Every single announcement/patch, there is always seemingly crowds of people expressing dissatisfaction about buffs/nerfs not being done properly. Isn’t this at least partially caused by internal human biases from developers by the classes they favor more than others? What if that can be taken out of the equation so no human can influence the buffs/nerfs at all?
What if Blizzard started implementing AI into their balance system? For example, imagine having all the pit clear times and skill damages from each class pooled into an AI. Then, dynamically scale up and down damage for skills in REAL TIME until an equilibrium is met. Over time, every single class and the skills they use will basically reach a relative balance so the tier lists will naturally auto balance itself, and this includes everyone trying new variety of specs.
Since it takes all the data in real time and adjusts accordingly. Pits will essentially be forced into perfect class distribution at all times. Every class can clear just as well, and can top leaderboards just as equally because no biased human decision is influencing the buffs/nerfs. We just let AI handle all the balancing and force it.
Essentially, what you could end up with is a game that literally requires ZERO balance patches until the end of time because it is self balancing. Blizzard can just point to the tier lists and say “look, everything is in perfect balance, as the data shows” and never have to release a balance patch ever again.
2 Likes
AI sucks. I tried Google copilot and about 90% of the code it suggested was wrong or sometimes it was very close but introduced bugs and security holes. Not to mention AI would send all their proprietary information to a server they don’t control.
5 Likes
First, its “Microsoft” not Google that has co-pilot.
Next, Blizzard would be keeping the game data numbers on their own servers and running their own balance algorithm, not sending it to a 3rd party to calculate for them is what I would suggest.
well, google’s AI thing told people eating rocks is healthy, I’d say we should wait on giving AI any actual desicion making
1 Like
At this point, I feel like with how balancing is going, I don’t mind giving a different method a try.
Internal biases towards certain classes are really screwing things up. We need to somehow eliminate that from the decision making.
1 Like
The less AI in the world, the better.
1 Like
I am against replacing humans with AI.
Automating a system to react to players means there will be ways to manipulate said system to cave to dark intentions.
You would be giving way to competitors to look into automating some player behaviors, to then manipulate the AI to do their bidding and lose retention.
AI doesn’t have morality or knows what it’s doing. It just has a task and an expected output.
The quality of the game will heavily be affected in the negative sense, if you adopt this approach.
If they (humans) did something similar with the PTR data before the season started and caught an underperforming class, that would be more what I would like to see. Maybe we’d have a season where they’d vow, “No class left behind!”
That might explain why our class balance is so poor. Maybe they are using AI to balance the teams now. Anyway, today they discussed class balance. They keep saying we will get a class revamp in the future, but nothing can be talked about now. That is so frustrating.
There are things we can do between now and the future. Why are we updating and balancing one class at a time? In D3, we could do a mass class balance in the PTR every season, yet in D4 we are moving backward. Will future expansions or D5 result in them saying they will now balance one item at a time based on this trend?
1 Like
none of this is a concern atm is how i feel, the errors in the code on these hotfixes shows me they got all there eggs in the expansion basket atm, im still blown away they are selling an expansion with how in balanced the game is atm specially with a new class lol.
Good things come to those with patience. Didn’t the marshmallow test teach us that?
Balance in D3 was changing some numbers in a spreadsheet for a long time.
Balance in D4 involves changing mechanics and functionality beside the numbers game, which requires a greater deal of coding.
What do you mean?
Would you go to the baker and ask for the dough before it becomes bread?
The real joke here is that people still think the classes need to be balanced even though it’s only the items.
They could just add a unique that adds +X% “Skill” Dmg instead of buffing the skill.
As an example: instead of buffing pulverize to 70% dmg they could have just added those 70% to the shockwave aspect instead of the tree.
The system is bad thats the problem.
Now the real joke is that they have skills in this game that don’t even work until you use an aspect. For example Tornado doesn’t work with the skill tree alone. You will need stormchasers aspect.
A skill shouldn’t require an aspect to function.
The AI doesn’t have to have any morality. It just has to complete its goal of balancing.
Skill damage, cooldowns, etc. will be adjusted in real time based on the entire population median data harvested until the pit tier list has an even distribution of =/- 10%. Then it holds it there.
The dark intentions you talk about would require these forces to purchase millions of accounts just to automate them into doing what it wants to sway the algorithm to adjust the way they want which is economically unrealistic for anyone to really do. And even then, what is the result? Having some potentially super OP build which the rest of the population would eventually catch wind of and spec into, and thus plummeting it back down when the algorithm kicks in and rebalances the tier list?
The AI algorithm would essentially block such abuses long term anyways because its self balancing.
They would end the command with “keep xxx and yyy classes in 50% deficit”.
If you are interested in balance, I’m pretty sure you can guess the classes.
The problem is not a capability thing. It is intentional. AI will not help.
maybe at the very least, the AI’s math would be more accurate?
idk, i think i’d prefer an ai to do balancing instead of leaving it to some suit who just wants to leave work early and play golf instead.
You can use AI to suggest changes, but unless it’s trained on how everything works, and they intentionally specify what they want to change and provide statistics, all of which is possible, but unlikely, then it would be useless.
AI isn’t actually intelligent. So chances are it will give you nonsense and probably be even worse.
The developers have an idea, and they have changes that aren’t announced, things that are coming in S5 that aren’t in PTR, and things coming in VoH and S6 that we don’t know about.
For all we know, these balance changes ‘could’ be well placed, although currently they do look awful.
AI wouldn’t be able to easily take any of that into account though, so it would just a bad idea.
1 Like
But if AI only got to “suggest” and not “do”, then all we would end up with is the same developer bias.
AI will suggest something like Sorcs need about a 300% buff and Barbs need a 25% nerf.
And Blizzard would be like: “nah, we aren’t doing that. Nerf Sorc damage by 25% instead, and buff barbs 15% more this season”.
Then we don’t solve any problems. We have to forcefully remove human bias out of the picture in order to obtain true equality.
It would be more like:
Sorc needs to be able to dual wield 2h scythe and gain barrier when they claw an enemy.
1 Like
I’d be fine with new builds. Kind of like a 2H scythe based arcslash build?
Adjustment in real time is a big NO for me
Adjustment is something you do in a development/testing phase or at discrete and large intervals, not live.
What you do now as a player can then be tomorrow invalidated.
This means whatever gear you’d collect now is meaningless.
This means why even play the game.
±10% performance overall
Having the ±10% performance overall means you lose the build tiers. If the problem is dev bias, then management/philosophy needs to enforce healthy balancing. This means give every class an opportunity to shine, have each season spotlight a different class at each builds tier list. Tier lists would be adjusted based on types of content you want people to run in your season.
e.g. I like that we had open world (blood tides), then dungeons (vaults) and now open world. It’s healthy to give each system it’s time to shine, as long as that system is fun. It’s clear for me by retention and implementation that people enjoy open world more, for example. I was happy to leave S3 after season journey. I don’t even want to recall what S1 looked like.
AI can still be biased
AI itself cannot do anything without a proper prompt. AI needs revised prompts. And who does revised prompts? Humans. Again, you’re losing to human bias. I will never trust AI to dictate what my gaming experience will be. When it comes to balancing, we need democracy in design and rotation. No 1 person gets to dictate what’s the strongest class for more than 1 season.
We don’t need AI
- We need a healthy balancing philosophy and not sticking too long on 1 class/build being powerful, across the board. We need an algorithm that checks for this healthy balance, not enforces it.
- We have humans playing the game
- human feedback says what people want - it is expressed, but is it enforced by their actions?
- data says how they react, if you should listen to them again or not for now (this means that later changes reset their “take into consideration” allowance on a topic)
- Sometimes, you need game design politics to keep a game healthy and allow other builds to shine (and not give players always the same food, so to speak)
- Right now, it feels like catering to a vocal population (looking at you, Rob and barbs) that then spirals into the common player taking that popular build because he has no other option if he wants to do certain content (pit being a funnel). I enjoy barb, but I definitely don’t want 1 class to be top for that long. It’s just not healthy for the game.
- Blizzard needs people that deeply understand the classes (a impair number of few people for each class, so it’s never up to 1 person, with the majority consensus of a class giving a go to a change, then another group that checks for balance across all classes, with proper testing tools). I understand why Adam said “let’s have real conversations” and mentioning interviews, in that they’re also not happy with class balance.
- Every conviction that a class designer has could be validated both by theory and practice:
- knowing all the damage equations and/or having a “maths team” to give them this feedback
- understanding the code flow (emphasis on flow)
- experience: with hired testers that reflect most personalities of the mass of players; I can’t trust some streamers to not be biased towards what they like to play instead of a healthy game because you’d obviously want people to play what you play too, for more views and some are in for the business and popularity (that distorts their reality, they create their own bubble) - to some extent, they can become lobbyists
- validation should not be a blocker, but a reality check…and if your decisions hurt retention and ROI considerably…well that person need to change what they do, like yesterday, not now, in terms of urgency
When a release goes to the public, it is read as “Ok, judge my work”.
What I see is “here’s the ball, where we are, see how long you can play with it or if you play it at all”, not a conviction that “I’m so happy to share with you the ball (the game)” (and one shouldn’t be a people pleaser and lie).