Note: I said 6144 games in the video, but that's inaccurate. It's the perspectives of 6144 players' games, but it's 4P data, so the actual amount of games is 1536.
catancommunity.org/cms/Tournaments this is a good one for tournaments in the weekends. There is also a Discord community called 'Catan Championship' that's really good.
I wonder if you don't convolute your analysis when you define strategies by successfully executing elements of that strategy. For example a player picks a strategy that involves the sheep port, but they're blocked by another player and never get to use the port. That is an unsuccessful run of that strategy, but it wouldn't be included in your statistics. That might make port strategies seem more successful than they really are.
Yes, that's one of the flaws of this research. It's tough to make it perfect. If you have ideas on how to improve this part, I'm happy to try it in the future.
@@Treeckosaurus for simpler games you can just have automated players take sort of random actions each turn and then draw lessons from a much larger dataset. For games like chess you can generate a space of possible moves and search for promising paths. But Catan has that social element which makes me think it'd work different. One thing you could do is maybe try to quantivy the importance of luck. If you know the total pips each player had each turn, then compare that to how much they got, that might show being x percent above average translates into y percent greater chance of winning? My hypothesis is that a good player can snowball beginning luck into a win very effectively. But that later in the game winners have more villages and cities so they're not as exposed to the dice. Another is maybe there's something you can do with the social pressure. If you're in the lead, other players should act more against your interest. Maybe things like blocked roads or village locations become more frequent for players in the lead? Maybe you can think of a proxy for usefulness of trades and see if players in the lead tend to have fewer useful trades? Something like that? Might be interesting, because players behind will want to slow the lead player down, but not at the cost of their own prospects? I'm really not a high level player though, so I can't say how high-level players would play.
There's theory that in high-level Catan games you don't want to appear too strong at the start, because of blocks/steals/etc. Perhaps Green took that too far. Or perhaps Green made a deal with Red - that if Green left Red the 6/4/3, then Green would get a brick trade. I agree with you it doesn't look good though.
![Treeckosaurus [Catan]](http://yt3.ggpht.com/ytc/AIdro_nRVKXeGFBefgIvY06ORFQmO0c1cs3nZ5tcP_kSXaGz8Iw=s48-c-k-c0x00ffffff-no-rj)
Note: I said 6144 games in the video, but that's inaccurate. It's the perspectives of 6144 players' games, but it's 4P data, so the actual amount of games is 1536.
Hi! Where I can play tournaments?
catancommunity.org/cms/Tournaments this is a good one for tournaments in the weekends. There is also a Discord community called 'Catan Championship' that's really good.
Wow amazing video! Appreciate the shoutout and excited to grow this dataset further!
Thanks!
big fan of city road setups! Happy to see it as a top performing strategy. Thx for putting together this data for us!
Me too. Glad it got some glory.
I found this very interesting and useful. Hope to see you in div1 champs :)
Glad you enjoyed it! I will play a few games in around 2 weeks for Mind Sports Olympiad practice :)
Great video Treeck! Nice to see one of my fun setups get a feature 😂
Thanks!
I wonder if you don't convolute your analysis when you define strategies by successfully executing elements of that strategy.
For example a player picks a strategy that involves the sheep port, but they're blocked by another player and never get to use the port. That is an unsuccessful run of that strategy, but it wouldn't be included in your statistics. That might make port strategies seem more successful than they really are.
Yes, that's one of the flaws of this research. It's tough to make it perfect. If you have ideas on how to improve this part, I'm happy to try it in the future.
@@Treeckosaurus for simpler games you can just have automated players take sort of random actions each turn and then draw lessons from a much larger dataset. For games like chess you can generate a space of possible moves and search for promising paths. But Catan has that social element which makes me think it'd work different.
One thing you could do is maybe try to quantivy the importance of luck. If you know the total pips each player had each turn, then compare that to how much they got, that might show being x percent above average translates into y percent greater chance of winning? My hypothesis is that a good player can snowball beginning luck into a win very effectively. But that later in the game winners have more villages and cities so they're not as exposed to the dice.
Another is maybe there's something you can do with the social pressure. If you're in the lead, other players should act more against your interest. Maybe things like blocked roads or village locations become more frequent for players in the lead?
Maybe you can think of a proxy for usefulness of trades and see if players in the lead tend to have fewer useful trades? Something like that? Might be interesting, because players behind will want to slow the lead player down, but not at the cost of their own prospects?
I'm really not a high level player though, so I can't say how high-level players would play.
As the master of Ore/Wheat/Wood, I would like people to refer to it as the WOW strategy.
I'm here to spread the word
What kind of bot setup did Green take at 2:04, terrible!!
There's theory that in high-level Catan games you don't want to appear too strong at the start, because of blocks/steals/etc. Perhaps Green took that too far. Or perhaps Green made a deal with Red - that if Green left Red the 6/4/3, then Green would get a brick trade. I agree with you it doesn't look good though.