Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 1 Hour CCR Test by IM Larry Kaufmann / Patzer 3.51

Author: Peter Berger

Date: 08:42:25 05/11/01

Go up one level in this thread


On May 11, 2001 at 10:40:24, Roland Pfister wrote:

>I had Patzer run over night with 16 minutes per position
>on a Piii-450, 32MB Hash.
>
>1. Qb3  0 4:34
>2. Bc8  4 0:09
>3. Nh6  4 0:00
>4. b4   0 -
>5. e5   4 0:01
>6. Bxc3 4 0:00
>7. Rfb8 1 1:25
>8. d5   0 -
>9. Nd4  4 0:01
>10. a4  4 0:04
>11. d5  4 0:03
>12.Bxf7 4 0:00
>13.Nxe4? 4 0:01
>14.Nxc4? 0 10:25
>15. exf6 4 0:00
>16. d5   4 0:01
>17.hxg4? 4 0:00
>18. Bxf6 0 -
>19. Bxe6 0 -
>20. Ndb5 3 0:21
>21.Kxe6? 4 0:00
>22. a4   0 -
>23. Bxh7 4 0:01
>24. Nxe5 4 0:09
>25.Qxb3? 4 0:00
>
>That is 68 points.
>
>if it would have run with half speed, it would have found
>4 points less, with quarter speed, 8 points less.
>
>if it would have run with double speed, 2 points more,
>with 4fold speed 4 points more

Thanks for posting your results - as expected Patzer did well .

I have done a few similar experiments with different speeds during the last
nights and have got similar results for other programs ( those that can eat and
spit EPD ).

As soon as the programs are good enough ( or the hardware fast enough ) the
tactical ones will be solved - that's the reason all of the programs are so
close together now IMHO .

I am surprised Patzer needs _that_ long to avoid Nxc4 in CCR 14 .

What's left are the positional ones - and they are quite search-resistent : good
result in CCR 7 for Patzer I think .

Btw , I think CCR 8 isn't exactly a good test . The only program I have seen
that played it with the right plan of Nf5 was Chess Tiger 14.0 - all the others
might play Bxf6 , or just as well Nxc6 as most of them planned to trade both
pieces anyway ( that's luck ) .

Cheers.

pete






This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.