A bug yes but where?

Daniel Anulliero · Post by **Daniel Anulliero** » Tue Sep 16, 2014 7:14 pm

Hi all
My chess engine ISA is current in developpement
It has only alpha beta + PVS + quiesce , hashtables and eval with materiel , pst , passed , doubled ,isolated pawns , square rule , and rule for winning KBB vs K and KBN vs K.

I use this kind of position to see if hashtables work :
FEN "4k3/8/4pPp1/3pP1P1/2pP4/2P3K1/8/8 w - - 0 1"

Here is the list of moves by ISA :

Code: Select all

1. Kf4 &#123;&#40;g3f4&#41; +7.65/28 12&#125; Kf7 2. Ke3 &#123;&#40;f4e3&#41; +7.95/27 10&#125; Ke8 
3. Kf4&#123;&#40;e3f4&#41; +10.84/27 10&#125; Kf7 4. Ke3 &#123;&#40;f4e3&#41; +7.65/24 10&#125; Ke8 5. Kd2 &#123;&#40;e3d2&#41;+7.95/24 10&#125; Kf7
6. Kc2 &#123;&#40;d2c2&#41; +12.76/27 10&#125; Kf8 7. Kd2 &#123;&#40;c2d2&#41; +17.78/2710&#125; Kf7 8. Kc1 &#123;&#40;d2c1&#41; +8.96/24 10&#125; Ke8 
9. Kb2 &#123;&#40;c1b2&#41; +17.78/24 10&#125; Kd710. Kc2 &#123;&#40;b2c2&#41; +7.95/19 10&#125; Ke8 11. Kb1 &#123;&#40;c2b1&#41; +12.76/25 10&#125; Kd7
12. Kb2&#123;&#40;b1b2&#41; +7.50/18 10&#125; Ke8 13. Ka3 &#123;&#40;b2a3&#41; +17.78/23 10&#125; Kd7 14. Kb4 &#123;&#40;a3b4&#41;+8.96/18 10&#125; Ke8 
15. Kc5 &#123;&#40;b4c5&#41; +M9/20 1&#125; Kd7 16. Kb5 &#123;&#40;c5b5&#41; +10.84/1910&#125; Ke8 17. Kc6 &#123;&#40;b5c6&#41; +M9/20 2&#125; Kd8 
18. Kd6 &#123;&#40;c6d6&#41; +M7/16 1&#125; Ke8 19.Kxe6 &#123;&#40;d6e6&#41; +M6/14 0&#125; Kf8 20. f7 &#123;&#40;f6f7&#41; +M5/12 0&#125; Kg7 
21. Ke7 &#40;e6e7&#41;+M4/10 0&#125; Kh8 &#123;&#40;g7h8&#41; -M4/9 0&#125; 22. Kf6 &#123;&#40;e7f6&#41; +M3/8 0&#125; Kh7 23. e6 &#123;&#40;e5e6&#41;+M2/6 0&#125; Kh8 
24. f8=Q+ &#123;&#40;f7f8&#41; +M1/4 0&#125; Kh7 25. Qg7# &#123;&#40;f8g7&#41; +M0/2 0&#125; 1-0

It find the winning way but I've noticed two stranges king take back &#58;


7. Kd2 &#123;&#40;c2d2&#41; +17.78/2710&#125;
and
10. Kc2 &#123;&#40;b2c2&#41; +7.95/19 10&#125;

with a severe drop of the eval

Do you think it's a bug in search?hashtables?eval?
Bests
Dany

stegemma · Post by **stegemma** » Tue Sep 16, 2014 8:20 pm

Daniel Anulliero wrote:Hi all
My chess engine ISA is current in developpement
[...]
Do you think it's a bug in search?hashtables?eval?
Bests
Dany

Maybe a bug in the search has broken the hashtable causing a bad eval?

Daniel Anulliero · Post by **Daniel Anulliero** » Tue Sep 16, 2014 8:50 pm

ah ah
good but not so useful

cdani · Post by **cdani** » Tue Sep 16, 2014 9:45 pm

You can try to disable or reduce the hash and view if it plays this.
It's also possible a bug where it selects a line and discards other possibilities without analyzing them. May be excessive cuts. It's really difficult to say.

Daniel Anulliero · Post by **Daniel Anulliero** » Tue Sep 16, 2014 10:16 pm

I suspect some bad value saved in the hashtable because the big jumps of the eval +17 and +7

stegemma · Post by **stegemma** » Tue Sep 16, 2014 11:32 pm

Daniel Anulliero wrote:ah ah
good but not so useful

It could be almost anything but even it could be related to the horizon effect. Maybe you quiescence search has some problem, so that a certain level it won't works has expected.

We can suppose anything but onyl debugging and then debugging and then debugging you can find the bug.

A bug yes but where?

A bug yes but where?

Re: A bug yes but where?

Re: A bug yes but where?

Re: A bug yes but where?

Re: A bug yes but where?

Re: A bug yes but where?