Hi, exists somewhere in Internet space statistics about national wordlists? What I mean is : Language .. # of words OSPD .. x xxx xxx German .. x xxx xxx...
... In bag. ... It reduces variance. ... Yes. Moreover, without biasing the simulation's probability distribution. ... Greater accuracy. ... If you simulate...
sheppardco@...
Feb 1, 2002 10:59 am
315
Quick question: am I correct that PioBotJr simulations use the following cutoffs: Endgame: top 25 moves? Otherwise: top 10 moves? And the dictionary is...
sheppardco@...
Feb 1, 2002 11:03 am
316
... There biggest variable is whether a program has all the necessary features. I recall a thread in the chess programming newsgroups in which programmers ...
sheppardco@...
Feb 1, 2002 11:17 am
317
I agree that you're on to something here. Here's a suggested formula from a quasi-mathematician on how to implement it. First, setup... 1) Don't just track...
The PioBotJr dictionary is OSPD (at least it is supposed to be :) The Endgame, when the bag is empty, it uses the top 25 plays although I am going to increase...
John Babina
john@...
Feb 1, 2002 3:32 pm
319
A few days ago before I decided to write a Simming bot, I had an idea I want to run by you all... Instead of "simming" a play by drawing random 7 letter racks,...
John Babina
john@...
Feb 1, 2002 3:49 pm
320
... I believe this will work better than Monte Carlo. This is exactly what I wrote up in that paper I referred you to in personal mail a couple of days ago. ...
How does this give you what you want? If done properly this gives you the average score of all plays possible. The purpose of simming is to give you the...
If you use the ADJUSTED rack leave score using the SG formula, this would solve that problem -- I forgot to mention this in my example... -John ... From: "John...
John Babina
john@...
Feb 1, 2002 4:21 pm
323
I actually never got to read your paper Graham! :) So I'll have to give it "A READ" rather than "ANOTHER READ" .... (don't take it personally). To me it just...
John Babina
john@...
Feb 1, 2002 4:25 pm
324
... Agree 100%. Although with my own engine it did take me about 10 minutes to perform the exhaustive generation, and by the look of the code samples we were...
... Yeah: evey 100 or so iterations, you should see if any moves you're simulating are so poor that they can be eliminated from future iterations of this sim. ...
... Technically, I use the lookahead engine only in the pre-endgame, which Maven defines to begin when there are 16 unseen tiles (i.e., 9 in the bag). Key...
sheppardco@...
Feb 2, 2002 4:51 am
327
... avg ... Correction: 20 * [chances of oppt having HELLO]. * [Chance HELLO is best given that HELLO is legal] The problem is that the last probability is not...
sheppardco@...
Feb 2, 2002 4:55 am
328
... Not exactly--John's probability calculation is wrong because you cannot tell whether HELLO is best, only that it is legal. Your Global Analysis has a...
sheppardco@...
Feb 2, 2002 5:33 am
329
... I agree, too, and it is quite different from the approach I have always taken. It may be better. I consider inferences to be a move generation problem in...
sheppardco@...
Feb 2, 2002 5:54 am
330
Thanks for the comments; unfortunately it's late and I have <things to do> at home ;-) so this is a briefer response than your comments merit. re taking rack...
... Yes, and that's clever, but it does not address the problem. Consider Global Analysis of the opening rack. You order all moves from highest to lowest...
sheppardco@...
Feb 2, 2002 6:14 am
332
I find this thread on global analysis quite interesting, but a step back to the real objective might help reveal how global analysis could prove very useful. I...
Steven Gordon
sagordon2@...
Feb 2, 2002 6:46 pm
333
Ok, I have this idea, and it seems pretty simple (but needs some fine tuning)... I want to bounce it off of everyone... * Your simulation Scrabble program is...
John Babina
john@...
Feb 3, 2002 7:08 pm
334
Ok, I found a flaw in my logic but it made me see something else -- I decided to make up a test case where the best normal play was to play a high scoring (34...
John Babina
john@...
Feb 3, 2002 7:52 pm
335
hi guys
extreme newbie here.
am 3rd yr diploma student in software engineering. a major part of our 3rd
year is a project.
i have chosen to *attempt* to...
... Not quite. What we call a 'movegen' is the search for the playable words, given the tiles. A simple movegen may return all the plays, the highest scoring...
... Yes, the plan had an arithmetic error, but that isn't the real flaw. The fundamental flaw is that you want to compute winning percentage, and you are...
sheppardco@...
Feb 4, 2002 4:00 am
338
I'm wondering if a possibly better strategy would be if the computer assigned a higher penalty weight to the opponent's possible response plays when the ...
AmazingMazeMan@...
Feb 4, 2002 4:38 am
339
... Brian, is this a serious suggestion? I can't tell from the tone. I have often thought (okay, daydreamt) that there may be something to gain from using a...
... I have a different view towards neural nets and Scrabble. The world of backgammon is sooo small, in the sense that the number of possible boards is much...
Laeuter
laeuter@...
Feb 4, 2002 10:13 am
341
... I've been waiting for Mark Watkins to chime in, as I know he has already done this in BobBOT. He uses it to decide if a position requires deeper analysis....
I have now modified my program from simming the top 10 plays 2500 times to simming the top 50 plays in 100 sim intervals and am monitoring the SD as per JAC's...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
