[Gareth321]

I'm more sceptical about inference based detection methods because as they improve (using AI), so too will the cheater's ability to fake human movement. It will be trained on real humans and mimic how real humans play - just at the very high end of the range of skill and ability. Developers will be loathe to ban "good" players just because they're good.

[jimmydorry]

If cheaters were capped to mimicking good players, that's already an incredible win over the status quo. The players that are walling (as an example), are playing with more information than they should and this should always be detectable with enough observation, especially in terms of them displaying super-human reaction times and being pre-positioned to their advantage... so I'm not quite as pessimistic as you are about this not having good returns.

I'm sure there's a reason why they don't, but I wonder why games don't try implementing honey pots, like rendering a fake player behind a wall and automatically banning if a player's crosshair snaps onto them, etc.

[Gareth321]

It's true that it would be an improvement. I should not let perfect be the enemy of good. Your honey pot ideal is one of many solid ways to detect cheaters. Developers appear more interested in selling copies than they are ensuring players have a good time. Perhaps the motivation is more aligned in subscription games, where they care about the recurring revenue.