Sign in with
Sign up | Sign in
Your question

endgame TB discrepancy

Last response: in Video Games
Share
June 25, 2005 4:24:32 PM

Archived from groups: rec.games.chess.computer (More info?)

In Crafty 19.19, I entered the following EPD that I found on this
newsgroup:
8/6R1/8/8/1b3p2/5p2/7K/3k4 w - -; bm Rf7; id "ECE3.0766"

and it gave me Rg8 right away instead of Rf7 until I entered 'go' when
it returned with Rg5 instantly. I'm a little confused by this Crafty
output but I also wanted to know whether Rf7 is the authoritative
answer. I do, after all, have the KBPPKR TBs installed. Thanks.

White(1): 8/6R1/8/8/1b3p2/5p2/7K/3k4 w - -
1. Rg8 Bd2! 2. Re8! Be3! 3. Rb8 Ke2! 4. Rb1! Bd2 5. Rh1! f2! 6. Kg2!
Be1 7. Kh3! Bd2 8. Kg2! Be1
White(1): go
clearing hash tables
time surplus 0.00 time limit 30.00 (3:30)
depth time score variation (1)
time=0.00 cpu=100% mat=0 n=125 fh=100% nps=10K
ext-> chk=8 cap=0 pp=0 1rep=0 mate=0
predicted=0 nodes=125 evals=26 50move=0
endgame tablebase-> probes=17 hits=17
SMP-> split=0 stop=0 data=0/64 cpu=0.00 elap=0.00

mated in 20 moves.
White(1): Rg5
time used: 0.00
learning position, wtm=1 value=-32727
LearnBook() executed
puzzling over a move to ponder.
clearing hash tables
depth time score variation (1)
Black(1): Bd2 [pondering]
clearing hash tables
time surplus 30.00 time limit 30.50 (3:33)
depth time score variation (1)
time=0.00 cpu=100% mat=0 n=114 fh=100% nps=10K
ext-> chk=4 cap=0 pp=0 1rep=0 mate=0
predicted=0 nodes=114 evals=17 50move=1
endgame tablebase-> probes=17 hits=17
SMP-> split=0 stop=0 data=0/64 cpu=0.00 elap=0.00

More about : endgame discrepancy

Anonymous
June 26, 2005 11:28:30 AM

Archived from groups: rec.games.chess.computer (More info?)

marko wrote:

> In Crafty 19.19, I entered the following EPD that I found on this
> newsgroup:
> 8/6R1/8/8/1b3p2/5p2/7K/3k4 w - -; bm Rf7; id "ECE3.0766"
>
> and it gave me Rg8 right away instead of Rf7 until I entered 'go' when
> it returned with Rg5 instantly. I'm a little confused by this Crafty
> output but I also wanted to know whether Rf7 is the authoritative
> answer. I do, after all, have the KBPPKR TBs installed. Thanks.

'Authoritative answer' -- what's that? A decent engine would go
for Re7 as the best -- there are more important things to spend
cycles on than a lost game ...

I try the position in Fritz (using Fritz 8 as well as the Crafty 19.19
engines), and I get:

----------------------------------
New game,
8/6R1/8/8/1b3p2/5p2/7K/3k4 w - - 0 1

Analysis by Fritz 8:

1. -+ (-#20): 1.Rg6
2. -+ (-#20): 1.Rg5
3. -+ (-#20): 1.Rg8
4. -+ (-#17): 1.Rf7
5. -+ (-#17): 1.Rd7+
6. -+ (-#17): 1.Rc7
7. -+ (-#17): 1.Ra7
8. -+ (-#16): 1.Rb7
9. -+ (-#16): 1.Rg1+
10. -+ (-#16): 1.Rh7
11. -+ (-#15): 1.Rg4
12. -+ (-#13): 1.Kh3
13. -+ (-#12): 1.Kg1
14. -+ (-#12): 1.Kh1
15. -+ (-#11): 1.Rg2
16. -+ (-#10): 1.Rg3
17. -+ (-#8): 1.Re7
---------------------------------------

Looks like an EGTB response -- I suppose I must have
the appropriate endgame file around, though I wouldn't have
guessed it.

Not much choice between Rg5 and Rg8 ... Rf7 loses a bit faster, but
....

Your logs suggests that Crafty doesn't find the relevant file:
have you checked your path information?


--
Anders Thulin ath*algonet.se http://www.algonet.se/~ath
June 26, 2005 12:48:31 PM

Archived from groups: rec.games.chess.computer (More info?)

Maybe 'authoritative answer' was the wrong phrase. The terminology is
'best move' of course. I guess I was wondering what the consensus
was, considering there appears to be some contention.

I was pretty sure Crafty was finding the correct EMD files because it
gave both answers *instantly* (1st after entering the position and the
2nd, after entering 'go'. Plus, the 'probes=17 hits=17' log entry
suggests it found what EMDs it was looking for. Most of all, Crafty
had no qualms consuming 220MB of my RAM after entering the tbpath
directive for my library of ~200GB of EMDs.

I don't know how to interpret your Fritz 8 output since I don't own it,
but I suppose its giving Re7.
Anonymous
June 27, 2005 12:50:02 AM

Archived from groups: rec.games.chess.computer (More info?)

marko wrote:

> I don't know how to interpret your Fritz 8 output since I don't own it,
> but I suppose its giving Re7.

Sorry -- I should have explained that:

>1. -+ (-#20): 1.Rg6

If white moves Rg6, Black will mate in 20 moves.

>17. -+ (-#8): 1.Re7

If white moves Re7, Black will mate in 8 moves.

That is, white can't win unless perhaps Black makes some major
blunder. So any decent engine would commit suicide by 1. Re7 Bxe7 etc.
rather than prolong the agony for another 12 moves :-)

The 'correct' move you want is probably one of the top three, though:

>1. -+ (-#20): 1.Rg6
>2. -+ (-#20): 1.Rg5
>3. -+ (-#20): 1.Rg8

--
Anders Thulin ath*algonet.se http://www.algonet.se/~ath
!