Identifying Bots in ARAM using heuristics. ( Building a heuristic rule-set)
ARAM nowadays has been invaded by bots which are destroying that mode completely. Unfortunately the ability of the bot catching program for LOL seems to be lagging behind in recognizing certain behavior as bots.
I am an AI researcher and I want put down certain behaviors in these bots which can over a period of time be used to form a heuristic ruleset for banning such accounts . Of-course more bots will be built and more rule-sets would need to be created .
- Variations in spots and pathing . If you draw the heat map of the time spent by a champ at a particular spot of the map , Human players will have larger distribution than bots . Bots tend to occupy very specific locations on the ARAM map and take specific paths.
The bot creator can inject certain randomness in the spot occupied but this will cause higher aberrant behavior since spot occupation is not based on reasoning for bots.
-
Variation in Item builds : Typically for a human who plays ARAM maps , item variations would be higher than that of bots for their item build . We humans have a concept of boredom so we mix things up , bots dont.
-
Heals : I have rarely seen bots go for heals . The absence of any heals taken would again mean that this is a bot .
-
Randomness : All the above behaviors can be corrected in the bot code but at the expense of randomness. Our behaviours in-game are caused by two types of causation . Short term causation : Our need to get immediate goals , like flash kill low hp champs , tank hits only when tower hp is low . Long term causation : Overall desire to win the game and push the map when opponents are all dead .
The fourth point can be used to catch out any randomness that is introduced to counter the first three points. I will keep adding to this post as and when i get new ideas.
