Hi

I looked at this gcg and I see the problem.  In the gcg its clear that
opponent 1's racks are partial racks consisting only of the tiles
played.  The short summary of this email is that you can't expect
reasonable evaluation of a partial rack.

On the turn in question the gcg gives:

>Player1: AKN h11 NAKED

However what is really happening here is that the bag is empty and so
player 1 must hold AIKN (player 2 holds ORU).  By setting the rack to
only AKN the game is set up so that each player holds 3 letters and
there is 1 tile in the bag - clearly an impossible situation in real play.

When the championship player is used it tries to use a pre-endgame
solver since there is a tile in the bag.  However the pre-endgame
solver assumes that you have a full rack (since the bag isn't empty)
which in this scenario isn't true for player 1.  It therefore fails to
give a meaningful result.  Simulation and static evaluation are also
wrong since they are based on the faulty game position too.

I'm not sure how to give a nice solution here.  Putting AIKN on the
rack gives the correct results.  The question is how to get such a
nice result in general (the problem exists even before the endgame too).

Matt

--- In [email protected], "G. Vincent Castellano" <g...@...> wrote:
>
> Try mine (attached), it seems to show a similar issue.  On this
turn, Generate 
> Choices and Simulation show the correct win % (0), but Ask Champ
shows the 
> following.
> 
> JohLue: Turn 14
>     A B C D E F G H I J K L M N O   -> JohLue                   AKN
      413
>     ------------------------------     gvc                      ORU
      448
>   1|=     '       = F A Q I R   F| --Five Minute Championship
Player's choices 
> (your play: 3)
>   2|  -       "   R E D     A B A| best  J8  KA(I)N   12 100.0%
>   3|    -       ' O H O     J O  | best  J8  AK(I)N   12 100.0%
>   4|'     -       A       -   O X|  1.00*H11 NAK(ED)  11 100.0%
>   5|        -   B R U N T        |  4.00 K5  (T)ANK   8  100.0%
>   6|  "       "   I   "       " Z|  8.00 14B A(RE)    10 100.0% KN
>   7|    '     S O N G       ' T I|  8.00 J9  K(I)N    11 100.0% A
>   8|=     C L I N G       ' M A G|  10.0 10B (O)K(EH) 11 100.0% AN
>   9|  V ' R     ' S '     E E L S|  10.0 5M  AN       8  100.0% K
> 10|  O   E H "     L I P   W E  |  11.0 J9  K(I)     6  100.0% AN
> 11|  E   M A           O D   N  |  11.0 J9  A(I)N    7  100.0% K
> 12|' S L A W     '     P I   T '|
--Tracking-----------------------------------
> 13|    U T     '   '     T Y E  | IORU  4
> 14|  - R E   "   E V I C T E d  |
> 15|U N E D I t E D       Y A   =|
>     ------------------------------
> 
> --gvc
> 
> 
> John O'Laughlin wrote:
> > Chris,
> > 
> > Do you have a gcg for this game?
> > 
> > John
> > 
> > On Tue, Jan 6, 2009 at 7:16 PM, tendenc <tend...@...> wrote:
> >> Quackle AAGINR 451 pts
> >> Me NITROSO 368 pts
> >>
> >> New Player 1: Turn 12
> >> A B C D E F G H I J K L M N O
> >> ------------------------------
> >> 1|B E E C h N U T P =|
> >> 2| - " R E " U - |
> >> 3| - G O X O N - |
> >> 4|' - I D U T '|
> >> 5| N O D R |
> >> 6| " I F " I " |
> >> 7| Q A T ' ' P E ' |
> >> 8|= f I B R I L S L ' =|
> >> 9| O ' A ' Y E ' |
> >> 10| V T H E M E D " |
> >> 11| E E - D |
> >> 12|Z A G - S ' - W '|
> >> 13|I L ' ' O Y H A J|
> >> 14|T O U C A N W A V E R |
> >> 15|S M ' A L K I E ' F E|
> >> ------------------------------
> >>
> >> The winning play is NITROSO at O5 for 74 points. This wins the game.
> >>
> >> When I simulate the generated choices, NITROSO at M5 (62 points)
> >> gives a 100% win, when in actuality loses the game immediately.
> >>
> >> I don't pretend to understand all the intricacies of simulation but
> >> it would seem that a math function to calculate the score after an
> >> outplay to validate the win% would seem appropriate.
> >>
> >> Cheers!
> >>
> >> Chris ten Den
> >>
> >>
> > 
> > ------------------------------------
> > 
> > Yahoo! Groups Links
> > 
> > 
> >
>


Reply via email to