You’d be very surprised how much of the industry hasn’t discovered this. There’s lots of stats that can be used to reliably detect cheating.

* New account

* K/D ratio that’s multiple times better than the best players in the game

* Using incredibly off-meta “cheater builds”

* Consistent record breaking total kill count

* Inhumanly low TTK and headshot ratio

* Speed hacks that allow you to move around the map much faster than you should be

* Time to lock on/switch between targets

All of those are incredibly easy to statistically analyze. I play a fair amount of Warzone, and I often run in to cheaters that would fail a very basic “definitely cheating” stats check. You can even get a rather reliable indication of whether there’s an obvious cheater in the lobby by how quickly the surviving player count goes down.

A 3rd party implemented an app that looks up the stats of every player in your lobby, that players were using to detect the more obvious cheaters. There’s no non-cheating explanation for a double-digit K/D player getting 50+ kills for the past few matches in this game, but you’ll find players like that in your games routinely. The developer response was to declare the app against ToS and modify the public end points that provided those stats, breaking the app.

That’s one of the biggest games in the world, they’re clearly not utilizing even simple statistical methodology, and they don’t want users doing it themselves to decide what lobbies to back out of.

I don't think that level of snark is really warranted here.

Valve does this though. I'm pretty sure their VACnet is more effective than their classical anticheat. It's just a lot more effort.

VACnet IS more effective than their previous system, but that's a LOW bar. CS:GO is still full of obvious spinbotters and wallhackers that get to ruin a hundred competitive games before eventually getting banned, and just grabbing a new account. VACnet for some reason can't even get the obvious stuff; Spinbotting looks nothing like a normal person playing, to the point that you could probably write "if average rotation rate > some high bar then ban" and do better

Yeah except he's right, the vast majority of cheaters can be detected server side.

Literally no game does this, because ?????

Battlefield has been doing this, with FairFight, and it's worked okay-ish, but insanely slow

Sounds like a startup opportunity. Lookin' forward to the pitch deck!

What kind of massive startup ROI are you expecting to make from a highly niche bit of middleware?

Just because something is possible doesn't make it profitable.

Worse the game publishers make more money from cheaters than from legit players.

Cheaters have to keep re-buying the game when they get banned.

The incentives are for publishers to ban cheaters... but not too fast... and that's exactly what we see happening in practice.

Kinda interesting huh

I doubt it. Especially since the games where cheaters are most rampant are f2p.

Video game industry is massive, and cheating is rampant and ruining peoples experience in game. I don't think it's the craziest idea

Valve have been doing it with Counter Strike for at least 3 years. And the "VACnet" they have is still far from perfect. Especially with the less obvious cheats. They used to be a talk on youtube, but it seems to have been taken down.

`\dkliksphilip` reuploaded the talk it seems: https://youtu.be/SnRgW54EWwA

I did a few hundred cs overwatch (it's a system where people with 150+ won matches can review suspicious behaviours reported by system/other users - https://blog.counter-strike.net/index.php/overwatch/) cases and 95% of them were people using Spinbots. You can write a program/script that parses demos and detects those in few hours (assuming you don't know how to parse a demo file). Why is this even a thing? I want to help them with fighting less than obvious cheaters but if almost everything I get can be automated by any 1st-year cs college student then what is the point?

I am guessing extra computation? Maybe this sort of technology is in infancy?

The real interesting take away is it is always a mouse and cat game. The hackers will adapt, instead of reading memory they might just read packets, or use a m.2 memory reader card, because they have to -- and so will the anti-hackers. I am interesting in machine learning to simulate COD -- I saw a video of it and it's like a young child was playing the game -- not good but obviously on the path to competence.

They do. And after a few weeks/months there will be a ban wave. But there really isnt incentive to fix these problems because if you ban cheaters they have to buy new copies of games.

Because they can get people to install rootkits on their PCs and offload the computational burden on to them instead of paying $$$ to detect these themselves?

CS:GO does do this.

In one part because the false positives of picking up top players as bots becomes an issue.

"vast majority" means? 60% or 99%?