Re: [quackle] Question about Quackle Features

Sun, 18 Dec 2005 19:02:53 -0800

John,
 
I'm glad you found my suggestions useful.  I'll let you know if I think of any others.  Here are two other questions while I'm thinking of it.
 
1) John Hart posted on CGP today about knowledge of an opponent's rack and whether you can sim that.  I responded by asking whether you could tell Quackle what tiles you know the opponent definitely does NOT have?  Like if an opponent trades 1 late in the game, can you tell Quackle that the opponent does not have the Q?
 
2) Are you guys going to use the equity values from Maven, or are you going to generate your own?  More importantly, are you guys going to develop equity values for combinations?  I have the oldest version of Maven there is, I think, and the only combination it values is QU at 0 instead of Q -13 and U -4.  I imagine the K is not as negative with an A or C, and that may not show up in a one ply sim.
 
I guess the things I mentioned are harder, but might get done eventually.  Are these easier, intermediate steps:
 
1.  It gives you the equity value for the leave, so that you can figure out how much of your equity is due to score vs. board vs. leave.
2.  Can it tell you what percent of the time a certain tile gets played on the next turn if you don't play it now?
 
Let me know what you think of these. 
 
Thanks.
 
Mike Eldeiry

John O'Laughlin <[EMAIL PROTECTED]> wrote:
On 12/17/05, meldeiry <[EMAIL PROTECTED]> wrote:
>  Jason and John,
>
>  I have a few questions about possible features.  I don't know
>  whether they're possible, or even useful, since I don't know as much
>  about the game as you guys, and I don't know anything about
>  programming.  But here they are:
>
>  1)  Can it be programmed so that you just enter the whole game and
>  then it spits out an analysis for you, play by play?  It occurs to
>  me that this would be a huge improvement over Maven, which only lets
>  you sim one position at a time.  What a pain.

This is one of the features I'd most like myself, and I'm sure it will get done.

>  2)  Is there anyway the simmer can breakdown differences in equity?
>  How much of the 5 point improvement of this play over that one is
>  because it doesn't put a vowel next to a TWL?  How much is because
>  of a superior leave?  Etc.
>
>  3)  Finally, I think it would be really neat if the simming engine
>  could tell you when a certain play by you our your opponent follows
>  your candidate play a large percentage of the time.  This might be
>  particularly useful in the endgame.  For example, if I drop 1 tile
>  to fish, what percent of the time do I bingo?  I guess this would
>  require an inference engine to be accurate, since the opponent is
>  likely to respond by closing lines.  How often does a certain setup
>  work?  Is play X a mistake because 85% of the time opponent can
>  respond with QAT for 60?  (That would have to be a pretty dumb play!)

2 and 3 are harder to answer.  I often ask such questions myself, and
using Maven, I'm able to get the most of the answers by recording a
sim and putting the simulation log through a one-line script at a
command line.  At the very least, Quackle will be able to output a
simulation log like Maven's.  I realize that this is not a solution
for non-programmers.  We might be able to create some simple internal
scripting language or a GUI in Quackle to answer these questions, but
that would be a lot of work just to approximate what I can already do
with tools like grep and awk.

Most of the feature requests I've gotten from people so far are things
Jason and I have already thought of that will be easy to implement.
Yours would be a challenge, but that's the kind of thing we're looking
for.

John

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

SPONSORED LINKS
Cranium board game Child board game Chess board game
Milton bradley board games Scrabble board game Parker brother board game

YAHOO! GROUPS LINKS



Reply via email to