|
Post by Lars Ericson on Oct 7, 2015 6:09:05 GMT -8
In Monopoly, players buy properties and each property has a value. The player whose property value sum is greatest wins. So the "goal value" is the sum of property values. Suppose there are 10 properties. Then there are 2^10=1024 possible goal values for each player (actually more than GDL can express on a scale of 0 to 100).
Does GDL have any examples of games such as this where the winner is the one with the highest score (not just 0 or 100), and the score is the sum of individual captured piece values? Is there an efficient way to express this, or do you need 1024 clauses?
|
|
|
Post by Andrew Rose on Oct 7, 2015 13:18:49 GMT -8
I'm not exactly sure what your question is...
Reversi counts the number of discs of your colour ("the score") and then declares the winner as the person with the most. Same with some of the chess variants. (Skirmish?)
And in terms of games with lots of different combinations to consider for the scoring, the 6-player version of Guess 2/3rds of the Average has 10^12 different possibilities to consider.
Does that answer your question? If not, can you clarify?
|
|
|
Post by Lars Ericson on Oct 7, 2015 20:00:57 GMT -8
Andrew, thanks for the help! So Reversi is implementing successor up to 64 and addition up to 8 and doing arithmetic via succesor relation. Guess 2/3rds for 4 players has successor up to 400 and axiomatic definitions of |x-y|=z,x+y=z,2x=y,3x=y, and x<y.
So if I want to say Mike has 3 houses H1,H2,H3 with prices 3, 7 and 11 and Sam has 2 houses H4, H5 with prices 17 and 5, then I need to express Mike's ownership of each house and the integer value of each house and then a sum of the house values of houses owned by Mike in using definitions written in terms of succesor relation.
|
|
|
Post by Andrew Rose on Oct 7, 2015 23:25:10 GMT -8
That would certainly be one way of doing it. Another alternative is to look at different versions of the rules. In mine, for example, the game ends when all other players are bankrupt. No need to add up property values.
|
|
|
Post by Lars Ericson on Oct 8, 2015 3:14:29 GMT -8
Hi Andrew, do you mean your hypothetical version of Monopoly? I don't see a Monopoly in base, Dresden or Stanford repositories in games.ggp.org. I say hypothetical because the other thing missing in GDL is a pair of dice. I'm right about that, right? We don't have dice. I found just one paper from 2001 for non-deterministic choice in Answer Set Programming such as this one. However this doesn't seem to be implemented anywhere and I don't recall it in GGP class.
|
|
|
Post by Andrew Rose on Oct 8, 2015 6:57:06 GMT -8
Yes, hypothetical. We don't have dice in GDL but we do have a proposal with widespread agreement for how we would introduce it without breaking back-compatibility with existing players. I've got it on my (long) list of things to do at some point.
|
|
|
Post by Lars Ericson on Oct 9, 2015 7:05:27 GMT -8
|
|
|
Post by Andrew Rose on Oct 9, 2015 9:24:59 GMT -8
Sorry, I don't know enough about ASP to comment. But I can tell you about the "random" proposal.
In GDL, for games with randomness, there'll be an additional role which will always have the name "random". Its goal value will always be 0 in all states. Servers promise to fill it with a player which chooses uniformly at random between all legal moves (or, more likely, to simulate such a player themselves). That way, dice rolls, shuffling, etc. are provided by giving "random" a turn. The rest of the time it no-ops.
Existing players will be able to play exactly as before. Players which want to take advantage of understanding that "random" will always choose randomly (and not, for example, deliberately attempt to stitch up another player) will be able to do so because "random" will be considered a key word by the server and therefore won't be obfuscated.
|
|