Terrorbot - AI for TTT

@markus The bots aren't meant to be psychic. That hook (which only affects a small subset of traitor buttons) would still require a lot of additional checks and some randomisation to be close to what a human player would notice. Might be included in a future update, but it's very situational so not worth the time at present.

I know standard bots don't have normal SteamIDs (technically they do, it's just a generic not really valid id that all bots use). That's why it's noteworthy that these do, they'll be compatible with other scripts. The IDs are a little outside the standard range ( https://i.imgur.com/4sRohvu.png ) so they'll never conflict with other players, the downside being that they might not pass some steamid regex filters depending on the script.

On a vaguely related note, it was interesting to find ULX wouldn't ban the bots, instead stating "Cannot ban a bot". I always assumed they'd only check SteamID, not ply:IsBot().

@my_hat_stinks You can do some things (check which player used which traitor button) with the TTTTraitorButtonActivated-Hook. (See: https://github.com/garrynewman/garrysmod/blob/master/garrysmod/gamemodes/terrortown/entities/entities/ttt_traitor_button/init.lua#L93) What do you mean with “unique SteamIDs“? Normally bots have not really a SteamID. On Server:

1. 32 Bit: “BOT“
2. 64 Bit: something like “90071996842377216“

On Client:

1. 32 Bit: “NULL“
2. 64 Bit: no value

So how do you trying to realize this? :O

Very nice

@Chadnes You're right, evaluation who used which trap is incredibly difficult even for human players, so much so that some servers outright ban killing players based on that. That is only one possible piece of evidence, though. The bots currently use more reliable measures, such as shots, bodies, and callouts. The trick then is just to evaluate the reliability of the evidence, such as checking whether the callout by someone already thought to be a traitor.

Ideally we'd have traitor rooms marked out on the navmesh, but for some reason you can only mark areas with predefined names. Blame the engine for that one.

This seems like a tricky subject because how can a bot figure out whether someone is suspicious or not? Considering suspicion is decided on someone's behaviour around them. The only way for a bot to figure out whether someone is a terrorist is if they knew where the terrorist rooms / traps are and whether they have been used. It also is determined by who has used them as people who are close the the trap is inaccurate as anyone can be near a trap, esspecially a terrorist to trap a civilian. How have you done this and is it particularly accurate ?

I suggest you make a video with an actual player base to see how the bots perform against large amounts of people, just to ensure it does not generate havoc.

Damn, unique and intriguing as said. GLWS

Very intriguing, I'd like to see how this works when used with real players, looks good hat. GLWS

This shit seems interesting