with shareware games, if the player finds the game to be unplayable then they aren't going to be bothered to go back and twiddle with the difficulties. And what happens when they want something in between? Some games the easy setting is a bit too easy and normal is a tad too hard. Or perhaps hard is too difficult but normal is too easy.

If I was playing a game and I could notice DDA then yes it would probably diminish the fun. I wouldn't feel as much satisfaction when I win. But if I don't notice it at all, then the satisfaction isn't diminished.

So with a virtually invisible DDA system players will still feel the same sense of satisfaction however you won't have players who get "stuck" or cannot beat it.

Again, if players want the game to be harder for themselves then they could select a HARD difficulty. But with DDA instead of HARD being set to the developers standards, it is HARD for the player. Isn't that the whole point of using words like HARD and EASY?

[Robert Cummings]

I think most of you are making a lot of noise and forgetting the real point here.

The point is to make something that gives people pleasure. If a dynamic difficulty eases off the cpu's attacks just slightly when you're desperate it can really, really enhance the enjoyment.

That is the whole point.

There is nothing stopping this system being in place -with- 3 tier difficulty settings. You've got your Beginner, Intermediate and Advanced, or Easy, Medium and Hard but with difficulty load balancing on all three.

MY take on this is: don't think of it as a cop-out but something that enhances the game. When the hero runs and jumps and THEN everything blows up, it's just so much more fun than an ultra accurate timer that kills the player 1 pixel from finish, or blows up half an hour later.

You want to think in terms of a flexible game, where the game is the entertainer.

[DGuy]

The old game Gods from the Bitmap Brothers, would randomly drop health at fixed locations depending on how much damage you had taken up until that point. The poorer you did the more extra health you saw, the better, the less.

This worked well in that game because it didn't have a steady stream of enemies coming at you, but rather each enemy encounter was it's own little "scene": these scenes where strung together with short patches of down-time between each one (think Half-Life) and that made up each game level.

The emphasis was "get through each scene", not "get through each scene with enough health, else you'll have to repeat it because you'll die in the next scene".

I remember it kept the frustration level low and the game moving.

[Robert Cummings]

Thats nice, it's keeping things moving. I like that. You should not have to predict what you need for the next level.

Because you end up hoarding all the best stuff then suddenly you've completed it - where is the fun in that?

[arcadetown]

The fact is, players are not going to give your game "second chances" ... especially with shareware games, if the player finds the game to be unplayable then they aren't going to be bothered to go back and twiddle with the difficulties.
Great point. I've seen quite a few games that suffer from this and are great exit button click generators. JetsNGuns could have hugely used DDA as it starts off way too hard and never steps down, thus your normal user can't enjoy it. Once a user clicks the exit button in frustration you can forget about a sale, plus you can forget about popularity and word of mouth.

DDA was a huge part of the continued success of The Black Knight, sorta like another hugely successful game Gold Miner. Don't be confused, the game is challenging. It's carefully balanced so a newbie can play a few levels and enjoy it causing repeat plays, while an experienced player can play for numerous levels and compete for high scores.

[Anthony Flack]

I think some people have a sort of adversarial idea of Player VS Game. The goal of the game is for the player to have fun, not as a test. The only reason you challenge the player at all is because challenges are fun.

Yes, challenges are fun, and cheating at games usually isn't. DDA, if handled badly, is almost like auto-cheating. My point is that there is value in the game offering a rigid set of parameters to measure yourself against. Joggers like to time themselves on a run, but it wouldn't make much sense if their stopwatch went slower when they were feeling tired.

Limited use of DDA I could see as helping things along - a little bit of "oh, you very nearly got there, I'll let you get away with it". But to leave the difficulty in constant flux is not something that I can see working well most of the time. And I wouldn't enjoy it. And I think there are other, better ways (like actually giving people a choice at the end of each stage - "hard path or easy path?"). And games like Diablo where the mechanism is built in to the gameplay itself, naturally.

Everyone is taking great pains to make sure the player must never notice these tricks going on - because if they knew it, they wouldn't like it. Tread carefully. There are many people who think the catch-up mechanisms in Mario Kart 64 totally ruined that game.

And let's not forget, juggling and balancing gameplay elements is a tricky business at the best of times, even in a static game. To put all of this in the hands of some algorithms, even clever algorithms... well... it's all very well to talk about it, but who here is capable of making a game that is fantastically fun, well-balanced and dynamically adjusting?

[Black Hydra]

Some good points. If the players feel that the game is constantly changing its standards it will feel random which is definitely NOT fun. You have to choose variables that are incredibly subtle to difficulty so that the player never knows you've modified the game for them. Having other racers just jump to your position when your winning or having other racers stop when your losing would be incredibly unfulfilling so DDA can be ugly if done wrong.