It’s 2023, the technology to collect data en masse and continually update the game until all damage and healing specs are performing equally in PvP and PvE is here and has been here a long time.
A Neural Network collecting parses in PvE and examining class/spec distribution in PvP ladder rankings is the solution, applying hourly micro patch updates.
It would be real simple. Major class abilties/mobility/CD’s would be handled by humans each month as they are now.
Meanwhile, in PvP, for both 2v2, 3v3 and Rated BGs and Solo Shuffle (4 different brackets), the Neural Network would simply apply a class/spec unique buff/debuff of +/- damage and healing while in those instanced PvPs, and this buff would be different for each of those four modes. The cost function of the Neural Network would be simple “The Net Difference between all DAMAGE classes and specs in top ladder ranking” with the goal being to achieve uniform distribution of all damage specs in PvP in all 2400+ ratings; likewise the same cost function for healers (and tanks as well) run independently from Damage.
The same cost function would be run for instanced PvE (mythic and mythic raiding) by checking damage and healing performance of each class and spec in the top tenth percentile on raid logs.
Thus the Neural Network would quickly achieve balance, while major class overhauls )(such as mobility, haste buffs, GCD triggers, etc) would be balanced by human developers.
It would only be examining the top tenth percentile of PvE parses and PvP rankings, neither of which would contain bots.
As for actual players cheating with bot assistance in PvP…ironically it also be balanced, making all cheaters equal mathematically, and perhaps, with a Neural Network, easier to detect. If the NN is picking up kicks/stuns from player that either never miss or always come out between 0 and 1 server ticks, it would…know.
You’re aware that mathematical analysis is always done with 99% automation via Excel or Mathlab, even when a human is doing it?
A Neural Network cannot be fooled or hoodwinked. It would have PvE balanced in less than 24 hours and PvP within two weeks (relative to the integer number of players participating per hour).
What you’re afraid of is there not being a FoTM class.
That sounds like a lot of work for a game that’s already pretty balanced if people would just stop idolizing the top M+/Raiders and just play w/e tf they want.
The Code is this simple for PvE for Mythic+ damage dealing (writing in human terms for normies to the best of my ability).
Let N be the number damage class and specs.
Let D be an array of length N for each damaging class and spec.
Given the top 10% of all players (top decile) for element of N, how much damage are they doing on: 1. All Trash Pulls in Mythic Plus.
2. All Bosses in Mythic Plus.
Let T be the Trash Array of length N.
let B be the Boss Array of Length N.
Let each corresponding element of T be the average damage done to trash mobs in the top decile.
Let each corresponding element of B be the average damage done to bosses in the top decile.
We now apply +/- damage buff separately to trash and to bosses to achieve uniformity in arrays T and B, until all elements of T and B are equal to average of the initial values of T and B (attuning the change to the average of the Initial T and B array assures that this nerfs the strongest classes by bring them down to the average and buffs the weakest classes by brining them up).
A secondary function would be added, such as “We wish to achieve a negative hyperbolic distribution of Keys completed from 10 to 30, such that we translate the mean buff/debuff by X% and apply an additional and uniform modifier of F(y) where y is the key level, to all damage dealing specs.”
Ignoring that overall damage numbers aren’t comparable from 1 run to another, even of the same instance, so we’re starting out by giving the bot essentially junk data.
If you have a specific scenario you want all the specs to be balanced in, it’s not that hard to hit for a person. If you want perfect balance across all scenarios, all specs have to work identically, just with a new skin.
What makes you think a balanced game would help blizzard retain consumers at a higher rate than one with variable performance?
At least with shadowlands covenants there were opportunities to play off meta and excel in niche scenarios, with current tier implementation and talent system most classes are locked and have very little if no rotational variability.
Next season is going to be lit… if blizzard decides to give players the ability to choose between the 3 tier bonuses offered this expansion
People would just figure out a way to game a system like that to their advantage.
For example, organise to have a very large group of a single class log in at once, spec into the worst possible talents and deliberately underperform until the Bot applies an aura buff.
Classes and specs don’t have the same abilities as others.
However, each class/specs ability to complete content should be identical to all others.
This is for mathematical performance, not class design.
For instance, as a hypothetical, let’s say that Fury warriors had -30% debuff and Shadow Priests had a +30% buff in Mythic+ from the NN.
It is clear then that humans need to intervene and figure out why should gargantuan buff and debuff had to be applied to begin with and balance the class design itself.
You’re not going to get the top 10% of PvE players to participate in such activity, you would need tens of thousands of players working in coordination to do this.
It’s also only going to check the top 10% of performing players in those classes and spec. So if they did this, the next tier of players (second decile) would override them.
Third, the Neutral Network is adjusting to the Initial Mean Value of the T and B arrays given in the example…so this wouldn’t even work notwithstanding points 1 and 2.
Lifetime bans would go out to this hypothetical conspiracy of tens of thousands of players if somehow did achieve this.