Hacker News Reader

Comment by gabereiser | Hacker News Reader

At an individual level, if a player goes from 1.3K/D avg over a sample and then it jumps to 8K/D over the same sample duration later, it’s pretty obvious.

Wall hacking and aim bot can be caught with behavior and statistics. Not in real-time but damn near.

View on HN

Not quite. I understand that's a piece but if you've ever tried to solve this problem it's not even close to being as easy as you make it out to be.

1. The stats you cited (k/d) is well within normal variance

2. It's not super common for cheaters to run around spinbotting/blatant wallhacking. I would actually say that's the minority.

3. Depending on the game, that would never happen because MMR just matches you up with people based on your new "skill" i.e. 2 games with 8K/d puts me in harder games and I go back down to something reasonable. Again, most cheaters don't continue to just dominate, those are the minority and get caught easily.

So it’s not as trivial as I make it sound, yes, but that’s not to say it’s impossible. People are working on just that. These things wouldn’t get you banned, but these are “early warnings” that you’re cheating and warrants further review of game footage (which most games store in the cloud now). It’s not trivial, but it’s not rocket science. AI behavior patterns combined with game footage review combined with a judiciary would cause one to be banned. The future of esports though is in closed edge networks with dedicated hosts. Provide them with the Internet cafe appliance. I just want to play some Counter-Strike. If I were to average my K/D over 1000 games I would get a pretty good statistical spread of how I play on certain maps. Granted theres a degree of variability depending on my mental state, but still a good enough sample to get an idea of how I play Dust II. Sudden and consistent outliers above norms would flag me for review. 2,3 kill cams and footage review would tell you that I’m cheating. Someone in Burma or Thailand could do that job until the AI is trained to spot aimbots, cough I’m sorry, “flick shots”.

Measuring inter-account variability in performance doesn't work for players who only play with cheats on that account

Damn smurfs, good point. I would be defenseless. In reality you would go with a multiple angle approach but you are certainly correct. Behavior modeling would suggest you’re the GOAT. 100% of the time.

My own stats in games have fluctuations more extreme than your example.

Some days I’m multitasking while I play games. Some days my mouse battery dies mid game. Some days I’m really sharp and I’m up against people who aren’t as good. Some days I’m not as good and I’m up against people who are really sharp.

Wild fluctuations are normal for a human casually playing games.

People who play video games high or drunk shouldn't be punished when they're finally sober. :)

Which is why you sample an average. If you play high or drunk, it’s probably a pattern.

If only there was a way to teach computers to identify patterns based on a large dataset...

Yes. Obviously what happened is that the player is away - on holidays, in hospital, etc. - and their younger sibling uses the opportunity to play the game on their older sibling's machine. And the younger one is much better at this class of games.

That, or any other of many scenarios that go against the idea that an account is used by one and only one person.

Diablo 3 ladder top 20 or so on Nintendo switch’s

You can see that gear is hacked items. Invalid stuff.

Nothing is ever done about it. Absurdly blatant, and discouraging.

I’ve made top 10 on hardcore at early portions of a new season before. But only before these regulars get online and instantly get to the top.

Reply on news.ycombinator.com