Let the cat leave the house to hunt mice. 
How many of those 10,000 accounts belonged to rage cheaters and players who downloaded free aimbots from github or tried other stupid internal things?
Only stupid mice enter houses where cats live.
Statistics on this would be interesting. How many of the detected cheaters were using high-quality cheating software?
And how were they detected? Probably exclusively via machine learning models and player reports. In other words, this player can come back with the same cheat and a new account and possibly play undetected again.
Are behavioural analysis and machine learning used for players outside GM and Top500?
According to “rumours”, these techniques are only used for high-ranking players.
(this is about adding a kernel component to your anticheat, we need that!)