Jonathan Schaeffer Interview

[Jay H]

If you have an hour and a half to burn, check it out....
http://www.youtube.com/watch?v=aku4Zi_Mzq0&feature=related
Love him or hate him, the man has a unique place in the game's history......

"The former world champion....who is extremely egotistical...."

The "pot calling the kettle black".....

Jay H

[Jason Solan]

Thanks for sharing Jay, very interesting. That was before I discovered competitive checkers. I had always looked up to Dr. Schaeffer. The odd thing is that as far as heuristics go, I think Martin and Ed have surpassed Chinook with their "hobby" programs, but I guess thats no match for funding and top of the line hardware.

[Alex_Moiseyev]

In terms of perfectness and "solving checkers" we should probably address several statements and see in which one program hits.

1. Proof that program always makes best move and don't miss any wins or draws.

2. Proof that program handles properly each position in 24 pieces ending databas.

3. Proof that program don't lose in 2-moves and 3-moves games.

4. Proof that program today plays better than any other program and human.

5. Proof that program don't lose in GAYP.

As of today we should say that checkers were solved only at the lowest level - #5.

I had a phone call from New York Times when Chinook Team made announcement about solving checkers. They didn't want to find a true, but continue to ask one question only: "Checkers were solved ?"

[Jason Solan]

In terms of perfectness and "solving checkers" we should probably address several statements and see in which one program hits.

1. Proof that program always makes best move and don't miss any wins or draws.

2. Proof that program handles properly each position in 24 pieces ending databas.

3. Proof that program don't lose in 2-moves and 3-moves games.

4. Proof that program today plays better than any other program and human.

5. Proof that program don't lose in GAYP.

As of today we should say that checkers were solved only at the lowest level - #5.

I had a phone call from New York Times when Chinook Team made announcement about solving checkers. They didn't want to find a true, but continue to ask one question only: "Checkers were solved ?"

I'll admit I have not read his paper or book yet about the proof. But 2-Move and 3-Move are subsets of GAYP. If the program cannot lose at GAYP why would it not stand to reason that it would not lose at least 2-Move? I guess I could see it might be possible to lose at 3-move if the proof was done in such a way (as it sounded from the video) that only 1 line is needed to be known to draw to prove it's a draw. Therefore if the computations were done such that not all 3-move openings were calculated, only a draw found from each of the opening moves, then I suppose that would be possible. With 2-move though, the program would need to defend against every possible opening.... actually as I type this out I guess it would still be possible if the program only choose 1 line of defense to prove a draw (i.e. if playing red, only open 11-15 and if playing white and red opens with 11-15 then only ever play 22-18 [sounds like most GAYP tournaments doesn't it?]).

So I guess what I'm saying is I understand your point, but will give my own comments.
I don't think 1 is attainable, because the "best" move is subjective (is it better play with draw line X which has a more difficult draw or better to play with draw line Y which has more shot opportunities which could lead to wins).
2 will be achieved at some point. With machines able to hold terabytes of ram and running multiple cores each capable of hundreds of billions of instructions per second, all it takes is someone with enough motivation to dedicate the time to it. Granted that time will be measured in years, but it would not surprise me to see it happen before chess is proven at the #5 level.
3 I feel could be proved with some effort on their part if they wanted to.
4 I believe that it is impossible to prove that something is "better" than another when the perfect result is a draw.

[liam stephens]

1. Proof that program always makes best move

It certainly won't do that.

Set up one of the standard man down draws such as Payne's single corner draw.

Instead of trying to win with the 3 pieces the database will simply give up a piece making it a 2X2 easy draw position.

And another question to be added to the list:

6. Has the program "solved" 11 Man Ballot ?

[Jay H]

Yeah.....
My problem with Schaeffer was that he "solved" Checkers...he solved nothing.
He programmed a computer to play Checkers is what he did.
That he handed the game over to the media to be steamrolled is what I will never forgive. He wanted a sensational story for himself, at the expense of the game's reputation amongst the uninitiated.
The ultimate cut was that , in one of the last pieces of news coverage regarding Chinook "solving" Checkers was that the game (Checkers) is now in a class with tic-tac-toe, which, in all truthfullness, is even further down the scale then the negative comparisons the media has given Checkers when mentioned in the same article as Chess.
One thing Schaeffer/Chinook will never have is a victory over Tinsley............the 6 draws and forfeit do not count as a victory. Period.

Jay " " H

[tommyc]

Easy there Jay.........................no one can beat Tinsley ,hes dead.And while if he was alive today the computer prog just gets better while Tinsley as he got older gets worse then the machine wins.

[George Hay]

I like Schaeffer, and so did Tinsley. In a roundabout way, Schaeffer and Tinsley probably brought me back to checkers. On a personal note, I was born in 1955, the year Tinsley won the championship. My maternal grandmother taught me to play checkers as a strategic game. In high school I played chess, as Fischer won the championship. Even then, it was chess and checkers for me. Well, the years went by, but I heard reports of the Tinsley-Chinnook match, and nobody had to tell me who Tinsley was. Then somewhere I heard or read about checkers being solved, and here I am. I think on balance, the publicity has been very good for checkers. --George Hay

[George Hay]

Dr. Schaeffer uses algebraic notation (A1- H8) used in chess and Russian/Brazilian Checkers 64. He claims it is easier to use than regular checkers notation. Of course, for a chess player like Schaeffer it would be easier. One checkerist reviewer of One Jump Ahead found that algebraic notation was difficult to follow. So it looks like whatever notation you are used to is the easiest to use. But Richard Pask, who considers it "ludicrous" to use algebraic notation for checkers, asks the rhetorical question: Why refer to 32 non-playing squares?

I found the above Pask quote in 'Move Over' Supplement page 9. Pask's comments on the Tinsley-Chinook matches are in The Legendary Tinsley pages 3-4. Both electronic publications are available from the Checker Maven.
http://www.bobnewell.net/checkers/paskpdf.html

I hope Richard Pask writes a new checkers/draughts book for the general public. He could put the Tinsley-Chinook matches and what it means for checkers to be "solved" in perspective, as well as clarifying checkers tactics and strategy.
--George Hay

[Danny_Alvarez]

I personally think that Mr Schaeffer's work and most of the publicity was good for checkers..... now the fact that he grabbed a headline and claims it is solved wasnt good in my opinion.... also i dont see how it is solved yet.... no disrespect to Mr Schaeffer (whom i admire) or his team.
the example he gave about re-filling the pacific ocean is awesome and it shows the depth of our beloved game. the number of possible moves which from memory i think it was 5x10 to the power of 20 is also eye opening to the people that think it is a kid's game

cheers
Danny

[George Hay]

Did you ever read something and then had a hard time finding it!? Well, I did, but I finally found it!
This is an entry in Vocabulary Of Checkers by W.T. Call (New York, 1909). Under the entry Improving the board
W.T. Call quotes Dr. T.J. Brown as reported in "The Draught Board for November, 1871." Dr. Brown claims
that a board with all single corners would result in a win for the player that moves first. Conversely, for a board with all
double corners, the player that moves second would win. Dr. Brown used the example of a standard checkerboard with strips of paper over outer squares to produce a 4 single corner 7x7 board, and a 4 double corner 7x7 board. Dr. Brown concludes, "As our board is a combination of the two, the result is clear that the forces are equal, and, properly played, every game would be a draw." W.T. Call adds, "The theoretical conclusion that every game of checkers should end in a draw is supported by the facts of experience."
--George Hay

[Alex_Moiseyev]

W.T. Call adds, "The theoretical conclusion that every game of checkers should end in a draw is supported by the facts of experience."

My move is 11-15 ... Where is your move ?

Theoretically I understand that every game is drawn, however this scientific (or pseudo-scientific) statement has very little influence on specific game or move.

For instance, I know and this is scientific fact - some day I will pass away and join with majority ... so what ?!?!?! Knowing this fact doesn't help me much. I don't know even - when this happen

[George Hay]

My reply is 22-18, but not everyone would agree!
It seems to me that opening theory is still, well wide open! The mapping out of the game of checkers is part of the charm. It will take a 24-piece database to "settle" some questions, and perhaps "strongly" solve checkers!
Even then there is still the psychology of this great mind sport!
--George Hay

[Alex_Moiseyev]

My reply is 22-18, but not everyone would agree!
It seems to me that opening theory is still, well wide open! The mapping out of the game of checkers is part of the charm. It will take a 24-piece database to "settle" some questions, and perhaps "strongly" solve checkers!
Even then there is still the psychology of this great mind sport!
--George Hay

11-15 22-18 15x22 ... I have several couple questions regarding this specific order of moves in database(s):

1) What is the very first (chronologically) game in database with this specific order of moves ? Who played it and when ?

2) Since then, how many games in database were played with this order of moves ?

BTW. The theory about winning game status on the board with single corners only is inaccurate and I disagree. This is true only when both sides have 1 piece. But with just two Kings for white and red I can demonstrate you several positions where draw exists.

Alex

[George Hay]

Mr. Alex,
1. The Single Corner goes back to the Ancients, as Jim Loy would call them, the early draughts authors.
Many of those games are animated on Eric's http://icheckers.net/
In English, the first single corner games are in William Payne's draughts book in 1756.
2. There must be more single corner games in databases than I could possibly count!
3. I was wondering about that myself. I saw the posts about loosing checkers, aka give away, aka suicide.
viewtopic.php?f=8&t=2093
As it turns out, it is possible to get a draw in loosing checkers!
But I am not sure that disproves Dr. Brown's theory. His claim is for perfect play, not that a draw is not possible! For a 7x7 board with four double corners it seems the result of perfect play would be a draw. Except when you consider the game of Qubic, as explained by Roberto Waldteufel.
http://www.wylliedraughts.com/ >>Programer>>Qubic
I was shocked to find out that the 4x4x4 even numbered game of Qubic is a first move win with perfect play!
(If you do decide to download the Qubic Game, make sure you have an empty Qubic folder to save it to, so as not to have the files sprawl all over your desktop!)
--George Hay