Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EGTB question

Author: Robert Hyatt

Date: 09:04:51 10/01/01

Go up one level in this thread


On October 01, 2001 at 09:56:06, Mike Hood wrote:

>Eugene, I think I understand your answer in part, but please help me further.
>I've quoted the whole file kbpk.tbs below, and I have three questions:
>
>1) There is a Mate in 31 listed, but no Mate in 30 or Mate in 29. Is this
>because the Mate in 31 involves an immediate transition into another tablebase,
>such as kpk.tbs? Maybe if you could tell me the exact position of this Mate in
>31 it would be obvious to me.

Most likely it is a transposition into a KBN vs K endings.  In a position
where promoting to a queen would be a stalemate.


>
>2) Why are no broken positions listed for btm? I presume that "wka1 wba6 wpa7
>bkb1" is at least one broken position for btm.
>
>3) All the numbers ought to be even, because every position containing one pawn
>can be mirrored left-right. Is the presence of odd numbers because your
>statistics are the number of positions in your already simplified/compressed
>chess board?
>
>Thanks,
>
>Mike
>
>kbpk.tbs
>========
>wtm: Draws:              174804
>wtm: Mate in  31:             1
>wtm: Mate in  28:             1
>wtm: Mate in  26:            11
>wtm: Mate in  25:             4
>wtm: Mate in  24:            25
>wtm: Mate in  23:           167
>wtm: Mate in  22:           476
>wtm: Mate in  21:          1917
>wtm: Mate in  20:          6110
>wtm: Mate in  19:         17489
>wtm: Mate in  18:         36103
>wtm: Mate in  17:         85135
>wtm: Mate in  16:        150419
>wtm: Mate in  15:        186502
>wtm: Mate in  14:        209994
>wtm: Mate in  13:        283840
>wtm: Mate in  12:        406655
>wtm: Mate in  11:        460888
>wtm: Mate in  10:        442601
>wtm: Mate in   9:        440135
>wtm: Mate in   8:        436186
>wtm: Mate in   7:        406369
>wtm: Mate in   6:        303115
>wtm: Mate in   5:        157067
>wtm: Mate in   4:         68023
>wtm: Mate in   3:         29622
>wtm: Mate in   2:          9838
>wtm: Mate in   1:          3118
>wtm: Broken positions:   500513
>btm: Lost in   0:           117
>btm: Lost in   1:           873
>btm: Lost in   2:          2793
>btm: Lost in   3:          8892
>btm: Lost in   4:         29591
>btm: Lost in   5:         84157
>btm: Lost in   6:        226693
>btm: Lost in   7:        415728
>btm: Lost in   8:        455096
>btm: Lost in   9:        447616
>btm: Lost in  10:        430805
>btm: Lost in  11:        422716
>btm: Lost in  12:        394004
>btm: Lost in  13:        338834
>btm: Lost in  14:        282260
>btm: Lost in  15:        243794
>btm: Lost in  16:        209974
>btm: Lost in  17:        137192
>btm: Lost in  18:         70856
>btm: Lost in  19:         33990
>btm: Lost in  20:         13627
>btm: Lost in  21:          7010
>btm: Lost in  22:          4168
>btm: Lost in  23:          2492
>btm: Lost in  24:           944
>btm: Lost in  25:           384
>btm: Lost in  26:           129
>btm: Lost in  27:            48
>btm: Lost in  28:            24
>btm: Lost in  31:             2
>btm: Draws:              859923
>
>
>
>On September 28, 2001 at 13:37:24, Eugene Nalimov wrote:
>
>>DTM, not DTC. After capture or promotion resulting position is stored in the
>>other TB, not in the current one.
>>
>>Eugene
>>
>>On September 28, 2001 at 12:38:55, Mike Hood wrote:
>>
>>>Thanks for the reply, Eugene. I'm willing to accept I may be wrong, but how do
>>>you explain the discrepancies I named?
>>>
>>>Mike
>>>
>>>On September 28, 2001 at 11:11:33, Eugene Nalimov wrote:
>>>
>>>>There is no bug in TBSTAT, and statistics is correct.
>>>>
>>>>Eugene
>>>>
>>>>On September 28, 2001 at 08:27:37, Mike Hood wrote:
>>>>
>>>>>On September 19, 2001 at 04:23:39, Les Fernandez wrote:
>>>>>
>>>>>>Has anyone tabulated the total win-lose-draw-broken positions for first the 3
>>>>>>then 4 and then the 5 piece sets?  I know that the information is available on
>>>>>>Bobs ftp site but hoping that someone has already done it for both WTM and BTM
>>>>>>so I dont have to do them one by one.
>>>>>>
>>>>>>ie:
>>>>>>
>>>>>>EGTB     WINS     LOSE     DRAW     BROKEN
>>>>>>
>>>>>> 3
>>>>>> 4
>>>>>> 5
>>>>>
>>>>>This is theoretically possible, but not very sensible at the moment. Most of the
>>>>>TBS files on Robert Hyatt's ftp server are defective, probably due to a bug in
>>>>>TBSTAT.
>>>>>
>>>>>For instance:
>>>>>
>>>>>krrk.tbs -- On black's move there are 1032 positions that lose in 16, but on
>>>>>white's move the longest mate is Mate in 7.
>>>>>
>>>>>kqrkr.tbs -- On black's move there are 60 moves that are Mate in 5, but there
>>>>>are ZERO moves that are Mate in 4.
>>>>>
>>>>>kbpk.tbs -- On white's move there is 1 position that is Mate in 31, but there
>>>>>are ZERO moves that are Mate in 30.
>>>>>
>>>>>In most of the TBS files the number of broken positions is not listed.
>>>>>
>>>>>Over 50% of the TBS files are obviously defective, and I can't vouch for the
>>>>>validity of the others. And, to ask a naive question, shouldn't the number of
>>>>>Mates/Losses/Draws in a pawnless tablebase always be divisible by 4?
>>>>>
>>>>>I've reported the bug to Eugene by email a few weeks ago, but it hasn't been
>>>>>fixed yet.
>>>>>
>>>>>Mike



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.