« Last post by Memories on 14 August 2017, 09:48:51 PM »
got gold sub now
I see. I think this is the first time this ruling has been made public then (unless I missed it somewhere recently).It may not have been public; Stef has privately asked me some rules questions.
I don't see that this was a question according to published rules: You should be able to choose the order (since they happen at the same time - after this turn). So this ruling creates a special rule for extra turns - that they are always last. In effect it actually means there are now two distinct times after this turn: First you can order Donate and Mountain pass, then you can order extra turns.
Cross-posting an old post of mine:I don't think Lord Rattington "knows" that Smithy-BM is the move, to any degree, ever. It doesn't know anything. It just tends to play mostly money, whether it's good or not.QuoteI think better AI should be a top priority for Shuffle iT right now, probably second after eliminating all the log-only UI stuff. iOS and Android versions are important, but not having them is just putting off potential future players (who may already be totally put off by the subscription anyway). The bad AI is damaging player experiences right now. Here's a recent example of a new player who played a game where he lost to Lord Rattington, where Lord Rattington played Smithy-BM. The game AI should never ever play Smithy-BM, because it amounts to telling the player the game is broken. The game itself knew that the best strategy was this boring one! There should be a "bad" AI that messes around buying a bunch of stuff, like a newbie would. And then there should be some "good" AIs that play more complex strategies. No AI should be intentionally playing a boring strategy.
I just cross my fingers that Stef understands that a good game AI isn't one that wins the most. It's one that's challenging but not too challenging, and overall fun to play against.
I think the move is to make a bot that's as good as possible, and then use it as a base for other bots, by just handicapping them or lying to them (e.g. the attacky bot spends the first 6 turns overestimating attacks). I wouldn't start out trying to design a fun bot.
See this (https://en.wikipedia.org/wiki/Pareto_principle) for the origin of the 80/20 "rule", and of course it is a generalization that does not always work with those number.I see, interesting. Do you know this one? I measure something - let's say how tall a mountain is. I use some units I pick - whatever units. I report the number. What's the first digit? 30% of the time, it's 1 (and it's 9 only 5% of the time).
My main point is: It is much easier (though still a very complicated task) to program a complex rule system (as Dominion is), when you already know all the parts, then to start with a subset and then get 'surprised' by new effects when a new expansion comes around.That's totally true, but I have evidence there too.