The problem with your Ai example is that it's real time strategy. I'm actually a CS major finishing up as well (AI is what I'd like to specialize in and AIs in relation for games is a graduate class I just finished) so we have that in common, interestingly. RTS games give AIs a significant advantage in how fast they're processing and deciding actions; there's also the sheer breath of options to consider at any given moment. It's not that AI is inherently better than the player.....it's that the player can't keep up with the AIs processing power. In an effort to keep up, a player prioritizes and must be faster at deciding, this creates error propagation over time. Comparing AI for RTS to a turn based game is in error. There's also the fact that the type of AI used and how it's developed comes into play. Machine learning is generally how the best RTS AIs are made (correction: I'm not 100% sure on this as they might also use another method for just calculating risk and gains based on what they know so far given player actions)....we don't exactly have that advantage in EA where the devs are planning to try out different systems and permutations; the machine won't have enough time to develop high level skills for a given permutation if the devs are tru to the desire to try new things alot in EA.
Also, the air ability is their for a reason for mages to destroy mage armor and field control is more important than straight damage. Bodybuilding would regelate warrior CC the same way mage CC would get treated. And I think that it creates diversity to have a player decide whether they want to build up CC chance, damage, field control, and so on.
The only reason the last game devolved is cause of how powerful CC was. The armor system already mitigates this. Different skills and anilties also effect how this would effect things. Leaving out drain willpower would be an option, for instance. Or placing in options that increase chance to resist along side chances to cause effects.
If the defensive abilities were directly changed out and the chance system placed along side the armor in the current system, right this moment? ... I think a lot of people would feel better, myself among them. I'm not quite sure how'd it effect peoples build choices but thats what the EA is for. I actually think more diversity would open up, in some respects.
Would you concede that implementing the system combination for an amount of time right now while they'er trying things would shed light on what works better for all involved? Obviously the current system needs work cause it's not satisfying for many people....and not everyone wants the old system. Trying a quick permutation of both and other systems would help quickly narrow down whats best rather than constantly tweaking one. Think Big O: n^2 vs logn vs nlogn .... at the moment going through every possible tweak of a system type starting from the starting one sounds like it'd take n^2 time since it means we introduce a system then go through permutations and then do the same with another.
And you bring up a very good point about the AI and the current deterministic system. It works well, but the problem lies in that it'll never work better than a player, ultimately. And like chess we will eventually reach the limits of how "surprising" and how many "oh shit" moments such a system can give. Things Swen himself has expressed to want give more of in this game. He want's players to adapt and feel a constant pressure to improve to combat. He wants consistent difficulty that involves more than learning how to perfect a system of decision making; at least thats the impression I got. In the last game we all learning to perfect our decision making by using CC on round one, then combo, and so on.
The question at the moment isn't whether or not the AI can be made better. But if the AI will difficulty curve can ever really give a consistent challenge to all player levels? Ideally, yes, but realistically will fall short at master levels where "trick" plays come into account where strategy has more to do with learning to predict opponent moves and reactions. Will the a perfect AI make the system more fun? Certainly challenging, but fun is more ... intuitive. Will it make it like chess? Assuming perfect players on both sides -> yes it will. AI's in this game have lots in common with chess after all. Is it possible to make a perfect AI for this game? Technically, yes, but I don't know if I'd bank on it with all the variables and how long it might take an AI to go through all that in a reasonable time frame given current algorithms. Than that brings us down to the BIG QUESTION: Should combat be like chess? And can combat withing chess be improved?
This requires user data. I'd answer No and yes, respectively, but we won't really know till we try new things and take feedback.
At the current moment combat is more like chess than not and I don't think that's nearly the best way to do things; regardless of how hard you can make it
It's interesting that you're a CS major, maybe I can speak more freely because of this. Let me draw from experience to explain this to you. What AIs have you worked with? I've done AI work with modding for Crysis in CryEngine, Lua for various games, and most relevantly, in the Aurora engine, for Neverwinter Nights 2. That one, I'll go on about, because it is remarkably similar to Divinity: Original Sin, in that it is a real time turn based strategy game. The AI had the capability to be more than competent. In fact, it was crushingly difficult once you started customizing how characters acted and behaved. Individual AIs were not customized very much, though. Every ability had its own AI: fireball, aimed itself at the biggest crowd of enemies it could without hitting an ally, and had a weight for fire immune or absorbing enemies, and exceptions for allies that are the same. These AIs, when put together, formed a creature, together with the basic creature AI, which varied by type. The AIs had a priority order, and would enact themselves in the order you put them. There were special scripts for abilities that interacted, and use cases that didn't come up often, like if an enemy could tell a player had lots of negative levels or was weak, Enervate would shoot up in priority and get used.
The point being, AIs for games like these are remarkably flexible, and very much so doable. Perfect AI? I think so. In Neverwinter Nights, against player AIs in one certain arena module that was quite popular, the game was almost unwinnable. Because the AIs were so darn good. and they're flexible, too! If you level drained an enemy, and they could no longer use a skill, its AI would adapt because of it. If you put on Fire Shield, it would stop casting fire spells, if your team was mostly down it would switch to single target spells for those that were left, etc. It's not hard to emulate this. The AIs, from the jargon I worked with, were not very complicated. They were just laboriously and meticulously crafted, each and every skill, until near-perfect combat was achieved. It was beautiful. I'm sure something similar could be done for Divinity, since the games are so similar, correct me if I'm wrong. I would highly recommend looking up and reading about the aurora engine; it was a beautiful thing, and very relevant to this debate.
Also, the air ability is their for a reason for mages to destroy mage armor and field control is more important than straight damage. Bodybuilding would regelate warrior CC the same way mage CC would get treated. And I think that it creates diversity to have a player decide whether they want to build up CC chance, damage, field control, and so on.
I'm not sure I follow? Air abilities don't do much damage to magic armor. I know the skill itself helps, but without hail strike, and with Rage and Savage Sortilege, the highest amount of single turn damage I've seen, with 6 AP, is 400. That's not even enough to break the worm's armor, and it's immune to CC anyways.
I'm sure bodybuilding would be there as well. But it wouldn't help you against a 400 damage rage/crippling blow combo, and that's the damage that my mage does with Crippling Blow, not even my warrior. As for field control, Phoenix Dive is used by like half the warriors in the game right now, and we're only in Act I. It's clearly many times less important than it was in the first game, and magic armor only serves to emphasize this point.
The only reason the last game devolved is cause of how powerful CC was. The armor system already mitigates this. Different skills and anilties also effect how this would effect things. Leaving out drain willpower would be an option, for instance. Or placing in options that increase chance to resist along side chances to cause effects.
This kills the mage. Seriously. I can think of no quicker way to make a class less relevant than to take out the only thing that made them relevant in Honour Mode. Like I keep saying, who would play a mage when their damage is lower, their mobility is nonexistent (barring self-casted netherswap), their armor is low, their HP is lower due to having to pump their main stats, and Warriors can do twice the damage in one click than a mage can do in their whole turn, with three cooldowns and saved AP? And especially when warriors have more CC? Why would you? Because they have range? Archers have more range, more damage and special arrows will probably provide more reliable CC. Mages would be useless if their CC wasn't guaranteeable, just like every bodybuilding based CC ability was in EE.
If the defensive abilities were directly changed out and the chance system placed along side the armor in the current system, right this moment? ... I think a lot of people would feel better, myself among them. I'm not quite sure how'd it effect peoples build choices but thats what the EA is for. I actually think more diversity would open up, in some respects.
Uuuuuuh... I don't think I need to say anything.
Would you concede that implementing the system combination for an amount of time right now while they'er trying things would shed light on what works better for all involved? Obviously the current system needs work cause it's not satisfying for many people....and not everyone wants the old system. Trying a quick permutation of both and other systems would help quickly narrow down whats best rather than constantly tweaking one. Think Big O: n^2 vs logn vs nlogn .... at the moment going through every possible tweak of a system type starting from the starting one sounds like it'd take n^2 time since it means we introduce a system then go through permutations and then do the same with another.
Yes! I will absolutely concede that implementing the system and trying it out is a great idea! I would gladly drop another 100 hours on it if this was the case. But you can bet I wouldn't touch my favorite class with a 10 foot pole.
And you bring up a very good point about the AI and the current deterministic system. It works well, but the problem lies in that it'll never work better than a player, ultimately. And like chess we will eventually reach the limits of how "surprising" and how many "oh shit" moments such a system can give. Things Swen himself has expressed to want give more of in this game. He want's players to adapt and feel a constant pressure to improve to combat. He wants consistent difficulty that involves more than learning how to perfect a system of decision making; at least thats the impression I got. In the last game we all learning to perfect our decision making by using CC on round one, then combo, and so on.
The AI can certainly be better than a player. Have you played a pub lately? I could make an AI that does nothing but hit barrels, and it would probably teamwipe someone. The AI doesn't have to be perfect, again: it just has to provide a challenge. The Starcraft AI was hand-coded to take advantage of specific strategies and combos, and was able to beat professionals. Machine learning was used to eliminate only the worst strategies. I know it's not turn based, so it's different, but Neverwinter Nights has convinced me that this AI could be perfected, too.
In short, I believe that my #1 criticism, would be that you underestimate the potential of a good AI. My second criticism, being equally important, is that you do not seem to have a good grasp of the meta of DOS or DOS 2. I would recommend getting Enhanced Edition and trying an honour mode run. Try it with mages, then try it with warriors. You will see a night and day difference in difficulty, tactics, and general survivability between the two playstyles, and will most likely come to agree that CC was not the #1 problem in that game. Not by a longshot. You should also play some pubs in #2. It was eye-opening for me the first time I saw a warrior do 1200 damage to the last boss, killing him before our pyromancer could break his magic armor (he had already dropped a full combo, starting with Flesh Sacrifice). I absolutely insist that you become more familiar with the power discrepancy between powerful warriors and CC enabled mages before continuing on your current line of logic. And if you still think there is an issue with mages having 100% capable CC, then I will listen intently, as I cannot fathom someone familiar with the meta and capabilities in this game saying it is too powerful.
No offense intended.
