Good news about endgame handling in the newly-released v0.94 of 
Quackle.

Back in June, I had a game at club where I needed to play very 
carefully to prevent my opponent from bingo-ing out.  I was ahead by 
46 points with 3 tiles in the bag and both blanks unseen.  The 
remaining tile pool was 'ABDEGILR??' so I figured she had at least 
one blank on her rack, and I wanted to be careful to both (a) block 
the only open bingo lane and (b) not empty the bag.  I found a 2-tile 
play that was just the ticket, made that play for 17 points, and went 
on to win the game.

When I ran the game thru a previous version of Quackle (v0.92, I 
think) in post-mortem, the move I had found wasn't even listed as a 
choice.  What it did list as choices were a 36-point play that I had 
seen and decided against because it emptied the bag and opened an 
even worse bingo lane, and a 32-point play that left the existing 
bingo lane open and opened up another bingo lane.  I considered both 
those moves to be likely to lose, added my move to the list, and ran 
a Sim.  The moves of interest were SPOOK 36 and WOP 32, both very 
dangerous because they empty the bag, and (T)OW off a T, fewer 
points, but closing the only open bingo lane.  The results of that 
Sim were something like this:

SPOOK ~32, (T)OW ~31, and WOP ~30 -- not what I considered a 
satisfying result.

I'm happy to say that v0.94 of Quackle handles this endgame better!  
For one thing, (T)OW is now listed as a choice by "Generate Moves," 
and when I run a Sim of ~500 iterations on a list of 6 choices, I get 
this result:

place/word   score   win %age   equity
L4   (T)OW    17      75          1.3
10J  SPOOK    36      50         -3.9
14B  WO       23      39         -34.7
14B  WOP      33      27         -19.2
14A  SWOP     35      27         -20.5
14A  TWO      25       6         -28.2

I'd still like to see the gap between (T)OW and SPOOK wider, and I 
think WO (not emptying the bag) is better than the 4 bag-emptying 
plays, but having (T)OW sim best is still a *huge* step in the right 
direction.  Combined with playing almost-bingo opening racks 
like 'AEINORT' significantly better than before (exch O rather than 
playing RATION for 12 points), this version of Quackle is MUCH better 
than the previous version.

Nice work, Quackle team!

John Hart




Reply via email to