Level cap

[skltnk00]

Botfobia...

People are always affraid of what they do not know most.

[razfen]

Botfobia...

People are always affraid of what they do not know most.

wtf are you talking about

[skltnk00]

Seems like people are affraid of combat bots taking over the game world.

[skltnk00]

Botfobia...

People are always affraid of what they do not know most.

wtf are you talking about

Isn't it obvious?

[iamahh]

for real tho, the endgame is community markets, it's the spinal cord of the whole world because creates a positive interaction between players, unlike pvp...

CF is pretty much an organic capital at this point, if a new player setup in that realm and manage to survive and get currency he can get a lot of good stuff

besides some people actually like trading just for the thrill of crafting, selling, buying... if someday there's a "terroir" system that distributes qualities around the world terrain, so that each region produces special buffs, based on ageing, then economy game will become crazy good

[Potjeh]

Why wouldn't they? You don't even need complex pathfinding for the most part, the obstacle space is fairly simple. If anything, a bot should outperform a player's reflexes.

make any1 do that and i will pay him becouse its just bullshit

I'm not saying it's not a lot of work. I wouldn't do it for under 5k $. There are no real theoretical obstacles. I mean, FFS, they have a DotA bot that beats world's best players 1v1, do you really think Haven's combat is more complex than DotA? Just look at this baby: https://www.youtube.com/watch?v=x7eUx_Ob4os

[azrid]

So you'll create a self learning AI for 5k?

[Potjeh]

That's actually trivial. There's tons of papers out there, it doesn't take very much to write your own implementation (I wrote an implementation of NEAT over one weekend). Or you can grab a free one at https://openai.com/systems/. Creating a training environment wouldn't be all that hard either because haven's combat is fairly simple. Then it's only a matter of running it for a long ass time. The hard part would be properly integrating it into a client, and I won't bother to learn Java unless I'm paid handsomely.

[svino]

I'm not saying it's not a lot of work. I wouldn't do it for under 5k $. There are no real theoretical obstacles. I mean, FFS, they have a DotA bot that beats world's best players 1v1, do you really think Haven's combat is more complex than DotA? Just look at this baby: https://www.youtube.com/watch?v=x7eUx_Ob4os

That's actually trivial. There's tons of papers out there, it doesn't take very much to write your own implementation (I wrote an implementation of NEAT over one weekend). Or you can grab a free one at https://openai.com/systems/.

That bot is missing a lot to play dota, and honestly i'd be surprised if they managed to get 5v5 working before next year. 1v1 SF is just about being mechanically good at the game, you need timing to get the last hits, and you need machine reflexes to raze and raze-cancel if you want to do it perfectly as the bot does. Sure though, it's impressive that it has good positioning and can creep block.

Making a swarm of haven combat bots working together probably wouldn't be impossible but it would be much harder than you seem to think it is. Those guys who built the dota bot are 8 deep learning experts working for 6 months on it, to ONLY play 1v1.

If you compare something like group combat in haven with sf 1v1, the haven combat is still more complicated. I'll give you that 5v5 dota combat has more parameters than 5v5 haven combat though.

[Potjeh]

Nah, even 1v1 DotA is orders of magnitude more complex than Haven. The actions space is a lot smaller in H&H, so training would take a lot less time and you wouldn't need as much expertise to train it.