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 > > > > > > >
