Spiral Knights - Making non-arena levels interesting again: Dynamic difficulty scaling

Making non-arena levels interesting again: Dynamic difficulty scaling

12 replies [Last post]

Wed, 09/28/2011 - 09:44

Nicoya-Kitty

So right now Spiral Knights has a fairly basic difficulty scaling mechanism: monsters take more hits to kill when there's more players in your party. It's not a bad system and it serves its purpose well, but I feel it doesn't go nearly far enough.

There's a problem in SK that we're all familiar with: the relative difficulty and payout balance between different levels in a given depth, and the (lack of) scaling of that difficulty with gear upgrades. People wait for 10-20 minutes at elevators for arenas to rotate in, not because of their unique and varied gameplay, but for their reliable and high payouts and, at least compared to other levels at the same depth, their challenge.

So the question naturally raised is how to make all the other level types equally as alluring in terms of difficulty and payout without A: overinflating the crown supply, and B: discouraging lower-specced or just plain unskilled players (we were all there at one point).

My answer? An AI director.

Anyone who's spent a decent amount of time playing through the L4D series is familiar with the 5th player on every team. The AI director acts behind the scenes to dynamically scale the difficulty of the level (within bounds) by varying the loot and triggering hordes and whatnot to keep the levels moving along at an appropriate pace, and to make sure that the group makes it to the saferoom at the end of the level, but just barely. It's that kind of precise difficulty scaling that gives players the rush of "oh my god how did we survive that, that was AWESOME".

So how would this translate into SK, without completely changing the game?

Well, we start with the basic input stats: Group size, gear level equipped, pickups/consumables carried, health levels. Then add in some advanced stats: DPS given and received per encounter, frequency of shield breaks, status inflictions given/received per encounter, pace through the level, frequency of consumable use and so on.

Taking this data, we can then tune the following parameters: Monster spawn rates, monster spawn types (ie status-inflicting vs normal, lower/higher tier monster of the same type), mender/silkwing spawn probabilities, monster drop types (ie hearts/pots/pods vs heat/crowns/mats), all within fixed limits determined by tier, of course.

So as an example, let's say that a small group it venturing into T2 for the first time. They've just got the basic 2* gear, probably of the wrong type for the stratum theme, and they're taking a lot of hits and losing a lot of health. The AI director tracks this and nerfs the spawns somewhat, throwing in some T1 monsters and reducing the swarm size, perhaps eliminating menders/silkwings. The drops are tweaked to give extra hearts, capsules and vitapods at the expense of crowns and heat. As a result, the party is barely able to scrape through the level, but makes it to the elevator with maybe a few revives.

As another example, let's say a guild party with full 5* gear decides to farm some T2 boss tokens. They're ripping through monsters with ease by working together, taking few if any hits, and saving their pots for the boss fight. The AI director tracks this and buffs the spawns somewhat, throwing in more T3 monsters, increasing the swarm size, and adding more menders/silkwings. The drops are tweaked to give almost no hearts and capsules, smaller if any vitapods, and in their place substitutes heat, crowns and materials. As a result, the party health starts getting whittled away when they make small mistakes, and by the time they reach the elevator they've made it alive but much worse for wear.

Ideally the scaling would be such that the average payout for the levels would be roughly the same as it is now to avoid crown balance issues, and should be evened out across the level types (like, I should expect to receive roughly the same number of crowns from any type of level, given my current skill and equipment), and the average survival rate should be more even across different gear and skill levels, within the appropriate window for the given depth. This doesn't mean someone should be able to squeak through to the core with proto gear; the minimum difficulty bounds of T2 or even the levels past basil in T1 should beat them down sufficiently to keep that from being a realistic option, and likewise T1 shouldn't start spawning epic waves of T3 mobs if someone fully geared decides to make a run through the lower levels to avoid the gate fees or to do a cradle and all run.

So the end result would be that players who do well get more crowns, mats and heat, and players that do poorly get more help just surviving the levels and making it through to the next elevator, and nobody needs to wait around at the elevator for an arena to rotate in, or a deconstruction zone to rotate out.

Wed, 09/28/2011 - 18:17

Antistone

You say you want to make it

You say you want to make it so that people don't always favor arena levels. The only way this plan would accomplish that is if, incidentally, it makes the payouts for arena and non-arena levels similar. Your plan might succeed at doing that, depending on several very subtle implementation details.

But making payouts similar is also sufficient to accomplish that goal, and can be done with far less work, far less complexity, and far fewer changes to the game difficulty by just...um...balancing the payouts. I described several approaches to doing so in this thread:
http://forums.spiralknights.com/en/node/23459

Your idea also has a potentially serious side-effect, which is that grouping with weaker players will cause the difficulty to scale down, and therefore reduce your payouts. I predict people would start aggressively kicking out of their group anyone with worse equipment or who does poorly in any battle, which will make the game less friendly in general and less welcoming to new players...probably not what we want.

Wed, 09/28/2011 - 19:15

Khamsin

I absolutely LOVE this idea.

I absolutely LOVE this idea. The only problem? It would take way too many resources from OOO to develop. If it could actually be implemented properly this would be beyond awesome.

Sadly, I don't see it happening. My honest suggestion is just that arenas and non-arena floors shouldn't compete against each other. If a depth has an arena, it should only have arenas. That's just IMO of course.

Wed, 09/28/2011 - 19:56

Nicoya-Kitty

@Antistone just equalizing

@Antistone just equalizing the payouts wouldn't work very well because then you'd either have to raise the difficulty of the non-arena levels to arena-like difficulty, which wouldn't be very welcoming for noobs or underequipped players, or you'd end up with everyone rabidly avoiding arenas because the difficulty no longer matches the payout. In a game like SK, you need to balance difficulty, crowns, and ce against player skill, equipment and expectation and I don't think that you can do that fully effectively with a static system like what's currently implemented.

As for weak players dragging down the group, that's something I've considered, and there's ways around it depending on how you tune the AI. For example, you could have it effectively ignore the weakest player (not really, you'd use a statistical model that would reduce the influence of outliers) and let them die, counting on the stronger players to revive him or have him revive himself with CE. That way the difficulty and payout would still be roughly in line with design goals. Now, if you've got 3 noobs and a pro on one team, then yeah, the pro is probably going to go solo but there's already incentive to do that in the game currently, as the monster health scales with team size, so it's not that big of a change.

@Khamsin well, they could always just hire me to write it. I am a programmer after all. ;)

And even if they segregate arenas off into their own depths, or hook them up like the graveyards to be mystery levels, there's still going to be loot and difficulty balance issues between different levels. The spiral court levels, for example, are pretty weak on both payout and difficulty compared to a good aurora isles level or even just other scarlet fortress levels.

Wed, 09/28/2011 - 21:43

Antistone

I find your arguments glib.

We already have levels with "arena-like difficulty" (i.e. arenas), and no guarantee that newbies can avoid them, even if they somehow knew that they ought to--so any predictions that making all levels like that would destroy newbies are pretty tough to swallow. And I can only assume that you didn't actually read the post I linked, since most of my proposals wouldn't require making arena-like challenges mandatory.

Furthermore, current payouts are OBVIOUSLY not commensurate to difficulty in general--arenas are clear outliers, to the point where even struggling groups will routinely wait for them (while even groups doing very well often avoid danger rooms). If the devs were making rewards commensurate to difficulty, tier 3 would give a lot more than tier 2 (currently gives roughly equal), danger rooms would give more than they do, graveyards & treasure vaults would give substantially less, and arenas would give a LOT less.

And why do we "need" to make sure that whatever level players happen to play is commensurate to player skill and equipment? The overwhelming majority of games don't, most of the ones that try do it very badly, and we've gotten this far without it. Even your exemplar game, Left 4 Dead, seems to make almost no changes in overall difficulty (you do realize it still asks you to choose a difficulty level when starting a game?), just in pacing, and for all its hype I've seen little evidence that it performs better than a naive random encounter table. I think you are severely overestimating the value of such a system and severely underestimating the challenge in designing a good one. And even if we assume you're capable of producing one, saying they could hire you to do it doesn't solve the cost problem, unless you're proposing to do it for free (and pay for all the QA that would be needed to verify your work). That money could be used to pay someone else to add some other feature.

Your proposal for ignoring a weaker player in calculating difficulty doesn't make sense. If a party of 3 good players and 1 bad player can beat the level (especially while letting the bad player leech health through revives), then it obviously isn't at the limit of skill for 4 good players. Plus, you're giving the weak player high-level rewards for low-level competence, which will make the low-level players who are actually in low-level parties feel like suckers and demolish any claim that the system is "fair". You would need to somehow accurately rate the contribution of each member of the party and give different people different payouts based on their individual contribution--and if you think you can do that, you haven't thought about it very much.

People rarely kick allies under the current rules because the other players need to be REALLY bad to actually make things harder. Monster health increases with players, yes, but not in proportion to players; a group of 4 faces only around 2x the monster health of a solo knight, not 4x. Plus, hearts are multiplied (while monster damage is not), and each knight is targeted by fewer monsters at once, making many fights significantly easier.

Wed, 09/28/2011 - 23:02

Nicoya-Kitty

@Antistone you make some very

@Antistone you make some very sound arguments that are unfortunately completely disconnected with reality.

The most challenging phase of the arena is entirely avoidable by simply going down the elevator after the first or second phase. Any noob who doesn't completely fail in the first arena section (which has a difficulty on par or below any similar level for the depth) has that option. I also did not imply that arena-like challenges would be required, only that the level of difficulty would scale up to what one would currently find in a complete arena playthrough.

The SK developers are obviously and overtly concerned with matching difficulty with payout, as well as keeping cr gain vs ce expended balanced. If you've read any of their posts on here, or update announcements, you'd know as much. With the current system, though, they often have to make tradeoffs between balancing difficulty and payout, and balancing ce and payout. The whole point of the recent danger room revamp was to try to rebalance the difficulty/ce/payout ratios.

SK needs dynamic difficulty scaling to keep the game interesting to seasoned and geared players while still being accessible to newbs. One of the cancers of RPGs is that player/gear levelling boxes you into a small subset of content bordered on the one side by extreme difficulty and on the other side by tedium. Scaling the payout with the dynamic difficulty gives players back the reward for levelling up that would otherwise have to be covered by repetitive grinding of lower level content, like T2 arenas for example.

I really find it difficult to believe that you've spent any reasonable amount of time playing L4D if you haven't seen the very blatant difficulty scaling that the director AI performs. The equipment stashes, the horde sizes and frequencies, and so on all vary significantly depending on how well your team is doing. The difficulty setting only adjusts how forgiving the AI is, and where the upper and lower limits for the various parameters are set.

I think you misunderstood how the difficulty scaling would work in a team with one soft player. The calculations would not be significantly influenced by the outliers, but the outliers' effects on the progress and status of the team in general would of course factor into the difficulty scaling. There's no way to not make a soft player impact the success of a given team, but it would affect the difficulty and payout that the other players a lot less than you seem to assume.

Frankly, people rarely kick allies under the current system because it's rude. I usually solo because it's honestly a great deal easier for me than trying to go in a group. I end up taking a lot more damage and needing significantly more revives when I'm playing with a group, unless the group happens to be really top notch. If anything, a dynamic difficulty scaling system would improve that situation. Granted some people will still be dicks, but there's no point trying to seek a technological solution to a social problem.

You clearly feel very strongly about your own proposals and I respect that. However, I would invite you to not let that cloud your perception of the ideas I put forward here.

Thu, 09/29/2011 - 00:17

Niichi

I'm not sure if this idea

I'm not sure if this idea really addresses the original problem to be honest. I mean, if loot scales with it wouldn't people still wait for arenas? Because the majority of arena loot comes from the enemies so it would still be the most profitable area.

In terms of the idea on its own merits it sounds pretty fun and makes levels a bit less predictable. Needs to be customised so in some areas it has no effect. Like the last thing we need to do is make danger rooms more difficult just because we made our way through the level itself pretty well.

Thu, 09/29/2011 - 00:59

Antistone

Difficulty vs. Payout

I've read many developer posts and update announcements since I started playing, so if this is true, I'd appreciate some links. I'm aware they altered danger room rewards recently (as well as massively altering danger rooms themselves), but I saw no statement of policy accompanying that change, and do not recall anything else related to this topic.

I'm done arguing with you, though.

Thu, 09/29/2011 - 02:30

Australiumfish

Ive played L4D 1 and 2...

Having played L4D enough to understand just how different the same levels can be with the Director, I must say that I completely agree with the idea of adding it to SK.

For example: Im 4* standard (cobalt, etc, with some other stuff) and I make cr by doing twins runs and selling catalyzers. I have friends who I do such runs with, and we clear the floors far too easily, and I think we would gladly accept more enemies/less hp drops in exchange for more cr payout! Its not exactly useful when we clear a danger room mid-twins-run and there are three hp pills lying on the floor when we all have three in our inventory anyway! We would far prefer more crowns instead!

Also, I know some newer players, who have difficulty fighting down to Snarby, and would greatly accept reduced difficulty in exchange for less cr payout.

So, +1 from me, and about +10 from my high* and low* friends.

Thu, 09/29/2011 - 02:40

Ufana

Most resources come down to

Most resources come down to ce. If you give a group of players a higher payout, the rest will see raised ce prices not only absolutely but in relation to what they are able to earn and that - with their lower payout - makes the game harder and more frustrating for them. I don't think a "longterm Harder and more frustrating adjustment" is what "lowskill / equipment" players (that already have a "shortterm hard & frustrating" experience at times - exactly those are in the "lower payout group") need or want ;)

The only form of reward you could adjust was heat- and that would still be very frustrating / unfair towards players that don't actually need heat, because they would get less cr/ce per time while playing good and that's usually exactly what those players need - otherwise they'd craft something to lvl!

Thu, 09/29/2011 - 09:36

#10

Nicoya-Kitty

@Niichi I don't think I'd

@Niichi I don't think I'd change the arenas (at least under this suggestion), so it'd really just bring other levels more in line with the arenas. Danger rooms and boss fights would likewise remain relatively unchanged, since in general boss fights are expected to have a roughly fixed difficulty and reward.

@Ufana There's always going to be a range of payout for levels that straddles profitability relative to buying CE with CR. It's like that now with low-tier players stuck in T1, or in T2 but unable to reliably clear arenas. The rebalancing under this plan would just expand how much content is viable to make a profit on. Even if you're running 5-10% below profitability you're still converting 90 or so of your mist every day, so while you wouldn't be able to profitably run on CE, you still wouldn't be stuck in a hole forever.

Fri, 09/30/2011 - 03:41

#11

Ufana

Ok, let's just assume that's

Ok, let's just assume that's right because really we can just guess unless we try...

What are the benefits of your idea if you compare it to letting the experienced player do the "difficulty scaling" himself by choosing his actions carefully?
Which is...
- Already implemented to some extent and easy to build upon (e.G.: make lichen merge instead of killing them right away. Or killing devilite overseers first, and if possible exclusively until the last devilite turns into an overseer. Or wearing a swiftstrike buckler for more dmg / lvls per time. Using vortex in fsc instead of shivermist, which can be more "dangerous" but a lot faster.).
- Less prone to bugs (I can practically guarantee you, with an ai governor you'd get stuff like trojans spawning instead of zombies on graveyards on rare occasions in the beginning at least. While I'd LOVE to see that - I'm all for horse cemeteries - I'm not sure the majority of players would agree).
- An actual choice you make. Not "an AI messing with you", which can get quite frustrating and not all players might like it.

Fri, 09/30/2011 - 12:46

#12

Nicoya-Kitty

@ufana currently there are

@ufana currently there are some options, as you mention, to manually scale difficulty. Going T3 instead of T2, letting lichen colonies merge, entering danger rooms and so on. The problem is that it doesn't really provide the ability to scale the difficulty and loot on a given level to both match the arenas on the high end, and not grind noobs into the dirt on the low end.

I'm sure there'd be bugs at first, but that's pretty much par for the course with any SK update. Like when the Candlestick Keep levels came out, there were problems with dead-end traps, gates not dropping when an area was cleared, etc. But it's not like they can afford to just let the game stagnate forever.

I think it's a given that not everyone will like every game mechanic. Some people like things that are easy, predictable and boring. I personally think that subset is much smaller than those who like things challenging, unpredictable, and tailored exactly to their skill and gear.

Search

Making non-arena levels interesting again: Dynamic difficulty scaling