Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Interesting result from SSDF

Author: Amir Ban

Date: 14:07:14 06/10/00

Go up one level in this thread


On June 10, 2000 at 08:28:07, Robert Hyatt wrote:

>On June 10, 2000 at 04:03:07, Amir Ban wrote:
>
>>On June 09, 2000 at 23:17:44, Robert Hyatt wrote:
>>
>>>On June 09, 2000 at 15:18:01, Amir Ban wrote:
>>>
>>>>On June 09, 2000 at 10:33:33, Robert Hyatt wrote:
>>>>
>>>>>On June 09, 2000 at 06:26:17, Amir Ban wrote:
>>>>>
>>>>>>On June 08, 2000 at 16:10:16, blass uri wrote:
>>>>>>
>>>>>>>On June 08, 2000 at 15:25:22, Amir Ban wrote:
>>>>>>>
>>>>>>>>Ok, here's my two cents:
>>>>>>>>
>>>>>>>>The most likely estimate for the game outcome is the overall match result
>>>>>>>>(33.5/41). The probability for any *particular* sequence of 10 games (say the
>>>>>>>>last 10) to be 10-0 is (33.5/41)^10 = 13.3%.
>>>>>>>>
>>>>>>>>If you want probability for any sequence anywhere, the probability is larger, as
>>>>>>>>you point out.
>>>>>>>>
>>>>>>>>If you suspect the 10-0 result, you can argue that the most likely estimate
>>>>>>>>should discount this result, so this gives you a probability of (23.5/31)^10 =
>>>>>>>>6.3%.
>>>>>>>
>>>>>>>If you use the most likely estimate
>>>>>>>The probability for 10:0 is smaller than (23.5/31)^10 because there are draws.
>>>>>>>
>>>>>>>The result was +20, =7.-4 so if you use the most likely estimate you get
>>>>>>>probability (20/31) for Junior to win
>>>>>>>7/31 for a draw and 4/31 for a loss
>>>>>>>
>>>>>>>(20/31)^10=0.012493... so the probability is less than 2%
>>>>>>>
>>>>>>>I ignored the fact that the white probability is different from the black
>>>>>>>probability but not ignoring it does not change much the result because you get:
>>>>>>>
>>>>>>>(10/15)^5*(10/16)^5 that is 0.012558...
>>>>>>>
>>>>>>>If you do not ignore the last 10 games and ignore the colours you get
>>>>>>>(30/41)^10=0.043992...
>>>>>>>
>>>>>>>Uri
>>>>>>
>>>>>>All right, you are right about the draws.
>>>>>>
>>>>>>My presentation of the problem was more complicated than necessary. Here is the
>>>>>>question in its simplest form:
>>>>>>
>>>>>>A match of 41 games concluded with 30 wins, 7 draws and 4 losses. It is unknown
>>>>>>in which order these wins/draws/losses occurred.
>>>>>>
>>>>>>Out of all possible permutations with the same result, what percent of the
>>>>>>permutations have a sequence of at least 10 wins ?
>>>>>>
>>>>>>This is clear cut question for the combinatorics gurus in this newsgroup.
>>>>>>
>>>>>>There seems to be gross difference of opinion about the result of this, so until
>>>>>>we get an answer, we can have an opinion poll about the result.
>>>>>>
>>>>>>My guess: 40%.
>>>>>>
>>>>>>Amir
>>>>>
>>>>>
>>>>>As I mentioned before, I am simply worried that reusing the engine over and
>>>>>over might be the problem.  IE if you just crank up two copies of crafty and
>>>>>let 'em play, over and over, if one breaks, it will remain broken and will
>>>>>likely lose every game (or nearly every game) from that point forward.
>>>>>
>>>>>Since I don't test like this, it is very possible...
>>>>>
>>>>>And that would explain a sudden string of 0's at the end...  and it would
>>>>>explain things if the 0's stop after a restart or another tester plays games,
>>>>>since he would start a new crafty.
>>>>
>>>>Of course, that is a possible explanation for this otherwise incredible result.
>>>>
>>>>Another possible explanation is that there is a bug in Junior that something
>>>>doesn't reset correctly and the engine starts playing stronger than intended,
>>>>and wins every game. I definitely need to look into that.
>>>>
>>>>Amir
>>>
>>>
>>>Perhaps while you are looking, you can look into losing the sarcasm?
>>>
>>>eh?
>>>
>>>I don't ever recall you going 10-0 vs crafty on ICC.
>>
>>It's not easy to do this while being censored. Not to mention being noplayed
>>after 4 consecutive games.
>>
>>Amir
>>
>
>
>You know very well why you were censored.  You made the direct claim, here in
>CCC, that I specifically wrote crafty so that if it was getting mated, it would
>sit and run out of time rather than letting the opponent mate me.  You knew that
>was false.  I (and others) pointed out lots of games on ICC that ended in crafty
>getting mated without it flagging, but you _never_ retracted the ridiculous
>claim.
>
>If you want to try a match to see if you can go 10-0, just tell me when and
>where to show up.  I'll be there.
>

I think this match already took place at SSDF.

This is totally nuts besides. How can you refuse to play me and challenge me at
the same time ?

Amir



>And as I recall, you were automatically +noplayed for lots of disconnects,
>not for playing 4 games in a row...  This is all old news...
>
>
>
>
>>
>>> So _something_ is
>>>definitely "up".



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.