doesn't make any sense why they would need to "learn" your pattern and match it with other players.. blizzard is way too lazy and cheap to code something like that and it doesn't even seem necessary, if they are really checking your waypoints then.. lap 2 same pattern = busted.
this thread is just trying to sell profiles.. I wouldn't have anything to do with Chinese people in wow, they are nothing but scammers and leeches
as for questing, the profiles are not perfect.. there are MANY places where the bot fails to do something and ends up running back and forth waiting for player reports
Yes, it's trying to sell something, duh.
LCP might be the bogeyman, it's certainly going to be for most bots.
And if we're talking about superstitions, I'll introduce you to a new one, efficiency.
The concept of LCP is to look at frequency over a spread of data, to create a sort of Bayesian spam filter of data, that would flag your account. It's not easy to counter or design either, since its not CTM alone, it's the navigation system. Essentially, LCP's job, is picking up efficient routes and flagging them as suspect. It could just be LCP tied to a "known" player, a bot user, which plays and is used as a stalking horse to feed data into the aggregate. But it doesn't have to be. Certainly, like the Bayesian system, flagging words as spam, trains the system faster, but it still has to find those aggregates using math, and not dictionaries.
If that's the case, profiles are useless unless they add distractions, delays and wandering/detours, etc. and the more "random" the paths created, it can pick up trends with navigation markers and blacklisted areas as a sort of fingerprint around an area. So a profile, is only as good as the navigation used.
If LCP is that efficient, which i seriously question, there may need to be a series of chaotic error values added to navigation nodes, to create broken paths for breaking this efficiency detection.
LCP works by aggregating statistics not for each node, but by movement in a sample area, the paths taken and detours used, etc. 90% of most player navigation is landmark reference, with plenty of erroneous deviation from the waypoint. Players don't even notice that its inaccurate, because they use the terrain as a visual referee as to positioning, not the straight linear point to point route created for efficiency. A bot uses a entry and exit point that usually walks over the same areas in the mesh, it wears a groove, if you will.
If so, the less obvious approach to LCP-busting is navigation by elliptical paths to destination nodes, to break the linearity of navigating from node to node, which is just adding in false mid-route data so it detours around a false point.
Another similar idea is path smoothing, where a flight or mount path adds additional routing points to create smoother routes at the cost of efficiency and possibly creating bad paths too near blacklisted nodes/walls/stuck points.
another approach is "bad" routing, using nav jaunts, heading towards a higher or lower point and correcting towards the last 20% to 10% of a path. Using elliptical paths to route near a destination, using expanding error between nodes to create a zig zag effect in longer path routes, etc. Basically looking like players, involves doing things players do. Like turn to look, routing to landmarks, slow turns, etc. things that would make the bot look senile or blind, when viewed by players and users.
To be ruthlessly fair, damaging the navigation system to avoid detection, may also be the thing that flags HB faster too. There is no universal solution, except a series of unique and random ones.