Yea that was a critical turn 😢.

Fascinating fascinating

I liked 5 2 for black more, they don’t need the extra wheat and would be able to get to 6 2 relatively quickly with the starting sheep, high wheat production, and coordinated 5s

4 Player Catan is too social to create a "solver". If you want to play at the highest level you need to table talk and make deals like non blocks, non steals, future trades and working together with the table. Which would be next to impossible to implement into an AI model. It's also not perfect information cause of players stealing cards / buying dev cards. It's simply not realistic with current technology.

@@kidkid but you can definetely create an engine that computes the best possible decision based on probabilities. I think there are tools to be created that can push the game forward, even if ignoring the social component

​@@MHNK77 I disagree. I don't think you can make an engine to play the best move consistently without having it understand how the game is played. First of all there's the problem of the best move assuming your opponents are playing optimally vs the best actual move. If you were playing against AI's who didn't have any emotion and was just trying to make the best move it could no matter what then that's one thing. But that's not how most humans play the game. Often if you don't know the people you're playing with you don't want to plow someone unless it's a significantly better move than the alternatives since they can easily screw you over later in the game with the robber if they feel like it. The best play will depend on your ability to table talk and communicate with other players. Buying a city might be better if you know the opponents wont overestimate you and block you to death. While if you think your opponents might get scared of you and start blocking you if you drop a city buying 2 dev cards first might be better since you will be able to protect yourself. Then there's the problem of how to actually train an engine like that. You can't use reinforcement learning in this case cause computers playing against computers will be different from humans playing against humans. Sure you could try to gather a lot of data from human games and see what move has worked in those games but the data would be pretty hard to feed into the model. If a good player playing against bad players picks mediocre settlement spots then goes on to win the game, that probably had more to do with their skill outside of placements rather than them placing settlements correctly; If you train a model on flawed data that it won't turn out that great. And that's just some of the problems with trying to create something like this, there is also things like: How much do you value wins compared to points? If you're playing in a tournament and you're guaranteed making it to semi finals with 7+ points you would want to play less risky compared to only going for wins. Should you prioritize win% or expected VPs? And predicting what settlement spots will be open for you on the way back without knowing how your opponents place.

@@kidkid 2 words: stochastic simulations :) Simply simulate a bunch of *random* scenarios and see which placement/ decision leads to the better average outcome. Again, it wouldn't be the best move 100% of the times, like a chess engine, simply something that on average gives the best result. You're overcomplicating things to sound smart

@@MHNK77 But how would you see which decision leads to the better average outcome when your engine doesn't know how to play the game? You see what I'm trying to say here? To simulate something like that you would need the engine to be able to evaluate its own position and understand what is good and what is bad. And I'm not trying to overcomplicate anything. It's just the fact that it is that complicated to begin with.

DyLighted had a brief foray into Poker before coming back to focus on Catan.

@@froseph85 he's definitely got the logical decduction for it. Online you need to be super studied but live low stake games are pretty soft. I watch a ton of poker vlogs and would definitely watch his content if he ever decided to dabble in poker content as well.

For road chance yes, but it would have made him a road target instantly

The next several videos are already recorded with the bad mic quality, but today I went and fixed it. In the future the audio will be a lot better!

LOL and then the asshole accuses _you_ of crying all game. What a little punk.

By plowing him at the 9 wheat

What does OJ mean?

Oh yes, my comment was stupid@@jondonjr

Orange @@seshuis96

@@seshuis96 Orange