Author: Robert Hyatt
Date: 09:06:32 10/21/02
Go up one level in this thread
On October 20, 2002 at 22:32:10, Mark Young wrote:
>On October 20, 2002 at 19:34:54, Bob Durrett wrote:
>
>>On October 20, 2002 at 19:23:38, Mark Young wrote:
>>
>>>On October 20, 2002 at 19:12:39, Bob Durrett wrote:
>>>
>>>>On October 20, 2002 at 18:33:25, José Carlos wrote:
>>>>
>>>>>On October 20, 2002 at 18:16:38, Uri Blass wrote:
>>>>>
>>>>>>On October 20, 2002 at 17:55:55, Bob Durrett wrote:
>>>>>>
>>>>>>>Why Fritz and Junior not among Top computers at ICC?
>>>>>>>
>>>>>>>I just looked at "best computers" at the Internet Chess Club [ICC] and noted
>>>>>>>that the top blitz programs are NOT Fritz and Junior.
>>>>>>>
>>>>>>>{{Note: at ICC, "best" is a command.}}
>>>>>>>
>>>>>>>
>>>>>>>Why is that? Maybe Kramnik was playing against the wrong engine?
>>>>>>>
>>>>>>>Bob D.
>>>>>>
>>>>>>ICC rating is a joke.
>>>>>>
>>>>>>People can maximize their rating by having a list of
>>>>>>players that they do not to play.
>>>>>>
>>>>>>I think that it is better to ignore that rating because
>>>>>>it has no meaning.
>>>>>>
>>>>>>Rating should be based on tournament when players have
>>>>>>no right not to play against special opponents.
>>>>>>
>>>>>>Players who choose not to play against special
>>>>>>opponents should get no rating.
>>>>>>
>>>>>>Another point is that the time control for blitz is not
>>>>>>constant time control and it can be 3 0 or 12 2.
>>>>>>
>>>>>>I think that all games should be played at
>>>>>>the same time control to have rating with meaning.
>>>>>
>>>>> ICC rating is not totally meaningless. It depends on what you want to know. I
>>>>>mean, if I usually play the same time controls and a variety of opponents, I can
>>>>>test my program with two different books and see which one works better. In that
>>>>>case, ICC rating is useful _to me_.
>>>>> But I agree is doesn't always show which player is stronger because of the
>>>>>reasons you mentioned.
>>>>>
>>>>> José C.
>>>>
>>>>OK, so some manipulation is possible. People who used Fritz or Junior could do
>>>>the same thing!
>>>>
>>>>What strikes me is that Fritz and Junior seem to be missing entirely!
>>>>
>>>>Do you suppose somebody has talked the ICC management into boycotting Fritz and
>>>>Junior?
>>>>
>>>>Joke or not, I think it's very strange that Fritz and Junior cannot compete in
>>>>that arena. They could play by the same rules as the others, too!
>>>
>>>
>>>If you think that Ruffian, Yace, and Crafty are stronger then Deep Fritz and
>>>Deep Junior, I think you should stick with them. But don't ever play them
>>>against Deep Fritz yourself or your world will come crashing down.
>>
>>You suggest a computer versus computer match to see which program is best
>>against humans???????
>>
>>Truly, I would lose to all of them! : )
>>
>>But, to be serious, can you say why you feel that way? Are you basing this
>>solely on results of computer vs computer tournament results?
>
>No.
>
>>
>>Also, what about my point that the same rules apply to Fritz and Junior at ICC
>>as apply to the three you mentioned? [Assuming that Fritz/Junior owners are
>>allowed to compete.] Do people owning Fritz and Junior CHOSE not to compete,
>>and if so why would that be true?
>
>I play Fritz and Junior but not on ICC. I have played on ICC in the past and I
>have been ranked #1 more then once on ICC with my computer account. I can tell
>you from knowing. ICC rating are meaningless. It is more about knowing how to
>"spike" the rating system. Scrappy for example does this by only playing against
>certain human players, it will not play computer programs such as fritz because
>it will get killed. Programs can have a much higher rating on ICC then Fritz or
>other programs whos operators don't care about or do not know how to spike the
>rating system.
That is wrong on two counts.
(1) scrappy doesn't play "certain humans". It will play _any_ human with a
rating
that is "reasonable". There is no selectivity at all, other than the rating
limit. If
you play someone rated 800 points _below_ you, your rating is going _down_
because
if you win, you get zero, if you draw you lose 16, if you lose, you lose 32. I
avoid that
simply to "stabilize" crafty (scrappy) so that I can tell if changes have hurt
or helped.
(2) Scrappy doesn't use ! computer to avoid losing to computers. If I was doing
that,
I would also have !computer in "crafty"s account. I have explained the
reasoning
before. I (A) wanted to see what the difference was between playing humans with
a
program that is really tuned to play against humans, vs playing computers with a
program
that is really tuned to play against humans (crafty). (B) More than a few
players asked me
to set up a second account that would only play humans because a few extremely
obnoxious
computer operators would break into a match between a GM and Crafty as soon as a
game
ended, which infuriated IM/GM players. That can't happen.
That is the _only_ reasons scrappy exists. I took it off for a while, in fact,
and got many
requests to "bring it back" so I did. I'd just as soon not have a second
machine running all
the time, but I do...
>
>>
>>It is not at all clear to me that Fritz and Junior are the best programs against
>>humans.
>
>It will never be "clear" as computers do not play in tournaments against rated
>human players. Until this happens we can only use the data we have.
>
>
>I don't know about the others, since I'm not strong enough to test them
>>anyway. But don't discount those programs which are not currently in the
>>limelight. One of them may, indeed, be the best in the world versus humans.
>
>This could be true, but can you show where the SSDF results do not translate
>into the best program playing humans. Fritz is top of the SSDF list. A version
>not to much removed for it drew GM Kramnik in a match. ChessTiger 14 Top the
>SSDF computer rating list it crushed all GM players in a tournament in south
>america. Beating all GM players it faced and going unbeaten in the tournament
>will at TPR of 2788.
>
>This data suggest that Computer vs Computer results translate well for playing
>humans!
>
>What data can you show that supports you views.
>
>>
>>Don't you agree, at least a little bit? Please?
>
>
>>
>>Bob D.
>>
>>>
>>>
>>>
>>>>
>>>>Bob D.
>>>>
>>>>>
>>>>>
>>>>>
>>>>>>It is better if you look at the ssdf rating list.
>>>>
>>>>I doubt that the SSDF ratings have anything at all to do with the rating of the
>>>>computers against humans. But the ICC computers ARE playing against humans.
>>>>
>>>>>>
>>>>>>Uri
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.