Original Post — Direct link

Edit: Sourced and annotated for people who like to diffuse arguments by screaming "WHERE'S YOUR SOURCES!!!! MY ASS!!!"

Premise: Cheat makers won. Underground discord servers selling $400 a month undetected obfuscated cheat code and people running DMA card cheats feeding out to a second monitor etc. PUBG devs knew the cheaters were out there but the old tried and true method of "find process OMGLEETHACK; ban player" wasn't working any more. Cheating is big business and cheat providers have grown as sophisticated as their state-sponsored hacking counterparts. (9, 10, 11)

So, PUBG started using the player pool and statistical analysis. All the various data is kept for all the various games leading to a massive dataset. The old dataset and previous gen banlists can be fed into an AI learning model, and the AI is taught by giving it a reward when it statistically catches a cheater was previously confirmed as a cheater. The model uses interesting data sets such as movement and positioning around the map, proximity and movement of teammates, etc. (data well beyond traditional statistical analysis of headshot %'s and ADR). (3, 5, 7, 8)

PUBG dev's pat themselves on the back, wrap up the AI learning model, put it to work on live data coming in the day's matches, and they all head down to the bar to celebrate a job well done! (4)

Here's where things go sideways. "Camp AI" and the more conservative with faith in humanity put cheating at 3%; surveys of real people puts cheating at anywhere from 12-60%. A decent number of competitive and regular hours players have been using private, undetected cheats for years. Anti-recoil for starters, radar hack, wallhack, humanized aimbot, etc. Having gone undetected, PUBG devs have just unleased an AI Police Force that is now blind to the "Tryhard cheaters" - the guys cheating on a main account (or a second account aged to look like a main account) who work hard to pretend they aren't cheating. These players wouldn't snap to the head of players hiding behind a wall (that would get them reported and banned under the old manual review system); but their squad would definitely "wander over to that part of the map on a hunch" and always just happen to be looking the right way, right when it counts. (1, 2, 8)

And now that the AI Police Force is blind to them, their numbers are growing. Guys who used to have to be careful all the time due to the manual report and review process, can now happily scream around the map scoring chicken after chicken after chicken after chicken. Players report them; the AI Police take a look at the player, and determines the player is clearly legit based on its learning model. (4, 7, 8)

Hidden cheaters slowly start ramping up thier illicit recoil control and other skullduggerous methods. Slowly the AI Police force continues to reinforce that those illicit methods are legal.

The rage cheater population diminishes somewhat due to AI Police being able to quickly swoop in even without a reporting system to ban the rage cheaters. But the epidemic of hidden tryhard cheaters starts to slowly grow. The remaining honest players don't know they're playing against cheaters - most people are nice, honest, and willing to give the benefit of the doubt. Some of those players get frustrated, and just like your buddy at the gym on roids, start to dabble in anti-recoil. They're able to keep up better. But the rage cheater population is shrinking and the hidden cheater population is growing, thanks to selective breeding by the AI police. Casual players drop off for other games, competitive players buy new mouses, bigger video cards, move closer to the datacenter, get on ADHD medication and up their game. Pros and streamers who can't cope, cheat. The competitiveness of the game goes sky high. Pretty soon some of these casual players, some who may have never cheated, ask their best player in their squad how they stay competitive. Small-time streamers watch their revenues slowly drop off monthy by month. One by one in confidence these players are directed to underground discord servers filled with cheat providers for every game imaginable. The chatrooms are filled up not with hundreds, or even thousands, but tens of thousands of players subscribing to cheats to various games. (6, 8, 12, 13, 14, 15, 16)

And so this unseen population of hidden tryhard cheaters grows day by day. The line between competitive player and hidden cheater becomes razor thin. Every once in awhile PUBG devs come back to the office and feed some more recent learning data into the AI Police model. And the more recent data again is taught to ignore hidden cheaters (who weren't detected) and to focus on rage cheaters (who were never a problem in the first place because even the manual process and gross data analysis called them out). And once again the AI Police goes back to work on detecting cheaters in the live pool, and turns a blind eye to what's now surely a playerbase consumed by hidden cheat assisted playing. More players who refuse to just give up and find another game find themselves in a seedy discord, trying to figure out how to buy bitcoin... and the cycle continues.

At this point there's no way to fix the data to correctly teach the AI Police, but devs have already fired the teams who processed reports from the reporting system. You wonder why the reporting system was expanded... it was expanded to feed addtional data into the AI model. The experiement failed. Welcome to the new age of cheaters. Hail Spez!

-Consider the player pool is reportedly around "400,000 peak per month" and has been for almost 3 years

-Consider the rate at which number of banned players has shrunk compared to player pool circulatiojn

-Consider that the banned cheaters aren't leaving the game... they're just getting smarter.

-Consider that players staying in the pool are: a.) pros or b.) moving from the rage cheater to hidden cheater category, increasing the competitiveness

-Consider that new players entering are getting absolutely crushed and either quit or join the rage cheating pool

-Consider that the overall % of cheaters would grow in this self-mutiliating life cycle, especially when incorrectly trained AI Police were given total control over the population

In a tangents, next time you are on twitch, consider some up-and-coming streamers who employ one cheater in the squad who helps carry them to final circle time and time again. A player who, if they are the last one alive, tends to kill themselves right away to save the lead streamer the embarassment of being caught. A player who's always making awesome callouts just before the enemy comes over the hill, for a streamer that otherwise seems generally clueless about where to point their gun. Plausible deniability is all that is needed. Especially for fapping Chads watching their totes faves gaymergurl streams. I don't care what you do baby just keep clicking that mouse...

Sources were requested:

(1) Per 2023 survey, 61% of all respondents have been caught cheating - https://time2play.com/blog/is-it-cheating/

(2) Per a 2018 industry survey by Irdeto, A staggering 25% of players "Sometimes, often, or always cheat on-line" - https://resources.irdeto.com/irdeto-global-gaming-survey/irdeto-global-gaming-survey-report-2?_gl

(3) Anti-cheat used to ban about 1,000,000 players per month back in 2018 - https://www.destructoid.com/pubg-has-banned-over-13-million-users-in-the-past-14-months/

(4) PUGB Using AI Anti-cheat, 400,000 accounts banned per month in 2023 - https://www.ggrecon.com/articles/pubg-is-using-ai-to-hunt-down-cheaters/

(5) According to researchers, a mode exists for cheating to infect a gaming ecosystem; estimating cheater prevalence by data analysis is difficult - https://www.cambridge.org/core/services/aop-cambridge-core/content/view/5FE990B414BD1CE5D57B7BE7E0645F36/S2050124221000199a.pdf/cheating_in_online_gaming_spreads_through_observation_and_victimization.pdf

(6) A disturbing survey by Kaspersky shows only 68% of eSports pro are fully against cheating (what's up with the other 32%?) with 91% wanting improved anti-cheat, indicating pros are willing to fight fire with fire and supporting contagion theory - https://www.kaspersky.com/about/press-releases/2022_e-sport-committed-to-fair-play-and-transparency-68-of-esports-pros-against-cheating-to-win

(7) AI anti-cheat research paper with efficacy estimates for certain methods -https://www.cs.cmu.edu/~jweisz/courses/docs/weisz03_DetectingCheaters.pdf

(8) According to researchers, a mode exists for cheating to infect a gaming ecosystem; estimating cheater prevalence by data analysis is difficult, puts statistical cheating at 3% which wildly disagrees with anonymous surveys - https://www.cambridge.org/core/services/aop-cambridge-core/content/view/5FE990B414BD1CE5D57B7BE7E0645F36/S2050124221000199a.pdf/cheating_in_online_gaming_spreads_through_observation_and_victimization.pdf

(9) Cheat dev AMA - https://www.reddit.com/r/VACsucks/comments/erkh8f/ama_i_am_a_cheat_developer_ask_me_almost_anything/

(10) One cheat maker fined $4.5M - https://www.engadget.com/2018-04-30-chinese-pubg-hacker-arrest.html

(11) Catfish claims $350,000 in monthly revenue selling cheats - https://www.svg.com/772747/the-untold-truth-of-the-biggest-game-cheating-ring-in-the-world/

(12) Pro ban article link 1 - https://esports.gg/news/dota-2/valve-bans-dota-2-players-including-knights/

(13) Pro ban article link 2 - https://liquipedia.net/pubg/Banned_players

(14) Pro ban article link 3 - https://www.pcgamer.com/pubg-pro-cheating-bans/

(15) Pro ban article link 4 - https://www.techradar.com/news/dirty-dozen-pubg-esports-pros-get-banned-for-cheating

(16) Pro ban article link 5 - https://kotaku.com/pubg-bans-pro-players-30-000-others-for-cheating-1831623073 and https://twitter.com/PUBGEsports/status/1079844118136729600

(17) PUGB Ban count currently at 60k-100k per month - https://pubg.com/en/news/6471

External link →

Okay, since you don't know what the CCU Peak number means, let me explain

https://preview.redd.it/lbzcf8yepxrb1.jpeg?width=1053&format=pjpg&auto=webp&s=7be491132de289d5dd89e3d32d2c84f248a40e09

CCU stands for concurent users. Daily CCU peak means that at one point during a day, there are 400k players online. This means just that. It is not a total number of players every day, let alone every month. It is a total number of players every day at 10:24am UTC (made up time).

There are millions of players world wide constantly coming in and out to play for 30 mins, 1 hour, 2 hours, 10 minutes - constantly logging in and out every month.

There is no 400,000 players (of which 200k start their gaming day at 5am, waiting for other 200k to come in at the 10:24am, then slowly going down to 200k later in the day again. That's not how this works.

Players, especially the mass % of casual players, will not play for more than an hour or two, not even every day.

There is no official source for the MAU (monthly active users) number, but when there is a peak of nearly half a million players world wide EVERY day, that indicates that there are millions of players world wide, dozens of millions.