Damage does not have (yet) the 7 Piece Endgame Databases

[Ed Gilbert]

Hi Gerard,

Are you referring to the 25-move draw rule? I have not given this any thought at all, and I ignore it when generating endgame dbs.

Why does this multiply the disk space by x2 or x3?

-- Ed

[TAILLE]

Hi Gerard,

Are you referring to the 25-move draw rule? I have not given this any thought at all, and I ignore it when generating endgame dbs.

Why does this multiply the disk space by x2 or x3?

-- Ed

No Ed. if you look for example at the game Baliakine-Heusdens (last NL championship) the following position were reached after the 61 black move :


White to move

obviously it is a draw but the two players continued the game until the 76th white move to reach the following position :


Black to move


Are they becoming mad ?

The reason is the following new rule :
When a "standard" draw is reached you count the remaining pieces, assigning 1 point to a man and 2 points to a king. If one player obtains an advantage of 3 points or more the result of the game is "draw+" for him and "draw-" for the opponent. It is then a kind of small win and, of course, this limits the number of "draw"

By the way I do not know the result of the above game (in order to obtain a draw+ white has to prove that he can capture the black man which seems not obvious)

Gérard

[TAILLE]

After a look at the KNDB site the result of the tournament is the following :
Baliakine 16 points with 1 draw+
Heusdens 16 points with 1 draw+ and 1 draw-

Baliakine wins the tournament due to the new rule.

note : when two players reach the same number of points then, the number of draw+ is more important that the number of voctories

Gérard

[Ed Gilbert]

The reason is the following new rule :
When a "standard" draw is reached you count the remaining pieces, assigning 1 point to a man and 2 points to a king. If one player obtains an advantage of 3 points or more the result of the game is "draw+" for him and "draw-" for the opponent. It is then a kind of small win and, of course, this limits the number of "draw"

I was not aware of this. My initial reaction is that I hope it does not apply to computer vs. computer games. If it does, then I think I can improve by making a small modification to the way db draw scores are propagated, rather than changing the databases. At the moment kingsrow does not distinguish between an easy draw and a hard draw.

-- Ed

[ildjarn]

By the way I do not know the result of the above game (in order to obtain a draw+ white has to prove that he can capture the black man which seems not obvious)

It ended a draw+ for white, because Heusdens didn't want to play any further.

[TAILLE]

I was not aware of this. My initial reaction is that I hope it does not apply to computer vs. computer games. If it does, then I think I can improve by making a small modification to the way db draw scores are propagated, rather than changing the databases. At the moment kingsrow does not distinguish between an easy draw and a hard draw.

-- Ed

Another solution could be to keep the current db as they stand but to generate another db in the same manner but starting only with the draw positions in order to distinguish draw-, draw and draw+.
It looks quite similar to the existing db; you generate the db beginning with the 3 against 1 piece db (the first database allowing to obtain a draw+) and you continue as usual.

There are no impact on the current db. If the current db give as a result a "draw" then you look for the new db (draw+, draw, draw-) and, if this db is generated, you give the new result to the engine.

Gérard

[TAILLE]

By the way I do not know the result of the above game (in order to obtain a draw+ white has to prove that he can capture the black man which seems not obvious)

It ended a draw+ for white, because Heusdens didn't want to play any further.

If I understand correctly that means that Heusdens would have won the tournament if he had continued and had obtained a real draw ! Dramatic !

Gérard

[BertTuyt]

To be honest, im in favor of the old system, so a draw is a draw.
Don't like to win a tournament this way.
So for the time being, I will also not implement it, or raise the question.
If one changes tournament rules, i will follow, but i will never take the initiative.

Bert

[TAILLE]

To be honest, im in favor of the old system, so a draw is a draw.
Don't like to win a tournament this way.
So for the time being, I will also not implement it, or raise the question.
If one changes tournament rules, i will follow, but i will never take the initiative.

Bert

Hi Bert, Ed.

For the time being I have no real preference for adopting or not this rule.
Nevertheless I plan to generate this supplementory db for at least two reasons :
1) For helping humans in analysing such endgame
2) For a challenge to play against a very strong player

It was claimed one day (not that long ago) that a top GMI is always able to impose a draw against a computer. In addition it was said that in the middle game this top GMI may be able to take an advantage against a computer but but he will not be able to eventually win because it is relatively easy for a computer to find the draw by using its large db. Finally it was said that with the new rule (draw+) a top GMI will probably win a match against a computer because he will be able, following an advantageous middle game, to obtain some draw+ results !

I confess that such challenge motivate me.

Gérard

[ildjarn]

If I understand correctly that means that Heusdens would have won the tournament if he had continued and had obtained a real draw ! Dramatic !

Correct.

[wellnesswrotter]

The Draw+ and Draw- rule is only applied in the Dutch Championship for Human Players for now. It is an addition to the tournamentrules to avoid barrage.

It eventually decided this years Dutch Championship.

For game analysis, it would be "nice" to take this in account in the end-evaluation.

[ildjarn]

Personally, I think I would be against draw+ in computer draughts matches. It's more intended for human players to satisfy the 'but I was better!' thoughts after a drawn game. Just like the 25-move rule shouldn't be taken into account for database results (IMO), since that rule is there to avoid unnecessary prolonging games, something which doesn't happen when you have an objective position evaluation which you can prove.

[BertTuyt]

Ed, my counts for the 5K - 2K and 2K - 5K are the same, thanks for the info.

Will now start with the 7p containing 1 man (on the older machine).
On the new one i will start with the 4-3 part.

Do you have slice info of the 7p databases containing 1 man, so i can check results, or do you only have the info for the total slice package.

Next to that, as i will start on the new i7, i need to transfer files.
Do you apply a check (64bit xor or something like that), to make sure that the databases are not corrupted during transfer.
As these calculations are time consuming i don't want to take the risk.

Do you have a bad experience with this , so is it wise to to these checks?

Bert

[ildjarn]

I have some 5v2 results available on http://sanguis.xs4all.nl/dammen/. I used Michel Grimminck's Dragon draughts database generator. Unfortunately it crashes on the 3K+2P/2K database generation, because there are wins that are longer than 251 halfmoves (the maximum for the databases that Dragon uses).

[BertTuyt]

Ildjarn,

thanks for your post.
Do you also have the WDL stats for the static positions?

Bert