next »

[Sam]

Games that match their difficulty to the ability of the player.


Often we see adaptive difficulty being accidentally implemented and acting inversely, with the game becoming easier for good players.  You've really mastered the controls for a platformer and manage to get up into a hidden area with a 1-UP.  Well done, you're good at the game!  Here, have some resources to make it even less challenging for you.

What we want is for the game to realise that it's dealing with a strong player and give them a suitable challenge, or to lend a hand to a less experienced or able player who might otherwise leave in frustration.  Let's discuss neat ways to do this!

Interestingly, whilst researching this stuff I found someone talk about hidden power-ups as an example of good adaptive difficulty:  They theorised that a player getting stuck on a part of a game may wander back through previously explored areas and manage to find the hidden power-ups, which will help them get past the next area.

Stuff I think doesn't work:
Bioshock makes you find nicer loot if you're low on money.  So the most effective way to play is to keep your money level just below the trigger point.  That is not a fun mechanic, and it doesn't make sense for it to be a concern of a confused man wandering an underwater city.  I'm sure many other games have a similar system, but this one came to mind.

Oblivion's choose-your-own difficulty slider.  I can deal with "easy, medium or hard," but how do I choose my difficulty on a 100-point scale? In addition I feel like I'm cheating if I move it once the game is underway (you can instantly kill pretty much anything if you set it to the lowest.)  Especially problematic as there's no real disadvantage to setting it to the easiest other than rather boring combat.  Of course, complaints about Oblivion's difficulty are cliché.

Stuff I think does work:
When you're at very low health in Crysis, enemies become vastly less accurate.  When you're at low health your vision is also blurred, filled with red, and your suit's computer is beeping and flashing at you (so it's not a viable tactic to get to low health then charge into battle.)  You manage to drag yourself to where you think there's some cover, and let your suit repair itself ready for your inevitable dramatic come-back.  This replaced a lot of "You died, press quick-load now" situations with dramatic firefights, both reducing frustration and increasing awesomeness.  Far Cry 2 not stealing this system left me very disappointed - especially as they have a whole set of cool "healing at low health" animations which I too rarely see.

The hardest fights in Final Fantasy games are typically only unlocked if you do some obscure and optional set of side quests.  This kind of falls into the trap of rewarding the best players with resources to lessen the challenge of the rest of the game (inevitably the fights reward you with SUPER-OMEGA-XCALIBUR-4,) but usually these extra quests are only done after completing most of the rest of the game anyway.

Extending the Final Fantasy example, "hidden extras" in general that provide an extra challenge seem like a good move.  If you love a game so much that you play it enough to find extra bits, you're likely to be looking for more challenge rather than yet another 1-UP.  Bonus levels that are hard rather than full of power-ups.

God Hand (a beat 'em up) has a kind of multiplier system.  Do well at combos and blocking and all those other beat 'em up things, and the multiplier goes up.  Higher multiplier means more points, but also more/harder enemies attacking which makes it harder to maintain your multiplier.

Puzzle Quest (complete puzzles to gain xp and level up on your quest to complete more puzzles) gives you some xp even when you fail a puzzle.  You then have to face the puzzle again and again until you beat it, but you keep getting that xp even when you lose.  That xp will make you level up, which will make the puzzle easier.  Neat!


I tend to find difficulty of my own games very hard to judge and as most of us don't have access to a vast QA department to judge it for us, it seems like well implemented adaptive difficulty could be a real boon.  But how to do it?

[Traveller]

Games that match their difficulty to the ability of the player.
Often we see adaptive difficulty being accidentally implemented and acting inversely, with the game becoming easier for good players.  You've really mastered the controls for a platformer and manage to get up into a hidden area with a 1-UP.  Well done, you're good at the game!  Here, have some resources to make it even less challenging for you.

I don't think this is a bad thing for platformers that the player is supposed to lose repeatedly before winning.  Games where you can save have an entirely different class of difficulty than ones where you can't, and they need to be balanced differently; they can't be compared.

For platformers where you may well get a game over on level 3 quite a few times and start over before advancing, the 1-up you grab on level 2 becomes a bigger deal.  Sure, it's suspended over some spikes in a place that is hazardous to go; a novice player should pass it by.  But an expert player will trust his skill and take that risk for a benefit much later in the game, when it gets REALLY hard.

I think that examples like that are a way of smearing out the difficulty throughout the entire length of a game.  If the end boss is really hard, and you have to play through the whole game to get to him, well, that can make the rest of the game boring.  However if there's all sorts of little challenges throughout the game that can give you a boost against that final boss (like scattered 1-ups), the game gets more interesting.

If you don't like games that can give a 'game over', well, this mechanic isn't worth anything.

[Problem Machine]

Ah, this is a topic that interests me. Most of the work that's been done on 'adaptive difficulty' is pretty terrible when it masquerades under that name IMO; but there's a lot of game features which achieve the same effect without being artificially imposed (a few of which you mention).

One thing that occurred to me about externally imposed dynamic difficulty (where the game says 'oh, he's not doing so well' and makes itself easier) is that basically the only thing keeping you from being unkillable are the reaction times and granularity of the system. If the goal of the system is to make it so you can always succeed, then IF it reacted instantly and appropriately to every incorrect decision you made it would be impossible to die. I actually have a sort of half-baked game concept floating around in my head about a game where the 'adaptive difficulty' is such that reality itself warps whenever you're in danger of dying. I forget who it was (Daniel Benmergui?) who demonstrated a game prototype at the experimental gameplay sessions where the game actively tried to keep the player alive no matter what (though in that case it was to avoid paradoxes), but that was a similar idea.

I think JRPGs are actually a good place to look for ideas on how to have more organic difficulty. Control. For instance, contrast how leveling up is used in most JRPGs vs how it's used in most western games; in JRPGs, it's usually a method by which you can train up to face challenges, but those same challenges tend to still be approachable if you don't put in that training time. So the players can choose their own level. Conversely, leveling systems in western RPGs largely fall into one of two types, area control and meaningless. The area control type is where you more or less HAVE to have reached a certain level in order to be able to explore a given area (WoW). Meaningless is when the levels just don't fucking matter (Oblivion). So, the classic cliched exp/level system can actually be an effective and simple adaptive difficulty system.

Alternately, there are negative feedback systems, where the player is given powers inversely proportional to how well they're doing. An obvious example would be Final Fantasy 7's limit breaks, ultra powerful attacks which charge based on how much damage the player characters are taking. This promotes a volatile match with a lot of turn-around. A western example of much the same system can be seen in the warrior class in WoW, who also generates ability-usage capacity through receiving damage.

I'll post more ideas as they occur to me. I think that there's so many ways to play with this, and most of the 'adaptive difficulty' systems I see proposed are just so ludicrously artificial and inelegant when it's just as easy to propose actual gameplay rules which achieve the same effect.

[Ivan]

Adam wrote a nice blog post about it recently:

http://gamasutra.com/blogs/AdamSaltsman/20090507/1340/Game_Changers_Dynamic_Difficulty.php

[Glaiel-Gamer]

Have each level have 2 exits to find. An easy one and a hard one. Harder ones take you to harder levels. Possibly have a world map, like super mario world.

[ஒழுக்கின்மை (Paul Eres)]

the way metal gear solid 3 and such did it was to invisibly make the bosses etc. have slightly lower hp if you die on them and retry. i took that idea and in immortal defense if you die and retry a level that level's enemies have slightly less hp. it's a *really* insigificantly smaller amount though, like half a percent less. but it adds up if they keep trying a level and keep failing on it.

[Alec S.]

Have each level have 2 exits to find. An easy one and a hard one. Harder ones take you to harder levels. Possibly have a world map, like super mario world.

I like this idea.

Often we see adaptive difficulty being accidentally implemented and acting inversely, with the game becoming easier for good players.  You've really mastered the controls for a platformer and manage to get up into a hidden area with a 1-UP.  Well done, you're good at the game!  Here, have some resources to make it even less challenging for you.

I think one way to get past this is to have optional harder sections that give the player some abstract reward rather than an in-game benefit.  Like points or achievements.  Although I do think that sometimes getting an in-game benefit for your hard work can be a good thing.  Take games like Robotron, where you have to get extra lives in order to survive.

[I_smell]

Just gonna throw another example into this conversation:

Devil May Cry 3 is a really hard game. There are two modes: Normal and Hard. If you die 3 times on Normal though, a screen comes up asking WOULD YOU LIKE TO TRY THIS ON EASY? Personally, I did this and it worked fabulously.

The bad news here is that I lent it to a friend and he gave up halfway through because it was too hard. I said DID IT TELL YOU ABOUT THE EASY MODE? and he said YEAH, BUT I STAYED ON NORMAL. I said OH- WELL YOU SHOULD'VE NOT DONE THAT.

So the moral is to try and do it subliminally or consistantly, because most people will be offended if you explicitly tell them they're not good enough to play the Normal way.

[Alex Vostrov]

I've become very suspicious of dynamic difficulty adjustment recently.  My major problem with it is that there's a high chance of lying to the player.  It's a bit like someone throwing a game to make you feel better.  With that in mind, there are several approaches that I think avoid that danger:

1. Making negative feedback part of the game mechanics.  Maybe being low on health makes you turn into the hulk or something.

2. Explicitly stating the difficulty level, either at the beginning of the game or as the game detects that the player is struggling.

3. Tiered goals with increasing difficulty.  Only the best players will be able to accomplish 100% of goals.  Total completion is not necesary for advancement.

[Alec S.]

I've become very suspicious of dynamic difficulty adjustment recently.  My major problem with it is that there's a high chance of lying to the player.  It's a bit like someone throwing a game to make you feel better. 

That's the thing you need to watch out for with difficulty adjustment.  Sometimes possitive feedback loops can be quite satisfying (for example, in a strategy game, you can end by utterly annihilating your opponent's side), and negative feedback loops can feel like the game is trying soften the experience (for example, rubber band AI in racing games).

I think a good balance is to have the player get benefits in game from hard work that allow them to reach even harder locations (such as how it's done in Metroidvanias and Legend of Zelda games).  The player shouldn't feel like they're being punished for success.

[Anthony Flack]

Powerups in shmups are the classic example of the problem of positive feedback. The game gets easier for good players and harder for bad ones. Not cool. I think that powerups are just about the worst, most balance-fucking element you can put in a shmup. I know people want them, but they all-too-often break the game.

Radiant Silvergun approached this problem in an interesting way. Not only did you not have standard powerups in the normal sense, but if you play in Saturn mode, your weapons actually power up more and more, the more you play the game. And you get extra continues for each hour of play time.

Which seems counter-intuitive, just giving people stuff for nothing like that, and letting them keep it all next time they play. But what it effectively does, is it artificially enhances your skill curve - you get further into the game each time, not only because you're getting better, but because the game is slowly getting easier. This is actually really enjoyable as you feel like you're improving faster than you ordinarily would. It means your first game is short (as it should be - you don't want to go too far into the game on your first run because it's too much to take in all at once), but each subsequent game takes you a bit further. Exactly the effect you want to achieve, only a little bit more so.

[Alec S.]

I think one way to get the best of both worlds is to have benefits that help the player with different skills than they used to obtain the benefit.

In other words, say there's an action adventure game, and the player is good at the platforming elements, but not as good at combat.  They are faced with a boss battle that they're not able to beat.  So they go on a side-quest to get an item that improves the amount of damage dealt.  That item is held behind a series of platforming challenges.  That way, the player still needs their own skill to get benefits, but those benefits aren't augmenting the skill that they are good at.

[Alex Vostrov]

Radiant Silvergun approached this problem in an interesting way. Not only did you not have standard powerups in the normal sense, but if you play in Saturn mode, your weapons actually power up more and more, the more you play the game. And you get extra continues for each hour of play time.

Have to be careful here and not commit the cardinal sin of progression.  Players can be obsessive about hoarding power, and you don't want to force them to waste tons of time on boring and repetitive activities.

I think one way to get the best of both worlds is to have benefits that help the player with different skills than they used to obtain the benefit.

Interesting idea; this could serve as a basis for a cool powerup system.

[Anthony Flack]

Have to be careful here and not commit the cardinal sin of progression.  Players can be obsessive about hoarding power, and you don't want to force them to waste tons of time on boring and repetitive activities.

I think it worked in this case because it was a shmup; learning and replaying the levels is a part of the game's structure, so it happened naturally and didn't feel like you were just grinding to boost your stats. It also had the side-benefit of making every play-through worth something to your overall progression, even if you didn't do so well (Outrun 2006 handled this nicely too, with every game giving you at least a few more Outrun points to spend). Also, with this kind of game, I actually don't *want* to get very far on the first playthrough. I want to start by getting my head around level 1 properly.

But the game I most admire in this respect is the original Sega Rally. When you first play it, it's this knockabout, fun racing game where you bash your way around the course in a fairly haphazard way and generally have a good time. You won't win, but each time you play you can try to reach one checkpoint further than last time, and maybe finish in a higher place. It's a fun game.

Then, you get better at it, and you manage to finish the championship. Woo-hoo! But, you didn't finish in first place. Now it's time to get serious. You've learned the courses, and you find that you aren't really bumping your way around anymore, but are actually driving with a high level of precision. You're trying to nail 9th place on the first track, so you can nail 5th place on the second track, so you have a hope of finishing first by the end of the third track. You're playing the game quite differently now. The time limit is no longer an issue - you're good enough that you don't have to worry about not finishing.

Finally, you achieve first place, and you unlock the bonus track. And it is rock-bastard-hard. You don't have enough time on the clock to have a hope. Now you're racing against time again. You keep improving your skills on the earlier tracks, trying to improve your cornering, focusing on exiting the corner and regaining traction as quickly as possible, looking to shave an extra second off your time for the first course, because that's an extra second you can spend on the bonus course. Now you're playing with a really high degree of skill and precision - the slightest bump; a moment's carelessness can completely ruin your run.

If the game demanded that you played like that from the outset, you'd probably never bother. It's very unforgiving; an extremely tense and demanding game. But it wasn't like that at the beginning - you had different goals, you were bumping your way around the track in a haphazard, arcade-racer kind of way, with little suspicion that you would one day be carving up those same tracks with such surgical precision.

And the real genuis of it all is, that the game itself hasn't changed at all. Only your goals, and the way you think about the game has changed. It's been challenging and engaging all the way from the beginner to the expert level, without any kind of adaptive difficulty being applied at all.

And the whole game is less than five minutes long.

[magnum_opus]

Go ask the Shmups board about rank systems...

The problem with adaptive difficulty is that it's really easy to punish the player for being  good at the game. This leads to the problem that the best way to beat the game is to get really really good, then die and do poorly as necessary to bring the games' perception of your skill down.
Battle Garrega is a good example, a game largely considered impossible to beat with out dying a couple dozen times, and it gets worse because you pretty much have to die at just the right spots too or you end up in bad position to get enough points to get that extend.

[Lyx]

Unless i'm missing something, i think one can summarize it like this:

For difficulty adaption to be non-annoying, it has to be at least one of the following:

1. Be part of the "reward" of playing the game - similiar to how levels get harder if you finish a level.
2. Be compensated with other rewards which make it worth it, so that the player actually wants the difficulty to increase. This may be the same as 1., just differently phrased, not sure.
3. be irrelevant to the "reward"-mechanics
4. Not actually being a difficulty in-/decrease, but just a modification in "how" the difficulty is created (i.e., enemies adapt to player tactics all the time).
5. Be player-requested (Example: Pirates! offers the player - after a very successful voyage - to change the difficulty level. Its up to the player to accept or not)

[Chrissketch]

Its not EXACTLY what the topic is about, but did anyone read this escapist article?
http://www.escapistmagazine.com/articles/view/issues/issue_195/5910-String-Theory-The-Illusion-of-Videogame-Interactivity

Basically, at the end of one of the Half-Life 2 episodes, the difficulty seems to ramp up into an epic last stand until the player "wins" at a the last moment. The result is a scenario that feels like you worked very hard (and you did, technically) to beat the game, but the outcome would have been the same no matter what.

Essentially it IS lying to the player, but in exchange they get an unforgettable experience that felt fun and exciting. I imagine it'd be better for story-heavy games rather than a shmup or something more skill based, but it is an interesting ethical quandary.

[Mr Dumle]

Agreed. I only get annoyed when I know that the game is changing difficulty. The less I know, the better.

[roBurky]

Flow was originally made as an experiment in adaptive difficulty. Read about it here.

The idea was that choosing to increase or decrease the challenge was something the player would choose to do, through the mechanics of the game. If a player wanted more challenge, they would go straight to the next level. If they found a particular level so challenging it was frustrating, they could move back. I don't think Flow itself demonstrated that particularly well, considering the stuff to eat or do in each particular level was limited, but I think the idea is good.