2-8 might be better, as the white score is not increasing. The PV is 2-8 43-39 7-12 49-43 23-28 41-36 20-25.
Yes Fabien with 2-8 the black position seems far better to reach the draw.
BTW I have now identified clearly several basic improvments in order to help Damy to solve such position. I will now take several days to program and run some new tests.
Though I have some other improvments in mind my new version is far more satisfactory. I am now able to contiue our discussion.
Let's take your last proposal i.e. 2-8. Instead of 43-39 my answer is 30-25 in order to follow by 42-38.
Black to play
For the time being I did not find any convincing move to draw with black.
What is your proposal?
Is this position a losing one?
I propose the following defense:
8-12 with the following idea 49-43 4-9 40-35 23-29 34x23 18x29
Yes, 8-12 is a serious candidate. Another possible sequence is: (8-12) 39-33 32-38 33-28 38x27 28x10 27-32 10-5 32-38.
Hi Fabien,
Obviously your sequence 8-12 39-33 is a little difficult to handle for black. Though I doubt white can find a win with this sequence I prefer to change the order of move and begin by the move 4-9 (instead of 8-12).
Is it still possible to continue the game?
After 35...7-12 36.40-35 the position is the following
Black to play
In the game black played 15-20 and my view is that it is a losing move but what about
the move 4-10.
The justifcation is the following : if white continue as in the game it could follow:
4-10 49-43 23-28 34-29 14-19 and we reach the position
White to play
If now 30-25 then 19-24 29x20 15x24 26-21 18-23 31-27 10-14 27x38 12-17 draw
TAILLE wrote:I am still not quite sure 35...7-12 is a losing move!
After 7-12 35-30 (as in the game) I propose 23-29 34x23 18x29
white to move
30-25 [49-43 14-19 looks like changing the order of moves] 4-9 49-43 14-19 42-38 12-17 38x27 19-23 draw
Your line is convincing, but you still have to consider alternative white moves, such as (7-12) 39-33.
Fabien.
Position after (7-12) 39-33
Black to move
For the time being I do not see any good move for black.
Does that mean that the sequence played 7-12 40-35 15-20 was a succession of three mistakes?
I confess I did not imagine Scan could have made a mistake in this game!
For the time being I do not see any good move for black.
Actually black has good chances, but it's probably hard to prove: (7-12 39-33) 3-9 49-43 9-13 33-28 14-20 28x37 20-25.
Does that mean that the sequence played 7-12 40-35 15-20 was a succession of three mistakes?
I confess I did not imagine Scan could have made a mistake in this game!
It seems to me that you are trying to analyse, with infinite resources, a game that had to be played at about 30 seconds per move. Apparently, this game required extreme precision from both players (fight between win and draw); of course you'll find mistakes.
For the time being I do not see any good move for black.
Actually black has good chances, but it's probably hard to prove: (7-12 39-33) 3-9 49-43 9-13 33-28 14-20 28x37 20-25.
Does that mean that the sequence played 7-12 40-35 15-20 was a succession of three mistakes?
I confess I did not imagine Scan could have made a mistake in this game!
It seems to me that you are trying to analyse, with infinite resources, a game that had to be played at about 30 seconds per move. Apparently, this game required extreme precision from both players (fight between win and draw); of course you'll find mistakes.
Yes Fabien I do not forget that it was a rapid game. For your information, even with a lot of time, Damy chooses 40-35 instead of 39-33!
My goal is to try and understand if an improvment of my program is possible!
Maybe I can say that 39-33 looks like a great centre move compare to 40-35 and it seems it remains true through the different sequences. When analysing the work done on this game I conclude that the centre is far more important in the endgame. The draws indentified are always the results of a big center in compensation of the man down.
I will experiment some modification of my eval function to give more importance to the center.
BTW is your centre evaluation depending of the number of pieces remaining on the board?
TAILLE wrote:Yes Fabien I do not forget that it was a rapid game. For your information, even with a lot of time, Damy chooses 40-35 instead of 39-33!
Scan also likes 40-35 a lot, as it appears to give white a huge advantage. But black can draw; kudos to you for finding the line.
BTW is your centre evaluation depending of the number of pieces remaining on the board?
In Scan, the scoring of any evaluation feature can depend on the game phase. It is up to the learning algorithm to decide whether to use that option.
I have, however, no idea what the features are. Even if you could describe "centre evaluation" precisely, I would have trouble finding out what Scan thinks of it.
