Author: Christophe Theron
Date: 13:53:03 10/23/00
Go up one level in this thread
On October 23, 2000 at 13:35:35, Bas Hamstra wrote:
>On October 23, 2000 at 00:10:39, Christophe Theron wrote:
>
>>On October 22, 2000 at 17:46:47, Bas Hamstra wrote:
>>
>>>On October 22, 2000 at 17:07:54, Uri Blass wrote:
>>>
>>>>On October 22, 2000 at 16:16:19, Bas Hamstra wrote:
>>>>
>>>>>On October 21, 2000 at 14:30:52, Thorsten Czub wrote:
>>>>>
>>>>>>Don't kill me - i am only joking ! Please don't start another
>>>>>>ideology battle :-))) here.
>>>>>>
>>>>>>Seems Gambit-Tiger plays not the weakest games...
>>>>>
>>>>>Well, maybe I exaggerated slightly when I said GT plays boring. But speculative
>>>>>eval is ANYTHING but a new paradigm.
>>>>>
>>>>>Only a pity that it thinks it needs killerlines in a tournament against
>>>>>amateurs. I mean bookmoves that are NO theory. I mean the continuation of
>>>>>unplayable moves, until it's basically mate in 10. See Tiger-Tao.
>>>>
>>>>The book moves in Tiger-Tao are theory moves.
>>>>
>>>>I looked in an old book like f5book.ctg and there are 2 replies for 12.exd5(Qa5
>>>>and Bg7)
>>>>
>>>>Uri
>>>
>>>Yes. But Nxc2 is NOT theory, because it's deadly. It's the move to avoid. And GT
>>>had it in it's book, as well as even moves AFTER Nxc2. Up to the point that
>>>NeuroChess could have easily mated Deep Blue.
>>>
>>>Look, Cock the Gorter (opening specialist) stood behind me and said: "if you
>>>play Nxc2 the game is over. You will be mated instantly." Afterwards he
>>>seriously complimented me, for not being mated for 8 moves or so.
>>>Though I don't really blame 'em, it IS frustrating when you sit there.
>>>
>>>Now compare that to Roland (Patzer), who played with normal book in stead of
>>>killerbook against the amateurs. Had he killed amateurs like that, Patzer would
>>>have been ranked even higher.
>>>
>>>Ciao,
>>>Bas.
>>
>>
>>
>>4 points:
>>
>>1) I have been myself the victim of better books for a long time, because I had
>>written my book myself and it was very small. For example I remember I have been
>>caught by Diep in the blitz world championship (Paris, 1997). You have to live
>>with this, it happens. If you don't want it to happen, you have to find somebody
>>to write a strong book for you, and you can maybe also write automatic
>>algorithms to adapt an automatically generated book to the style of your
>>program.
>>
>>2) Jeroen has not written a killerbook. A killerbook is a book in which you
>>insert games won by your program so it can replay them against KNOWN opponents.
>>Your program as well as the other programs in this tournament are not known
>>opponents.
>>
>>3) Jeroen inserts novelties he has studied himself in his books. This is a hard
>>job to find novelties, and I don't see why it should not be rewarded also.
>>
>>4) Gambit Tiger is going to be available soon. You can try to let your program
>>play against it with books disabled for both, and inform us about the results.
>
>I realize I whined. I will stop that. But a few tiny comments:
>
>It depends on you definition if you call that line a killerline. My definition
>of a killer line happens to be exactly the one that Ed Schroeder has, isn't that
>a coincidence? In my book a line is a killerline if it contains several
>non-theory moves that are only meant to trap the opponent engine. Of course you
>CAN call such moves "novelties".
I have been caught by this myself when I was amateur.
An amateur program can indeed have this problem. On the other hand, a commercial
program has other problems. It is obvious that we cannot re-write our opening
book for each tournament, and some people take advantage of this to insert
killer lines aimed against a known commercial program, hoping that the faulty
line is still not fixed in the tournament book.
It happened to Ed some years ago, and it's not more funny than the problem you
have...
>Unfortunately you are right in stating that these tournaments nowadays are not
>only fights between engines, but also between books.
>
>Finally: Jeroen is a cool guy. After I resigned, we played a 30/0 game right
>away with random (non killer) book, just for fun. Tiger won. It seems you wanted
>to hear that?
You caught me on my defensive. It's because the book argument reminds me too
much of Vincent's whining. Sorry, I see you are different.
There is something else to take into account.
Current amateur programs are not too far from top commercial programs. Often the
difference is 150 elo points, sometimes it is much less.
So the winning chances of the commercial programs over amateurs is probably
between 55% and 75%. Out of ten games, the commercial program can expect to be
losing 3 to 5 games (approx.).
If an amateur program loses to a commercial one, nobody notices. It sounds
normal. Actually it is not. What is normal is that the commercial program loses
from time to time.
Now if the commercial one loses, BIG DEAL! Wow, look! This small amateur program
has won the big name! Incredible!
Actually it is a normal thing and it is expected to happen in 25% to 45% of the
games, but people don't understand this, and instead consider the commercial one
has been ridiculed and the amateur has made a great achievement.
You have nothing to lose when you face a commercial program. The commercial one
can lose much more than you.
So I don't feel guilty at all to use all the means I have to try to avoid such
"disasters".
Christophe
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.