What's the best?

[CheckersStrongplayer]

so Chinook better than Nemesis?

[CheckersStrongplayer]

I never got an answer for this? Anybody's theory would be valid i guess

[matthewkooshad]

Whichever one that you cannot beat is the best.

[CheckersStrongplayer]

i don't know, which one i can because i have never played none of with their full databasese meaning the 8-piece

[Gene Lindsay]

Who knows which is best? If you had a ty with them and cake and kingsrow it would be a tossup as to which would win. You can download cake and kingsrow for free and they are just as good in my opinion.

Gene Lindsay

[CheckersStrongplayer]

Cake is stronger than KingsRow, it jsut depends on who has the better database to play, the six piece is pretty strong, but i remenber when i had Chinook 8-piece database, it whooped Cake like a rookie, Cake struggled so hard that it used to take it like 20s per move.

[Jason Solan]

i'm not sure where you're taking your information, but i wouldn't trust anything older than 1 or 2 years.

These programs advance so fast that the next version is much better than the last. Also a bigger database is an obvious advantage. Several 10-piece databases are available now i believe.

I would venture to say that none of the afformentioned programs has a distinct advantage over the others at this point.

[CheckersStrongplayer]

the 10-piece for what?

[Gene Lindsay]

Cake is stronger than KingsRow, it jsut depends on who has the better database to play, the six piece is pretty strong, but i remenber when i had Chinook 8-piece database, it whooped Cake like a rookie, Cake struggled so hard that it used to take it like 20s per move.

If Chinook's 8-piece database "whooped" cake like a rookie, what program were you using to play Chinook's database, it won't play by itself.


Gene Lindsay

[CheckersStrongplayer]

i was using CheckerBoard obviously with KingsRow...

[Ed Gilbert]

I have been reluctant to join this conversation, because I am obviously biased. But I can give you my opinions, and I also have some hard data that most of you do not.

For most of the past few years, I believe cake has been the best crossboard engine. With opening books turned off, and with comparable endgame dbs and RAM allotments, I don't think any other engine could touch it. Comparisons change when you turn on opening books, or add 10-piece dbs to engines that now have them, but for equally configured engines cake has been my pick as the best. KingsRow has been playing catchup for a long while, and now it has gotten closer, but as the following data show, cake still has an edge over KingsRow. Before I released the latest KingsRow on the web, I ran 3 engine matches against cake. In these matches cake and kingsrow each played both sides of the 144 'normal' ballots with opening books turned off, both using 8pc dbs and equal allotments of ram for hashtables and endgame db caches, with these results:

kingsrow 1.15u vs cakem1.09d at 5 sec/move:
23 wins, 19 losses

kingsrow 1.15u vs cakem1.09d at 10 sec/move:
9 wins, 19 losses

kingsrow 1.15u vs cakem1.09d at 15 sec/move:
11 wins, 12 losses

As you can see, engine match results are not always perfectly consistent. It takes a huge number of games to get results that are statistically significant between 2 engines that are not that far apart in playing strength.

When engines are clearly separated in engine strength, it doesn't take as many games to observe a clear pattern. Witness Suicide Cake's recent 4-0 victory over the Canadian challenger Roshi47.

Nemesis is another one that at least has the potential to be in a class with the strongest engines. I do not have Nemesis, and my only knowledge of it is what I oberved at the 2002 LV computer checkers tournament. At that time I felt that the 3 entrants -- Cake, Nemesis, and KingsRow were all fairly evenly matched, and that feeling was reinforced by the results of the tournament. Cake and KingsRow have improved markedly since then, while Nemesis has not been changed, so on that basis alone I am guessing that Nemesis might not now be as strong as Cake and KingsRow. But that is my highly subjective, and clearly biased opinion, so take that for what it's worth.

Another checkers program is WCC. This used to be a commercial product, but now is available from Bob Newell for basically his cost to duplicate the 13 CDROMs and ship them. The web site for this program is still active, and among other things it contains some rather inflamatory statements like it's "the program that eats other programs for breakfast". I was curious to see how accurate that is, so I set it up to play some 3-move games against KingsRow. I've been spending a lot of time watching college basketball during 'March Madness', and I've been making the moves in the checkers games as I watch the basketball games. The parameters of the match are: both programs running on a 2.4GHz P4, 10seconds/move, opening books off, 8-piece databases, 32mb hashtables and 600mb endgame db cache buffers. I've finished all the normal 9-13 and 10-13 ballots, and I am part way into the 10-14's -- a total of 106 games so far. The results:

KingsRow version 1.15u vs WCC Platinum:
25 wins, 2 losses

I don't think I'll play many more games, the relative strengths are pretty clear now. If anyone is interested, I have the pdn files of all these games, you're welcome to have them.

As far as Chinook goes, I don't think anyone except perhaps Schaeffer knows how it performs relative to the best programs that are publicly available now. AFAIK it has not played a public game for something like 8 years. I have no reason to think it would play any better than any other program given a level playing field, with all programs configured with 8-piece databases, opening books turned off, and all given equal allotments or RAM. But that does not really matter much. Jonathan Schaeffer made an huge contribution to the advancement of computer checkers when he invented the scheme for constructing an endgame database in such a way that it can be used during a lookahead search. He also invented some important search improvements, like history tables and enhanced transposition cutoffs, which make game programs more powerful. All of today's checkers programs would be much weaker without his contributions.

-- Ed

[CheckersStrongplayer]

Very interesting results you got there Ed, is there any chance that Chinook can be awaken to go against the best matches ever which would be Chinook Vs Nemesis, Chinook Vs KingsRow, and Chinook Vs Cake Manchester, this guy lives in Canada right? Is there any chance that you can e-mail him Ed, if i do i know that won't put much of matter in my challenge, but with you it's way different since you're quite known for building a really good engine, you can probably talk to that guy that made Cake Manchester to see what he says. Tell me what you think?

[Patrick Parker]

chinook isnt coming back atleast not until it solves everything and by then shouldnt lose?

[CheckersStrongplayer]

so it's major database which is i think 8-piece or 10-piece is not good enough to beat them since it hasn't been updated? And how come Nemesis hasn't been updated?

[Alex_Moiseyev]

To the best of my knoweldges - Chinook retired (so far) from competitive play only, but not from "scientific reserach".

Although, as far as I know, Prof. Shaeffer started building 11 pieces ending database and started another project - "solve 3 move deck" at low level - prove scientifically (full research up to the ending database) soundness of each opening in deck.

2 openings, Black Doctor and 9-13 21-17 5-9 already solved. I assume the next opening in line will be 9-13 21-17 6-9

Alex