Maximus vs. Schwarzman draughts match

[BertTuyt]

In below diagram Game 2 , black to play (Maximus), is the move 44. .... 7-11 a draw ??



Think that Maximus played 24-29.

Possible move sequence:

44. ....... 7-11
45. 27-22 11-16
46. 22-17 13-18
47. 43-39 24-29
48. 39-34 18-22
49. 34x23 22x11
50. 23-18 11-17
51. 28-22 17x28
52. 18-12 16-21
53. 12- 7 21-27
54. 7- 1 28-33

Bert

[Ed Gilbert]

In below diagram Game 2 , black to play (Maximus), is the move 44. .... 7-11 a draw ??

Kingsrow does not find a draw. After 44. .. 7-11 43-38 13-18 40-34, kingsrow thinks white will be up a man.

Edit: what about 44. .. 13-18 45. 28-22 7-11 46. 22x13 19x8, looks drawn.

-- Ed

[Ed Gilbert]

The position after 37. 27-22.


black to move

Maximus played 6-11 and lost. Kingsrow thinks that Maximus could have drawn with 12-18 instead. 12-18 22-17 7-12 17x8 13x2 32-27 18-23 37-32 26-31 probable draw.

-- Ed

[TAILLE]

The position after 37. 27-22.


black to move

Maximus played 6-11 and lost. Kingsrow thinks that Maximus could have drawn with 12-18 instead. 12-18 22-17 7-12 17x8 13x2 32-27 18-23 37-32 26-31 probable draw.

-- Ed

Though Damy choses also 12-18, the move 6-11 seems to have almost the same value. After 12-18 22-17 7-12 17x8 13x2 32-27 18-23 37-32 26-31 28-22 Damy continues to consider that black have a lot of difficulties and must play to find the draw.
In any case that means that Schwarzman managed to create in the middle game an advangeaous position and this point is really the challenge we have to face as programmers. We all know that our programs are very strong in the tactical area and in the endgame but against a very strong player like Schwarzman it may often happen that the human will be able to avoid tactical mistakes and will be able to take positionnal advantage in the middle game.

BTW what is your answer to the following question:
Against Schwarzman who has the best chance to avoid losing a game: Georgiev or Maximus?

[Wieger Wesselink]

After 25 moves Maximus had only a minor problem (the ugly piece on 5), but apart from that the black position looks healthy. 26... 3-8?? is a positional mistake that a human player would not make. After this move the black left wing is locked, and this is the main cause of the black loss.

[Rein Halbersma]

After 25 moves Maximus had only a minor problem (the ugly piece on 5), but apart from that the black position looks healthy. 26... 3-8?? is a positional mistake that a human player would not make. After this move the black left wing is locked, and this is the main cause of the black loss.

Hi Wieger,

What is your opinion an effective counter-computer strategy for humans? Was it a smart choice of Schwarzman to let black occopy square <36> to create positional opportunities? Are there any other structures where humans might outplay computers?

Rein

[BertTuyt]

After 37 27-22 also Damage wants to play 12-18

12-18 22-17 7-12 17x 8 13x 2 32-27 18-23 37-32 26-31 28-22 2- 7 27-21 6-11 21-17 7-12 16x18 23x21

The move sequence also seems to lead to a drawisch position.

So I tend to agree with Ed.
Where does Damy want to play something different??

Bert

[BertTuyt]

3th game 1 - 1

Bert

[Eric van Dusseldorp]

The position after 37. 27-22.


black to move

Maximus played 6-11 and lost. Kingsrow thinks that Maximus could have drawn with 12-18 instead. 12-18 22-17 7-12 17x8 13x2 32-27 18-23 37-32 26-31 probable draw.

-- Ed

At the end your program gives (12-18) the score: -7.
That means: surely draw, and not probable draw. Am I right?

But also for Kingsrow it takes a very long time to conclude that (12-18) is a better move than (6-11).

[Ed Gilbert]

At the end your program gives (12-18) the score: -7.
That means: surely draw, and not probable draw. Am I right?

No. Unlike database win or database loss scores, which are conclusive indications of those results, database draw scores are not necessarily conclusive. The reason is that during a search, database draw scores of +/- 1, 3, or 5 get compared to other eval scores, some of which are endgame db scores and some heuristic eval scores. If the PV has a database draw score, it only means that no other variation leads to a better score, but those other variations can have both db scores and heuristic scores. Heuristic scores are of course subject to being inaccurate.

In practical use, you can usually tell when a database draw score has a high probability of being correct, and when you should not trust it. I look at the number of pieces that are on the board, and during the iterative deepening search I look at how many successively deeper searches return a database draw score. If for example there are 20 pieces on the board and after a long search you finally get the result of db draw, this is not a conclusive result. But with 16 or fewer pieces on the board, if you get db draw scores returned at depth N, then N+1, N+2, N+3, N+4, N+5, then this is an indication with high probability of it being a true draw. But it is not a certainty.

+/- 7 scores are used in kingsrow to indicate a draw by repetition. These results have the same uncertainty as database draw scores, and for the same reasons. Also, in some small percentage of cases, repetition draw scores can be incorrectly indicated due to something called the Graph History Interaction problem that affects most game programs using transposition tables.

There is more about this subject in the Kingsrow help file, in the section "Using Kingsrow", "How do I...?", "Know for certain that a position is a win or draw or loss?".

-- Ed

[TAILLE]

Hi Bert,

After 37 27-22 also Damage wants to play 12-18

12-18 22-17 7-12 17x 8 13x 2 32-27 18-23 37-32 26-31 28-22 2- 7 27-21 6-11 21-17 7-12 16x18 23x21

The move sequence also seems to lead to a drawisch position.

So I tend to agree with Ed.
Where does Damy want to play something different??

Bert

My point point was only the following : for Damy the choice between 6-11 and 12-18 is rather difficult. Damy is unable to calculate that 6-11 will lead to a probable loss or to calculate that 12-18 will lead to a probable draw. In both cases the black position seems difficult showing Schwarzman managed to outmatch Maximus in the middle game. This last point is the key point. The brut force of the current computers is not sufficient to take an advantage in the middle game against a very strong player! Though we all use various mechnisms to try and eliminate unintersting variations, programs are still wasting to many time on unintersting positions and this is a great lesson.

[Wieger Wesselink]

Hi Wieger,

What is your opinion an effective counter-computer strategy for humans? Was it a smart choice of Schwarzman to let black occopy square <36> to create positional opportunities? Are there any other structures where humans might outplay computers?

Rein

Hi Rein,

I think that letting black occupy square <36> was a good idea. In general the extra pieces on <36> and <47> reduce the probability of a draw. I think that is why many strong players try to tempt their opponents to occupy <36>. Apart from that the piece on <36> is considered to give white a slight advantage.

It's hard to say what is the best strategy against a computer. Many strategies simply don't work because it is too easy for the computer to calculate a draw. But I know certain game types where computer programs make very poor judgements of the position. A locked wing is one of them, on both sides of the board. The big question is how to get such a game type on the board.

Wieger

[BertTuyt]

4th game 1 - 1

Bert

[BertTuyt]

5th game 1 - 1

Bert

[BertTuyt]

6th game 1 - 1.

So match result 5 draws and 1 win for Schwarzman.

Anyway an exciting event which I "watched" every day/evening.
Good promotion for computer draughts.
And I guess some learning opportunities before we will give mankind the final blow.

JJ, thanks for this initiative, and I hope this will be a regular event every year.

Could you share some details, now the match is over?
Like hardware used, your learnings,new insights, and improvement required.

Like a famous general learned us, one battle lost, but we certainly will win this war in the end

Bert