Greetings, glorious adventurers! If you're joining in our Alpha One spot testing, please follow the steps here to see all the latest test info on our forums and Discord!
Options
What are your thoughts about self-learning AI in MMO's
Gubstep
Member, Braver of Worlds, Kickstarter, Alpha One
https://massivelyop.com/2020/03/20/microsoft-is-using-cloud-computing-to-make-unbeatable-raid-bosses/
"According to software engineer James Trott, who spoke as part of a Game Stack livestream, Microsoft has been working on an MMO-like project last year that featured a raid boss that was capable of noticing dominant player strategies and adjusting its behavior in nearly real time. Apparently, this new machine learning technology could be leveraged by game developers in the next console generation.
In fact, this new AI system is apparently so good at what it does, Trott advises developers to tone it down. “The challenge for developers is how to tone back that machine learning and simulation, given the amount of compute in the cloud, so that the NPCs and monsters aren’t perfect,” he says. “Because with enough training and enough compute, they will beat the player every time.”
What are your thoughts on implementing this into future MMO's and updates?
Do you prefer static raid encounters where there is no variation or would you like to see an encounter where cookie-cutter strategies are only viable for a short period of time before the boss learns it?
"According to software engineer James Trott, who spoke as part of a Game Stack livestream, Microsoft has been working on an MMO-like project last year that featured a raid boss that was capable of noticing dominant player strategies and adjusting its behavior in nearly real time. Apparently, this new machine learning technology could be leveraged by game developers in the next console generation.
In fact, this new AI system is apparently so good at what it does, Trott advises developers to tone it down. “The challenge for developers is how to tone back that machine learning and simulation, given the amount of compute in the cloud, so that the NPCs and monsters aren’t perfect,” he says. “Because with enough training and enough compute, they will beat the player every time.”
What are your thoughts on implementing this into future MMO's and updates?
Do you prefer static raid encounters where there is no variation or would you like to see an encounter where cookie-cutter strategies are only viable for a short period of time before the boss learns it?
Software Engineer
Addon Developer
OG Backer
Addon Developer
OG Backer
0
Comments
Raids are a puzzle between developer and player. They come up with a puzzle, express it in script form and players try to work it out.
If developers use AI to eliminate actual strategies as players find and use them, all raids will descend in to a free for all clusterfuck.
On the other hand, if they are able to program in an intended strategy, the AI can then step in to fill in any gaps left over by lazy developers while still maintaining the integrity of the puzzle.
Have you seen open ai vs dota pros? It fkn wrecked them.
I can see it working well in a single-player environment but not in a coordinated group environment.
Would be the first greatest step to actually making PvE dynamic then just a dull numbers game based on responding to a choreographed fight. Hell, I might actually enjoy PvE for once in an MMO instead of it being a means to an end. It would basically move closer to what I get out of PvP. A challenge.
Oh, see, I was talking about real raiding.
As in, not in WoW.
In games with real raiding, you don't "need" AI to make interesting encounters.
I doubt it would help make content interesting in WoW.
What games do you mean? Guild Wars 2, where people stack ontop of each other, then run into predestined zones to soak just to run to group once more?
Most raids in other games are exactly what wow does.
Challenges come in many forms. Both raiding and PvP can be hard but they require different skill sets. Raiding in most mmorpgs is like a puzzle. You have all the information available to you and it's about finding the best solution to the problem in order to succeed. This is a very different kind of gameplay to most PvP.
Anyway, you want raiding to be more dynamic like PvP right, but remember that Ashes raids will be 40 people. This completely changes the group dynamic and reduces coordination.
Now I haven't done a ton of large scale PvP but there is usually very little coordination and teamwork involved. At most you'll have a commander who pings an objective but after that the players pretty much just do whatever they like. You might get the odd few who stick together and work as a team but the rest will just be individuals who happen to be friendly.
This is very similar to what used to happen in historical battles. Sure, both sides would start off with very ordered ranks and clear communication, but as soon as the 2 armies clash more often than not it becomes a free-for-all where the objective is just to stay alive rather than work together.
Dynamic PvE could potentially work in a small group(2-5 people) environment but certainly not for a 40-man raid.
But for re-spawning, 40-man size raids the boss has more HPs thus more time to adapt and react should the groups' DPS not be high and sustained - This could (would?) doom PUGs as well as more casual groups.
Agree with those that would look forward to this AI in a solo, or even co-op, game. But for the current, standard MMO boss mechanics adding this AI could be much more than 95% of the players could handle.
EQ, EQ2, Rift, Age of Conan and Vanguard are five very easy and very obvious examples of games here.
And all of them sadly failed didnt they? Age of Conan was an utter mess, I watched a Vanguard raid JUST now and it was the exact same mechanics as a wow raid..., didnt really play EQ or EQ2 but both are dead, didnt really play Rift either because it was already F2P when i heard of it...
EQ has been going for almost 21 years, and has had 26 full expansions - the most recent of which was just 4 months ago.
EQ2 has been going over 16 years, and had it's 16th expansion release in December just been.
In a genre where games are often left unsupported only a few years after launch - or even when servers are turned off after only a few years - any game still getting full content expansions after that long is a standout.
There are many words and terms you could use to describe both of these games in terms of their success. A failure is not among them, though.
One term I'm sure Intrepid would use is "a goal".
Regardless of how successful EQ and EQ2 are, from what I'm seeing their raid bosses aren't much different to what you'd see in other mmorpgs like WoW.
https://eq2.fandom.com/wiki/Hragdold_the_Frenzied
In terms of mechanics this is a pretty simple fight in my opinion. What makes these raid better than WoW's?
You probably know from my posts on other topics that I am a fan of encounters being somewhat varied.
You've picked a fairly straightforward encounter there. It is one of the easier encounters from one of of the easier raid zones of that expansion.
That is understandable though, as the harder encounters - the fun ones - don't have much information about them out there.
These are possibly my two favourate encounters from my time in EQ2.
https://eq2.fandom.com/wiki/Byzola
https://eq2.fandom.com/wiki/Roehn_Theer_(Palace_of_Roehn_Theer)
Due to a number of factors - guilds wanting to keep stratagies of hard encounters to themselves, but by now it is more because stratagies for these encounters are too complex to communicate without actual discussion - you won't find much on them anywhere.
Even encounters that do have some strategy listed are missing absolute key aspects to the encounter everywhere you look for them online.
Well now I'm curious. What made these fights so special?
Is this the boss you are talking about? I'm watching it and from what I can tell you have a pretty standard fight here. Tank the boss in the corner facing away from the raid (I'm guessing he does some kind of cleave attack on the tank), then whoever gets the debuff needs to move away from the group in order to get dispelled, and then when the adds arrive you pull them next to the boss and aoe them down.
Am I missing something because I can think of a number of WoW bosses with mechanics exactly like that. Hell, most WoW mythic bosses these days are far more complex than this.
Edit; it's also worth pointing out that this was recorded 4 months after the encounter came out.
Since this was released as a stand alone raid zone, with the earlier boss encounters being somewhat easier difficulty, what you are looking at is the result of four months worth of pulling the same encounter - potentially multiple nights each week.
By this time, they have the kinks worked out. And this encounter does have kinks to it.
That sounds interesting, can you elaborate on that? Who can attack the adds and what happens if the wrong person attacks them?
That's pretty standard affair for progression raiding. The question is, did it take so long to defeat the boss because of the fight mechanics or because players didn't have good enough gear to defeat it immediately? It's quite common for raid bosses to be released overtuned to the point where it takes the players some time to get strong enough to beat them.
The bosses you were talking about were released around the same time as the Ulduar Raid in WoW:
https://wow.gamepedia.com/Algalon_the_Observer_(tactics)
What I find interesting about this is that we have so much information about the WoW bosses and the tactics for defeating them, yet hardly anything for the EQ2 bosses. It's very odd.
EDIT: So I found a list of abilities for Byzola and while they are intriguing it's nothing amazing and certainly not anywhere near the dynamic PvE discussed in the OP:
https://eq2.zam.com/db/mob.html?eq2mob=c180700ec1b6f4891fcb55f02936875b
Reading the ability list it's pretty much as I assumed from watching the video. Kill the adds asap when they spawn, move away when you have the debuff and decurse as much as possible. It's your standard pressure fight where the focus is on the healers to keep everyone alive and correctly prioritise who to decurse while the dps burn the boss down before the raid gets overwhelmed.
Look, I'm not trying to bash on EQ2 raids or anything, but you seem to hold them in very high regard compared to WoW raids and I'm not sure why. Granted I haven't played the fights out myself but from what I can see the EQ2 raid bosses have very similar mechanics to WoW raid bosses. Really the only difference between raiding in WoW vs raiding in EQ2 seems to be the tools available to you. Even back in 2009 the amount of information WoW raiders had access to due to addons made a huge difference in the success of the raids.
Maybe. I understand you have little experience in large scale PvP but there can be a little more to it. My guild, FML, tries to accomplish the most with less and relying on players that understand their class and work together helped us be successful in BDO before we left in 2017. We were involved in node wars, sieges and even the open world scene. Now, yes, there were times when fights typically boiled down to just following a ping, engaging that inevitably turned into a big brawl but most of us knew our roles in the fight. You would have multiple groups setup with specific roles in the fight such has Main group, secondary, flex, cannon team, and defense. You would then build those groups based on who showed up, their gear and skill but mainly their class kits. I don't know if you had a chance to play BDO and get deep in the pvp scene or not but I would call it organized chaos.
I think boss self-learning AI would be perfect for large scale.
What you can see though, really, is a whole pile of information that simply isn't there. You can see formatting and such on the page where information is supposed to go, yet no one has put the actual information in there.
This is because not all of the information is there.
I mean, there is no information at all on the adds, what they do, what abilities they have or what happens if you get them wrong - and they are what make the encounter what it is.
If you find the encounter interesting based on the less than 50% information on that one encounter that Zam has, I'm sure you'd be somewhat amazed by the actual encounter.
Add to that encounter though, the fact that EQ2 had a much more involved combat system - so even if you were taking on the exact same encounter, there is more going on in EQ2 than in WoW.
---
Also, I'd like to point out that the group make up on that page is far from an ideal setup to take on this encounter when it was new. The group setup doesn't make it impossible, but it absolutely isn't an ideal situation.
This is relevant to me as it tells me that this was not a first or second tier guild that wrote this, which doesn't surprise me as first and second tier guilds didn't write strategies - they would share loot tables openly, but that is about it. While this is somewhat true in most cases, since this particular zone was released as a mid cycle stand alone zone, guilds taking on Byzola had already spent months farming Veeshan's Peak and Trakanon's Lair, which is the only content the game had to provide gear for this encounter.
Because SoE knew guilds already had the gear, they made this encounter less about needing that gear (well, you needed the gear that a first or second tier raid guild would be expected to have at that point - you couldn't go in with group gear and expect to do well), and more about executing the encounter damn near perfectly from all 24 members of the raid - the reasons guilds would fail this encounter were never assisted by having better gear. Its not all that odd.
In WoW, guilds use DBM, and that basically determines the strategy they use. Due to this, strategies are common knowledge. You could go from raid guild to raid guild fairly freely, knowing that all encounters will be done basically the same.
In EQ2, guilds use strategies specific to their guild. Guilds in EQ2 have access to an insane amount of information, but that information isn't telling them anything - it is simply providing them with the pieces to a puzzle that has multiple solutions. Guilds have to sift through that information to even find out which of the pieces that are laid out in front of them are a part of that puzzle, and which are there to try and throw guilds off finding a correct answer.
I've fought Byzola specifically in three guilds, and when I told my guilds main tank the method a friends guild I had done it with had used, he didn't believe me that it would even be a possible strategy - that is how vastly different individual guilds take on top end encounters in EQ2. Now, to be fair, it was near the end of that content cycle, and the guilds I helped out with had time to work the kinks out of the encounter in their own way, but the method they ended up with using was incredibly different to ours.
When you add the fact that the best loot in every expansion in EQ2 came from contested raid mobs (encounter is locked to one raid at a time, are the hardest to kill of a given content cycle, first guild to get the kill gets the loot), it is obvious - to me at least - why guilds want to keep their strategies somewhat to themselves.
I do have to wonder if the reason it is still somewhat undocumented - 12 years after the encounter was released - is due to not wanting to spoil the content for people. EQ2 does do fresh start servers on occasion, so there are still times when encounters like this are current content - and a full strategy would demean the content to a degree.
WoW simply doesn't have the puzzle aspect to it's raiding that EQ2 does. If I were to simplify why EQ2's raid content is better than WoW's, that would be it.
As far as I am concerned, you're questioning, not bashing. I'm sure similar questions would seem like bashing if they were directed at someone that didn't feel they could justify their stance and answer questions, but I feel I can, and questions are good.
I know I do go on a bit about EQ2's combat system and raid mechanics, and it is easy to dismiss it, but I'd like to point out that the game is still there, still live, and still free to play. It would take some time (and some friends) but this is all right there for you to go and experience first hand if you want.
Due to this, I'm being somewhat cautious when talking about EQ2, trying to play things down a bit rather than play them up - as I don't know how much free time you have right now (though I do assume a lot, right now), and I don't know how interested you'd be in attempting to use some of that time to prove me wrong.
It isn't like it is a dead game that other people can only really refer to memories of... anyone that wants to fact check me is absolutely able to do so, with enough determination.
---
I'd also like to use this time to point out to you that on top of the spell list for the one class in EQ2 that you have looked at in the past, each class has access to 4 different Alternate Advancement trees that pertain to combat (it may be 5).
These trees add new abilities, add new interactions between existing abilities, and even vastly alter existing abilities. There is also no one set build for each class, as the balance on them (at least when I played) was close enough that your gear, target content and likely group/raid makeup could completely change the build you would want.
So while you may look at the spell list (or the boss mobs abilities, in this case) and think that is somewhat intriguing, as is the case with almost everything in EQ2 - there is still another few layers to go.
In my opinion, organized chaos has it's place, but not in top end raids.
Top end raids should be organized structure.
Organized chaos is fantastic in PvP (good luck having anything else in large scale PvP, in any game). It would also work really well for open world events like monsters attacking nodes and such - and this is somewhere I do think AI could be used rather than a script.
if done well, this could almost emulate how PvP tends to have a meta flow about it. In PvP, you'll have a dominant build, which will then become very popular. Then someone will come up with a build that has a big advantage over that build, and people that are sick of the first build killing them all the time will then go for the new build. Then that second build will become popular, and someone will come up with a build that always beats that second build.
Similarly, if done well, AI on world even encounters could counter the last strategy used on them, forcing players to use a new strategy, and since (if developed correctly) the encounter simply couldn't cover every possible strategy players could come up with, and will always have a weakness, it is then up to players to find where it's weakness is today.
This could work really well in a public raid type setting.
Unfortunately when it comes to WoW, Blizzard's reaction to everyone using addons like Deadly Boss Mods was to increase the complexity of the fights to the point where it's physically impossible to kill the bosses without it.
The amount of communication and coordination required for this fight just wouldn't be possible without addons and macros since you effectively have 3 teams all needing to communicate at the same time.
The result is that WoW raids are less about problem solving and more about executing a known strategy.
such as ensuring no two fights are exactly the same, so "Farming" a Boss
doesn't get boring.
But at the same time, if the AI is learning from every encounter, from every
player/group, then eventually it will be unstoppable since it will learn how
to counter every Ability thrown at it.
The only way to keep that from happening, I think, is to store information on
specific group makeups so that it doesn't remember what it learned from the
last group, but it does remember the stuff it learned from your group. And even
then, it will "forget" after a certain period of time, like a month or so. Preferably,
a random amount of time so that you can't keep track of what it knows and what
it doesn't.
Most guilds I've been in form a core of 4 or 5 people who do most of the problem solving. This usually includes the raid leader and main tank, and then who ever seems to have the best aptitude and/or free time. In some guilds I've seen people taking logs from an evenings raids with them to work so they could look them over in more detail.
I've said before on these forms that raiding at it's best is the developers creating a puzzle, the players solving it, then the developers creating a new, harder puzzle for players to solve.
Blizzard never understood this. Even before DBM, they never treated raid content as a puzzle for players to figure out.
On the other hand, SoE kind of invented raiding in a 3D MMO (well, they kind of invented the 3D MMO). They understood, eventually, that it shouldn't all just be group content but with more people, even if the early raid content was exactly that.
There are potentially things other than treating it as a puzzle that developers could do with raids to make them distinct from group content other than player count, but I have yet to come across anything else that developers have even tried. Based on that, it forms the basis of what I consider raid content to be at this stage.
If I take my definition of what top end raiding is - a puzzle presented in the form of a multi-group encounter by developers for players to figure out - I then have to say that WoW has no top end raid content at all. It mearly has group content for larger groups. This is why I continue to disregard WoW raiding.
All else aside, the encounter you linked doesn't look enjoyable at all. It doesn't look hard, it just doesn't look fun, either. 20 solid raiders and I would expect to have this encounter killed on our second evening at the latest.
It is adding unconnected layers to the encounter purely for the sake of making use of DBM and such. While I am a fan of players having tools and developers basing content on the assumption that players will use these tools, the whole thing needs to be done to deepen the puzzle (as a combat tracker will do), not to remove the puzzle (as DBM does).
They did certainly popularize 3D MMOs though, and it was something of the gold standard until WoW. EQ was my first MMORPG, and I loved the heck out of it.
However, I did miss-speak (miss-write?).
EQ was the first commercially successful 3D MMORPG. Meridian has effectively been dropped by two publishers - 3DO and Near Death. The only reason it still exists is because it is free to play and doesn't cost the current owners of the game anything to have it listed on Steam - although as the source code is open now, I'm sure there are also some fan based servers for the game as well.