Not true at all.
Of course data like centipawn loss/accuracy is taken into account, but there is a multitude of other data they mine in order to make informed decisions.
For example, most unsophisticated cheats simply relay from the engine GUI window to the browser window, which generally always takes the same amount of time. Even in highly complex positions. So if somebody uses the same 7 seconds to find a perfect move in a hairy, complicated position, this is taken into account. Average move time is a big deal. Humans spend more time on difficult positions, engines dont need to.
Then there is "in the window" time. When somebody leaves the liChess window (as an example) the green "online" light beside their name goes out. This indicates that they are outside the board window. The frequency of this "out of the window" behaviour is taken into account.
Then there is "base accuracy". If your average centipawn loss over 1000 games is 24 and you have a good week and average 20, they will take this into account and not ban you. If your average centipawn loss over 1000 games is 54 and you have a string of games with 20 centipawn loss, they will take this into account and if other variables are also troubling, they will ban you.
You get the idea. There are dozens of variables in their algorithm, and ban decisions are based on the sum of all of them and the moderator's assessment of them.
So in my own case, sometimes (especially with some good coffee and some good music lol) I will steamroll through opponents one afternoon with multiple "perfect" games.
But I won't be banned becauase:
1. My base average is decent
2. I think a lot more in complicated positions
3. I stay in the window for the whole game
4. I consistently play more or less the same openings
And so on.
If it was all about not making tactical mistakes, they'd never catch titled cheats!
