Hello,
I've been developing a bot for a game, which helped me to earn some money, but as of recent events, I've noticed, that other players using bots(not the same kind of mine), have been started to be banned massively.
Becouse of that, I figured that I should update my antiban system. As of now, I'm not using anything fancy, just some old randoms, though, it does appear not so bot-like.
Any ideas how to get this system better are appreciated. I know that I didn't give you any context, but I figured that there may might be some ideas, that I haven't thought about.
Thanks in advance.
wat
![Counter+strike_821b88_5238752[1].webm](https://i.imgur.com/fZlos6K.jpg "Counter+strike_821b88_5238752[1].webm")
I don't know if you played the latest mods of counterstrike source, but the bots they modified act nothing like bots. They taunt, they coordinate with other bots, they anticipate your moves and cunningly trick you. They are modified to imitate the actual behaviours of pro players.
I'd say research into that, see how they made it so.
>>7663108
Thanks for feedback. I think it's not so bad idea to record movements and behavior of a player depending of state in game(doing task, not doing task) and then analyzing and then somehow applying those settings. I will look deeper into that.
Not really /sci/-related (more /g/) but I have some experience making bots so I'll put in my 2 cents.
If you develop your own bot for personal use then you're in a good position off the bat because usually game devs ban bots by finding ways to uniquely identify the specific bot. That doesn't mean you're safe because botting often becomes an arms race and eventually devs look for ways to be one step ahead of botters.
First they'll look for abnormal inputs that can't be generated naturally (especially if you're playing directly with the API), so make sure the way you send input to the program looks like something a natural keyboard and mouse (or at least a macro) would look like.
Next they'll try and grab lists of programs running and accessing the game process and identify known bot programs or abnormal access of injection of the game's memory space. But that stuffs usually identified by standard anti-cheat software anyway.
If they can't find a way to uniquely identify bots, they'll probably look to activity (identifying players that play nonstop, 24/7). Once they identify one they'll generally try to contact the player with an administrator to see if he responds (if not, he's probably botting). So knowing how the administrator would go about contacting you and having your bot identify it and notify you when it happens is a technique that more advanced bots use in more widely-botted games (such as WoW).
Devs don't generally make heuristics for identifying behavior itself as "robotic" or "human" (although avoiding simple mistakes like sending signals at set intervals and reacting to things with 5 ms intervals should obviously be avoided) because it's too much work compared to the alternatives. But if it's a competitive game then you're up against humans who will constantly be doing impromptu Turing tests and will report to devs if they fail. So in that case you have to put a lot more effort into making it appear human.
>>7663129
I think a good idea is to always allow some degree of randomness for actions. For example, if a bot consistently performs an action every 0.3 seconds, then it's beyond human ability. But if you use a random number generator designed to produce a human like distribution, then it could work. Just record a bunch of normal game play that you perform and analyse it whilst looking for parameters which can be random. Like click location, click timing, typing, etc. Reaction times to an event shouldn't be perfect
>>7663084
>How to appear human?
Enter a captcha
>>7663108
Kek, this webm
Can you inform the CSS version that you speak of?
>>7663147
the latest CSS on steam. Although you can find those bots in specific servers. I encountered them on 7/24 deathmatch dust and office servers.
it's runescape isn't it
>>7663140
This. That's why programs like SCAR were so successful in Runescape. It used a combination of color and area specification which worked well to emulate the randomness of an actual player.
It's a browser game. The bot is runned by probability, which doesn't deviate from average probability, which is set in the game(sometimes it does, in extreme cases, trying to simulate "random" event).
The point is, if the bot runs for a long time, the frequency of those "random" events, should converge to the probability which is average of all the events, right? Should I make the those probable events occur progressively less?
