Howdy, Stranger!
Categories
- 6.2K All Categories
- 3.1K Everything Achaea
- 1.5K North of Thera
- 26 Archives of the Terraformer
- 257 The Matsuhama Arena
- 885 The Golden Dais of Creation
- 299 The Scarlattan Theatre
- 149 The Blank Canvas
- 2K Getting Help
- 412 General Questions
- 258 Quick Class Questions
- 1.3K Tech Support
- 313 Client Help
- 486 Curing Systems and Scripts
- 841 Off-Topic
- 255 The Wander Inn
- 586 The Universal Membrane
- 298 Class Discussions
- 298 Individual Class Sections
- 20 Alchemist
- 9 Apostate
- 30 Blademaster
- 9 Depthswalker
- 14 Druid
- 4 Infernal
- 20 Jester
- 19 Magi
- 34 Monk
- 11 Occultist
- 8 Paladin
- 7 Priest
- 29 Runewarden
- 18 Sentinel
- 26 Serpent
- 20 Shaman
- 9 Sylvan
Comments
But there were several cases where it was just not realistic to cure quickly enough. Apostate. Thoth's Fang serpent. BM and alchemist have continued a trend towards faster, more aff-heavy classes. There were plenty more cases where curing was possible, but the margin for error was punishingly low. Automated curing can be exploited, but it doesn't make mistakes: doesn't typo input, press the wrong button, or apply too early.
Assuming manual curing is possible, you still have the problem of handling your offense alongside your defense. That's two sets of complex, demanding input. Most classes have input-heavy offense, and tend to get more complex with skill revamps (not that I'm complaining). There are only so many macros you can map to your keyboard.
Also, @Ada: Where have you been?
But I still have no f'ing clue wtf a gmcp is or how to use it with out asking Karai, Daier, or looking up other peoples examples and f'ing it up.. took me like 3 months to figure out how to use gmcp to make sure I was wielding my rapiers before I dsl'd.. Just kidding Achaea totes poped out with wield rapier226 left wield rapier556 right before I could get it figured out.
Occultist is extremely easy to automate (in comparison to other classes) The main part about occultist that makes it so easy is your not using afflictions to lock, your not using afflictions to boost your damage your just stacking them to an instant kill. It doesn't matter what afflictions you really use so long as they're part of the set for Cadmus/Hecate/Whipsering/Enlighten.
-- At least my thoughts after understanding how Occultist abilities work (finally after like 2 years or something now) and then looking at combat logs to see how others do it.
While manualing Occultist on the other hand would be 10x harder than automating it since the pets randomly afflict different things you'd be needing to afflict..
Shield of Absorption: 800 credits
Thought both of these items would go up a little more than that, but still glad I bought them before it went up what they did.
[ SnB PvP Guide | Link ]
Momma needs her masquerade outfit!
That love soon might end You are unbreaking
And be known in its aching Though quaking
Shown in this shaking Though crazy
Lately of my wasteland, baby That's just wasteland, baby
On Aetolia, there's literally people who have 1-button 'fight' scripts that execute their entire offense until the opponent dies. These people proudly proclaim this on their forums, and will argue (similarly to some people earlier in this thread) that that's the fun of combat for them. They will defend this as 'working as intended' even if most of the game seems to agree it's awful.
The reason this is common in Aetolia, though, is design decisions. Over there, afflictions are not cured randomly if they share a cure, but always in a specific, set order. Additionally, when you use random cures like tree tattoos or whatever, your opponent gets a room message of what was cured. This means that it's already pretty easy for an AI to keep track of afflictions.
Second, every class on Aetolia ended up designed like Occultist over here due to the curing system. I played the druid equivalent over there and the offense consisted of 'stack goldenseal afflictions until X affliction is stuck, then use Y skill that becomes usable only with certain affliction there', then repeat to increasingly stack things in a linear fashion until you could execute the final blow (I may have the details slightly off, I was never much of a combatant and it was a while ago, but the general gist was that). Every other affliction class was like that, and essentially was completely linear with no chance of branching off because of the curing system.
I bring this up because even if you can't outright fight automation, you can certainly design around it. High level chess bots are easy and common, but high level no-limit texas hold'em bots are still in their infancy not just because people started earlier on chess. Chess is a game where you always have perfect information on what your opponent is doing, and there's always a semi-linear aspect to moves. NLTH on the other hand, relies on an extreme amount of 'unknown' information. A lot of this unknown information becomes player-dependent, especially as you rise in player skill, so AIs can compete against mediocre players, but have no chance against a top player, especially once they realize it's an AI. My information might be outdated since I rely on friends that play poker for this and haven't talked to them in ages, but last I checked the closest we had to a good bot could only do one-on-one at a semi-respectable level. Limit texas holdem, on the other hand, is very easy to make AIs for because it's much simpler to just rely on odds and there's less unknown information (a player can only bet/raise, call/check, or fold, you don't have to consider bet sizing on both players' ends). Obviously, omaha holdem (4 cards in hand) is even more of a nightmare to make an AI for.
The point of that tangent is that along with Aetolia's example, we can see that AIs are limited largely by the game design. Classes like occultist/every affliction class in Aetolia are easy to automate because they have no 'guesswork' to do, and they follow an extremely linear and simple kill sequence. Classes where the opponent can change what your optimal strategy is at any point (I think people point to infernal as a good example of this often, though I don't know for a fact if it's true) are much more AI-proof to an extent. Moving forward, I think Achaea needs to consider designing classes in ways where their offense is less linear, and defense can be more unpredictable if they wish to avoid more automization. The more information that is unknown (preferably information that can be guessed by a good player based on psychological edge rather than completely random), the harder time AI-coders will have while human players will generally just find that another way to hone their skill.
TL;DR: Linear offenses with little unknown information are easily automated, introducing psychology-based guesswork and branching offenses is the best way to avoid AI offense. The harder it is to figure out the perfect unchanging strategy the better.
That love soon might end You are unbreaking
And be known in its aching Though quaking
Shown in this shaking Though crazy
Lately of my wasteland, baby That's just wasteland, baby
That love soon might end You are unbreaking
And be known in its aching Though quaking
Shown in this shaking Though crazy
Lately of my wasteland, baby That's just wasteland, baby
I thought that Ayami's point was similar to what you are saying. That's what I meant by "speed poker" (vs. the analogy Ayami made with chess, which is linear with no unknowns, and which computers excel at, while humans can still win against a computer at poker). I think what you're saying is that Achaea already has at least some of those poker elements. My only real point in my original comment was that it avoids the AI problem, but also selects for a very specific kind of player (i.e. probably someone who could play speed poker).
EDIT: I think my use of the word "speed" is causing some confusion there. Computers are always faster than us, of course, and all I meant by that was that in Achaean combat (even if they slow it down some, as has come up in this thread), we still wouldn't be sitting there eyeing each other over a hand of cards.
That said, I disagree, Austere. The fact that Aetolia has 1button offenses (and I do mean hit one button then you can go afk), and Aetolia is not that different from Achaea, proves that it can happen. The only difference is that it's a lot harder in Achaea still because there's some unknown information. Like you said, expert diagnoser is probably a step in the wrong direction. Occultist is still easy to automate though, because like someone mentioned it doesn't really care about what afflictions you have for the most part, just a number. And all cures currently give a third person message, so it becomes simply 'have they cured more times than I have afflicted?' Because it has no real alternative offensive paths, there's no reward at all for not automating, either. You can skip steps sometimes, but there's not much truly different you can aim for.
That said, the ideal is not to rely on truly random things like "which affliction got cured?" but to rely on psychologically random things. You want people to be put in situations where by defending against move A, they're making themselves vulnerable to move B, and predicting what your opponent is going to do becomes the best way to choose which to worry about. Some of the new class redesigns seem to be trying this (magi, maybe, though I have no actual knowledge of the class, just the announce from a while ago + posted ab files), but if the goal is avoiding full offensive automation it needs to be a principle applied to all classes in an extensive way. This would maybe make Achaea's combat even more complicated, though.
You can disagree all you want, it doesn't change the fact that occie doesn't even need automation. A human can perform just as well as a script. Of the top occies right now, there is only one that actually has an accurate(accurate is subjective, I do mean perfect) affliction tracker(to my knowledge), and by your own admission, it doesn't even need to be accurate. It just needs to count afflictions. Quite honestly, I don't think having a script count your afflictions for you is really automation. I equate it to a limb counter. You don't even need it to select your offense if you understand your skills. There is little benefit. The less benefit gained, the less wide spread use there will me.
I released my script because I understand that at least half of the top tier is using automated offense selection to some degree(and don't blame this on one city, I have spoken to a lot). This is not a new problem. It's just more open now. Now that everyone understands how it works and what the problem is, we can either all play on the same field with automation or we can change how combat works. Quite frankly, I want combat to return to a simpler, more linear time, but I am prepared for either. I have offered my assistance in this thread to nerf(outlaw and track for shrubbing if it comes to that) affliction tracking once and received no divine response (they never respond to me anyway on forums). It's our place to play with the tools we have and within the rules they set. This is not something we can fix as players. It needs to come from the top.
Having "psychologically random" things doesn't really change much with respect to automation either. Either people are changing their offence and defence reactively, which is plenty easy to write an AI to do, or you have a rock-paper-scissors scenario where the changes aren't in reaction to particular states and an AI transitioning randomly will perform equally as well as a human (or if you want to get fancy, tracking the statistics of the state transitions and using that, since people probably don't actually play rock-paper-scissors, or any kind of analogous game, by playing perfectly randomly).
Games that computers can't play well are pretty exclusively games with ridiculously many permutations and extremely chaotic systems (in the sense that small changes can lead to very large differences in outcomes) - Go is a good example since (1) there are far, far too many possible moves to brute-force a solution for the vast majority of the game and (2) a single move can, at almost any point, completely change the state of the game. The problem with Go is that a small move can have huge consequences several moves down the line, but no AI can evaluate enough moves deep to see those consequences and work back to what small moves to make. Put another way, computers are bad at Go because we don't really understand how humans manage to be good at Go. Humans don't have the computational resources to brute force the game either, so expert Go players must ultimately be employing some sort of more-local algorithm, we just don't know what it is. Perfect knowledge or lack thereof is not what makes the game (or any game) difficult to write AIs for - relative uncertainty is going to be the same for players as for automated systems. And that's the sort of uncertainty where the alternatives don't necessarily have equal chance of selection, whereas things like uncertainty about herb cures can be quantified using simple percentages. Crucially though, what matters is that Achaean combat is supremely analytic - one of the reasons a lot of people are drawn to it - people do things for generally clear reasons and not in any way the sort of reasons that are difficult to emulate with a computer program. Even if people are changing their strategy simply for the sake of changing it, not in reaction to a particular state, that's an easy thing to model and implement.
The reason we haven't seen more full automation is because the people who are most knowledgeable about combat either don't have the programming ability (not that it would require anything particularly fancy) or don't have the interest to develop fully automated systems, and the people with the ability and interest haven't historically been knowledgeable enough about combat to generate anything worthwhile. But it is absolutely, unquestionably possible.
With a class like Occultist, a human has absolutely no advantage over a properly coded AI. There is no point to a human even bothering with its offense when an AI can do it as well or better for no effort and minimal setup/coding. Relying on 'well it's not absolutely needed' is a bad idea, because necessity isn't what drives people to code bots for chess, poker, counterstrike, whatever. You need a more active approach to dissuading automation. Otherwise, I guarantee eventually there'll be fights where human input is completely minimal and an AI absolutely will replace all human input for some players during a one on one fight. Obviously over a long enough time that'd happen anyway, but the proper path is to try to stall that as long as you can, or you run into chess's issue.
Making offense simpler/more linear unfortunately doesn't dissuade automation in the long run. Just because a human can play perfectly well doesn't mean someone won't come that is simply more interested in coding a bot than any other part of Achaean combat. The ideal kind of design is something that's both simple to grasp and favors human capabilities over computer capabilities, which is the opposite of Occultist.
Oh, I was just addressing the 'speed' poker part. It doesn't even have to play like poker or anything, really, it just needs variety in offensive actions.