I think there will never be a fully automated and 100% effective anti-cheat solution, even console gamers can use aim assist devices like the Cronus Zen, and false positives seem to be a much worse problem in client side anti-cheats than on server side ones.Indeed that's a good idea, as long as what you're actually tracking is an aimbot (on target percentage can be estimated to be uncharacteristically too high for objects moving fast at a close range, or indeed behind a wall).
If you aren't tracking for aimbots but just players the wallhack alone by the potential FOV and being behind a wall, I could see the potential of this producing false positives for players who are just plain good at listening for players behind walls and predicting movements.
Under that premise, the best solution is a combination of getting a decent number of cheats detected automatically + getting a small number of cheats reported by players or flagged as suspicious due to improbable statistics and manually detected by support staff + getting a really small number of false positives that support staff can quickly manually review and revert.
Most companies don't even try, they spend 0$ researching server-side anti-cheat measures. The few cheaters I have seen on PC are not even slightly difficult to detect server side, like people teleporting, flying or moving at absurdly high speeds in Battlefield games.
I think machine learning could be of huge help in this matter, as a neural network properly trained could identify players behaviour when using a wallhack with a high degree of accuracy, for example. But it is easier to infect the player's computer with rootkits that don't even require per-game implementation and are mostly generic.