Internet engine matches

[TAILLE]

Hi Bert,

Another interesting observation, but i dont have a clue how to explain it.
The 10 wins by Kingsrow was 2 times with white and 8 times with black.
The Damage wins was 12 times with black.

Bert

I cannot know what could be the exact reason why Damage never win with white but at least I can suggest a hypothesis: if, for the positions with 1 king against 4 men (plus an egal number of men in each side), your evaluation function gives always the advantage to black then you will always look for a draw, even if your position is really good!

Let me give you a clue : from my own experience the white moves seem more difficult to discover (for a program!) than the black ones! At least it is the case with Kingsrow which, IMO, play better with black than with white.

[BertTuyt]

Gerard, it was an interesting experiment.
Although I'm not sure what I learned from it
But the games were sometimes very weird, especially some ultra quick wins (from both Damage as Kingsrow) after 16, 17, 21, or 24 moves (in general 55% of the win before move 30, and in total 86% before move 40) . Do you see the same?

I assume I need to dig deeper into all the win/loose games to find some overall clues, which could improve the evaluation and/or search.
Anyway thanks for sharing.
Is this a known position, and proven to be a draw (or whatever).

In the mean time I fixed a few bugs in the damage GUI.
And now I'm working on some major architectural changes in the Damage engine (and also with consequences for the Horizon engine).
After this change i will post a new GUI and Horizon engine on the Internet.
And then continue with the improvements i plan for the damage engine.

Bert

[TAILLE]

Gerard, it was an interesting experiment.
Although I'm not sure what I learned from it
But the games were sometimes very weird, especially some ultra quick wins (from both Damage as Kingsrow) after 16, 17, 21, or 24 moves (in general 55% of the win before move 30, and in total 86% before move 40) . Do you see the same?
Bert

Yes Bert the majority of the games were also very quick with Damy

Is this a known position, and proven to be a draw (or whatever).
Bert

No it is only my own invention in order to work more easily on the evaluation of specific position type.

[Krzychumag]

Gerard, it was an interesting experiment.
Although I'm not sure what I learned from it
But the games were sometimes very weird, especially some ultra quick wins (from both Damage as Kingsrow) after 16, 17, 21, or 24 moves (in general 55% of the win before move 30, and in total 86% before move 40) . Do you see the same?

I assume I need to dig deeper into all the win/loose games to find some overall clues, which could improve the evaluation and/or search.
Anyway thanks for sharing.
Is this a known position, and proven to be a draw (or whatever).

In the mean time I fixed a few bugs in the damage GUI.
And now I'm working on some major architectural changes in the Damage engine (and also with consequences for the Horizon engine).
After this change i will post a new GUI and Horizon engine on the Internet.
And then continue with the improvements i plan for the damage engine.

Bert

Bert when you throw open programme Damage and Horizon.

[BertTuyt]

I'm working on some hard bugs which are difficult to solve.
Nevertheless there is already a previous version available , so Damage GUI and Horizon Engine.
I'm not sure if I already forwarded you the link.
If not, otherwise drop me an email..

Bert

[Krzychumag]

I'm working on some hard bugs which are difficult to solve.
Nevertheless there is already a previous version available , so Damage GUI and Horizon Engine.
I'm not sure if I already forwarded you the link.
If not, otherwise drop me an email..

Bert

I apologise but you sent to me nothing.

[Krzychumag]

I apologise that I am asking Bert but you when you will give new versions of program Damage.

[MichelG]

I assume/hope that also Michael (Dragon) reads this Forum.
It would be nice if he would be able to play a 158 games match against Kingsrow (but im not sure if he has this program).

Dragon's DXP implementation issues seem fixed at the moment, at least it can play matches against dam 2.2.

I don't have Kingsrow, but i am happy to play a match against it or other programs in some way.

Michel

[BertTuyt]

Michel, and did you already play a 158 game match against Dam 2.2X, and if so what was the result?

Bert

[MichelG]

Michel, and did you already play a 158 game match against Dam 2.2X, and if so what was the result?

Bert

I just did some small matches to validate the DXP. Last (very short) match ended with 3 wins, 0 loss, 2 draws. I'll try to do a 158 game match the coming time.

Michel

[MichelG]

I did a match against dam yesterday.

One issue to consider is what hardware you use. As dam does not support multi threading, using a modern computer gives dragon a major advantage.

I tested with these conditions:
- 6 minutes per game on an i7
- no endgame databases
- if after 100 moves the game has not ended, it is declared a draw (even if it is 4v1 endgame that is not won yet)
- Dragon used 2 cores (of 6 available), dam used 1
- dragon used random openings from the book

This gives 93 wins, 63 draws, 5 loss (77% score, +210 elo points) after 161 games. The match also turned up some glitches, so i have some work to do

Michel

[Krzychumag]

Hi Michel.

When the new version of programme Michel.

[MichelG]

When the new version of programme Michel.

That will probably there on friday

- Thinking on opponent time (pondering) should be fixed
- Some glitches involving endgame databases should be fixed.

Michel

[MichelG]

Hi Michel.

When the new version of programme Michel.

It's available for download now:

http://mdgsoft.home.xs4all.nl/draughts/beta4.1.html

Michel

[Krzychumag]

It's available for download now:

http://mdgsoft.home.xs4all.nl/draughts/beta4.1.html

Michel

It thanks Michel. I noticed that during game with programme Dam - programme shows losses,victories and draws. But if he is playing with other program isn't already showing it. During game programme or game white or black. There is as this way as in other programs so that he plays alternately, once no with whites time black still through the entire match.