The on-line engine blitz tourney for December will be on
Saturday December 3, 15pm EST (21:00 CET)
To connect:
winboard -zp -ics -icshost 80.100.28.169 -icshelper timeseal -fcp ENGINE.exe -fd ENGINEFOLDER -autoKibitz
For UCI engines, add -fUCI . Be there ~15 min in advance.
On-line blitz tourney December
Moderators: hgm, Rebel, chrisw
-
- Posts: 27817
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
-
- Posts: 918
- Joined: Mon Jan 05, 2009 7:40 pm
- Location: Germany
- Full name: Engin Üstün
Re: On-line blitz tourney December
good, i will be there
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: On-line blitz tourney December
Hmm... I was thinking it'd be fun to do one of these. But I'm afraid I'll be doing last-minute present shopping.
By the way, do you have a timeframe for the 10x8 tournament? I'd like to do a little more work on Sjaak before then.
By the way, do you have a timeframe for the 10x8 tournament? I'd like to do a little more work on Sjaak before then.
-
- Posts: 27817
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: On-line blitz tourney December
Code: Select all
:Tourney Players: Round 9 of 9
:
: Name Rating Score Perfrm Upset Results
: ----------------- ------ ----- ------ ------ -------
: 1 -RybkaGB [2241] 8.5 [2154] [ 0] +06w +04b +03w +07w +09b +02w +05b +10b =08w
: 2 +CraftyRB [1996] 6.5 [1994] [ 15] +11w =08w -07b +05w +06b -01b +04w +03b +10w
: 3 +Nightmare [2011] 6.0 [1917] [ 0] +09w +05w -01b +06b -04w +10w +07b -02w +12b
: 4 +Rookie [1873] 6.0 [1869] [ 138] +12b -01w +08b +11w +03b -07w -02b +13w +14w
: 5 -Tornado [1833] 6.0 [1844] [ 95] +14w -03b +10w -02b +11b +09w -01w +07b +13b
: 6 +GaviotaRB [1699] 6.0 [1869] [ 344] -01b +07w +12b -03w -02w +08b +13w +14b +11w
: 7 +Goldbar [1928] 5.0 [1878] [ 68] +10w -06b +02w -01b +08w +04b -03w -05w +09b
: 8 -Almere [1814] 4.5 [1716] [ 304] +13w =02b -04w =09b -07b -06w +14b +12w =01b
: 9 +Eichhoernchen [1562] 4.5 [1717] [ 126] -03b +14w +13b =08w -01w -05b +12w +11b -07w
: 10 -EveAnn [1476] 4.0 [1693] [ 3] -07b +12w -05b +14w +13w -03b +11b -01w -02b
: 11 +Myrddin [1479] 2.5 [1443] [ 0] -02b +13w +14b -04b -05w =12b -10w -09w -06b
: 12 +Spartacus [1448] 1.5 [1357] [ 15] -04w -10b -06w =13b =14b =11w -09b -08b -03w
: 13 +microMax [1292] 1.5 [1355] [ 199] -08b -11b -09w =12w -10b +14w -06b -04b -05w
: 14 +blik [1413] 0.5 [1252] [ 17] -05b -09b -11w -10b =12w -13b -08w -06w -04b
:
: Average Rating 1718.9
:
-
- Posts: 27817
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: On-line blitz tourney December
This will not be before January. I will decide on a date after the holidays. The current version of Spartacus also seems very sick; it was at -9 for 50 moves in positions likeEvert wrote:By the way, do you have a timeframe for the 10x8 tournament? I'd like to do a little more work on Sjaak before then.
[d]8/8/8/3k4/8/1b5p/p6B/K7 w
and I don't see how it could possibly get that with a correct search... Even given that it does not know this KBPK is a fortress draw, it should never score it worse than -6 or so (3.25 for the Bishop, and 2.5 for the 7th-rank passer). -9 means it thinks black can force KQK, which is just nonsense.
-
- Posts: 1357
- Joined: Wed Mar 08, 2006 10:15 pm
- Location: San Francisco, California
Re: On-line blitz tourney December
Ah yes, the game against Myrddin. For reference, Myrddin scored the position with pawns farther back as +3.88, and was scoring the position you give as +5.16. It does not have information about fortresses, although it does have 5-man TB support. So as long as it could avoid losing any material....hgm wrote:This will not be before January. I will decide on a date after the holidays. The current version of Spartacus also seems very sick; it was at -9 for 50 moves in positions likeEvert wrote:By the way, do you have a timeframe for the 10x8 tournament? I'd like to do a little more work on Sjaak before then.
[d]8/8/8/3k4/8/1b5p/p6B/K7 w
and I don't see how it could possibly get that with a correct search... Even given that it does not know this KBPK is a fortress draw, it should never score it worse than -6 or so (3.25 for the Bishop, and 2.5 for the 7th-rank passer). -9 means it thinks black can force KQK, which is just nonsense.
jm
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: On-line blitz tourney December
Awesome, that means I can get some tweaking/bugfixing in over Christmas.hgm wrote:This will not be before January. I will decide on a date after the holidays.Evert wrote:By the way, do you have a timeframe for the 10x8 tournament? I'd like to do a little more work on Sjaak before then.
Sjaak evaluates it as about -3, both statically and after a 12 ply search. I think that's just from the two advanced pawns, but I'd have to check. This will certainly be a useful position to test some endgame evaluation terms with, so I'll keep it around.The current version of Spartacus also seems very sick; it was at -9 for 50 moves in positions like
[d]8/8/8/3k4/8/1b5p/p6B/K7 w
and I don't see how it could possibly get that with a correct search... Even given that it does not know this KBPK is a fortress draw, it should never score it worse than -6 or so (3.25 for the Bishop, and 2.5 for the 7th-rank passer). -9 means it thinks black can force KQK, which is just nonsense.
As an aside, I find it interesting that the value for a seventh-rank pawn seems to converge on ~minor for most programs. It's not so surprising, perhaps, but I originally had it much higher in Jazz and after tuning found that it's very similar in other programs. In Sjaak I gave it that value without tuning, but there it's modulated by the value of the strongest promotion piece (so it gets beaten down very heavily in variants like Shatranj where promotion is to ferz).
EDIT
Wow!
Code: Select all
[ 2] -2.97 0.00 101 1 / 0 Bf4
< 3> 0.00 240 2 / 0 Bc7 Ke4
< 3> 0.00 459 2 / 0 Kb2 Ke4
[ 3] -3.06 0.00 503 2 / 0 Kb2 Ke4
< 4> 0.01 903 3 / 0 Bc7 Ke4 Bd6
[ 4] -2.99 0.01 917 3 / 0 Bc7 Ke4 Bd6
[ 5] -3.08 0.01 1548 4 / 0 Bc7 Ke4 Bd6 Bc4
[ 6] -3.00 0.01 2516 5 / 0 Bc7 Kd4 Kb2 Bc4 Bd6
< 7> 0.03 6714 6 / 3 Bg1 Ke4 Kb2 Bd5 Ka1 Kd3
< 7> 0.04 9098 6 / 1 Kb2 Ke4 Bd6 Bd5 Bc7 Kf5
[ 7] -3.09 0.05 9805 6 / 1 Kb2 Ke4 Bd6 Bd5 Bc7 Kf5
[ 8] -3.02 0.06 13396 7 / 0 Kb2 Ke4 Bb8 Bf7 Bd6 Bg8 Be7
[ 9] -3.09 0.09 21287 8 / 1 Kb2 Ke4 Bd6 Bc4 Bb8 Bd5 Bc7 Kd3
[10] -3.02 0.11 28737 9 / 1 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Bc7 Kd4 Ka1 <H>
[11] -3.09 0.20 54107 10 / 1 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Ka1 Kf5 Bd6 Kf6
[12] -3.09 0.26 85996 11 / 2 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Ka1 Kf5 Bd6 Bc4 Be7
<13> --- 0.52 197396 1 / 1 Kb2 Ke4 Ka1 Bf7 Bc7 Kf3 Bb8 Kg2 Kb2 h2 Bxh2 Kxh2
[13] -8.78 0.59 220882 12 / 1 Kb2 Ke4 Ka1 Bf7 Bc7 Kf3 Bb8 Kg2 Kb2 h2 Bxh2 Kxh2
[14] -8.78 0.89 344515 4 / 0 Kb2 Ke4 Ka1 Bf7 <H>
<15> --- 1.28 492751 1 / 0 Kb2 Ke4 Ka1 Bc4 Kb2 Kf3 Be5 Kg2 Ka3 h2 Kb4 h1Q Kxc4 a1R
[15] -11.33 1.33 512524 14 / 2 Kb2 Ke4 Ka1 Bc4 Kb2 Kf3 Be5 Kg2 Ka3 h2 Kb4 h1Q Kxc4 a1R
<16> !!! 3.44 1373124 17 / 0 Kb2 Ke4 Bb8 Kf3 Be5 Bg8 Bc7 Bc4 Bd8 a1Q Kxa1 h2 Be7 h1Q Kb2 Qh6 Kc3
[16] -9.96 3.67 1460669 17 / 0 Kb2 Ke4 Bb8 Kf3 Be5 Bg8 Bc7 Bc4 Bd8 a1Q Kxa1 h2 Be7 h1Q Kb2 Qh6 Kc3
<17> --- 6.92 2835198 1 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Bf7 Kb2 Kg2 Bc7 Bg8 Ka1 Bc4 Bb6 h2 Kb2 h1Q Kc3 a1Q
[17] -16.06 7.01 2870839 18 / 2 Kb2 Ke4 Bd6 Kf3 Ka1 Bf7 Kb2 Kg2 Bc7 Bg8 Ka1 Bc4 Bb6 h2 Kb2 h1Q Kc3 a1Q
<18> !!! 8.91 3622327 17 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Kg2 Bb8 h2 Bxh2 Kxh2 Kb2 Kg3 Ka1 Bg8 Kb2 Bf7 Ka1
[18] -8.94 8.93 3626506 17 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Kg2 Bb8 h2 Bxh2 Kxh2 Kb2 Kg3 Ka1 Bg8 Kb2 Bf7 Ka1
<19> --- 14.58 6045070 1 / 0 Kb2 Ke4 Bd6 Bg8 Ka1 Kf3 <H>
[19] -25.46 16.04 6673809 6 / 0 Kb2 Ke4 Bd6 Bg8 Ka1 Kf3 <H>
[20] -25.46 36.67 15323793 14 / 0 Kb2 Ke4 Bg3 Bg8 Bh2 Kf3 Be5 Kg2 Kc2 h2 Bxh2 Kxh2 Kb2 Be6 <H>
-
- Posts: 2929
- Joined: Sat Jan 22, 2011 12:42 am
- Location: NL
Re: On-line blitz tourney December
Ok, that seems to have been relatively easy, an evaluation bug when dealing with a lone king. The scary thing is that this bug has been there since the very beginning...Evert wrote: Wow!I have work to do...Code: Select all
[ 2] -2.97 0.00 101 1 / 0 Bf4 < 3> 0.00 240 2 / 0 Bc7 Ke4 < 3> 0.00 459 2 / 0 Kb2 Ke4 [ 3] -3.06 0.00 503 2 / 0 Kb2 Ke4 < 4> 0.01 903 3 / 0 Bc7 Ke4 Bd6 [ 4] -2.99 0.01 917 3 / 0 Bc7 Ke4 Bd6 [ 5] -3.08 0.01 1548 4 / 0 Bc7 Ke4 Bd6 Bc4 [ 6] -3.00 0.01 2516 5 / 0 Bc7 Kd4 Kb2 Bc4 Bd6 < 7> 0.03 6714 6 / 3 Bg1 Ke4 Kb2 Bd5 Ka1 Kd3 < 7> 0.04 9098 6 / 1 Kb2 Ke4 Bd6 Bd5 Bc7 Kf5 [ 7] -3.09 0.05 9805 6 / 1 Kb2 Ke4 Bd6 Bd5 Bc7 Kf5 [ 8] -3.02 0.06 13396 7 / 0 Kb2 Ke4 Bb8 Bf7 Bd6 Bg8 Be7 [ 9] -3.09 0.09 21287 8 / 1 Kb2 Ke4 Bd6 Bc4 Bb8 Bd5 Bc7 Kd3 [10] -3.02 0.11 28737 9 / 1 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Bc7 Kd4 Ka1 <H> [11] -3.09 0.20 54107 10 / 1 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Ka1 Kf5 Bd6 Kf6 [12] -3.09 0.26 85996 11 / 2 Kb2 Ke4 Bd6 Bc4 Bg3 Bd5 Ka1 Kf5 Bd6 Bc4 Be7 <13> --- 0.52 197396 1 / 1 Kb2 Ke4 Ka1 Bf7 Bc7 Kf3 Bb8 Kg2 Kb2 h2 Bxh2 Kxh2 [13] -8.78 0.59 220882 12 / 1 Kb2 Ke4 Ka1 Bf7 Bc7 Kf3 Bb8 Kg2 Kb2 h2 Bxh2 Kxh2 [14] -8.78 0.89 344515 4 / 0 Kb2 Ke4 Ka1 Bf7 <H> <15> --- 1.28 492751 1 / 0 Kb2 Ke4 Ka1 Bc4 Kb2 Kf3 Be5 Kg2 Ka3 h2 Kb4 h1Q Kxc4 a1R [15] -11.33 1.33 512524 14 / 2 Kb2 Ke4 Ka1 Bc4 Kb2 Kf3 Be5 Kg2 Ka3 h2 Kb4 h1Q Kxc4 a1R <16> !!! 3.44 1373124 17 / 0 Kb2 Ke4 Bb8 Kf3 Be5 Bg8 Bc7 Bc4 Bd8 a1Q Kxa1 h2 Be7 h1Q Kb2 Qh6 Kc3 [16] -9.96 3.67 1460669 17 / 0 Kb2 Ke4 Bb8 Kf3 Be5 Bg8 Bc7 Bc4 Bd8 a1Q Kxa1 h2 Be7 h1Q Kb2 Qh6 Kc3 <17> --- 6.92 2835198 1 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Bf7 Kb2 Kg2 Bc7 Bg8 Ka1 Bc4 Bb6 h2 Kb2 h1Q Kc3 a1Q [17] -16.06 7.01 2870839 18 / 2 Kb2 Ke4 Bd6 Kf3 Ka1 Bf7 Kb2 Kg2 Bc7 Bg8 Ka1 Bc4 Bb6 h2 Kb2 h1Q Kc3 a1Q <18> !!! 8.91 3622327 17 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Kg2 Bb8 h2 Bxh2 Kxh2 Kb2 Kg3 Ka1 Bg8 Kb2 Bf7 Ka1 [18] -8.94 8.93 3626506 17 / 0 Kb2 Ke4 Bd6 Kf3 Ka1 Kg2 Bb8 h2 Bxh2 Kxh2 Kb2 Kg3 Ka1 Bg8 Kb2 Bf7 Ka1 <19> --- 14.58 6045070 1 / 0 Kb2 Ke4 Bd6 Bg8 Ka1 Kf3 <H> [19] -25.46 16.04 6673809 6 / 0 Kb2 Ke4 Bd6 Bg8 Ka1 Kf3 <H> [20] -25.46 36.67 15323793 14 / 0 Kb2 Ke4 Bg3 Bg8 Bh2 Kf3 Be5 Kg2 Kc2 h2 Bxh2 Kxh2 Kb2 Be6 <H>
That said, there is still something I don't quite understand. Even with that bug, the evaluation should have preferred pawn+bishop over a queen (for white)... so there may be another bug that is now hidden somewhere.
EDIT: Oh... no, not a bug as such. But if the material score is off by a factor of several, all futility and lazy evaluation margins are wrong, which does funky things to the search.
-
- Posts: 27817
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: On-line blitz tourney December
I think that is quite natural: such a passer binds a minor to prevent promotion, making it almost entirely useless for any other purpose. But you want to prevent the side stopping it from sacrificing a minor for the Pawn pre-emptively, because you never know if you will be eventually able to get it cheaper.Evert wrote:As an aside, I find it interesting that the value for a seventh-rank pawn seems to converge on ~minor for most programs.
When you have no minors, opponent passers might actually need to get an extra bonus. Although major pieces that have to stop a passer can still exert more influence over the board than minors do. This assuming the promotion piece will be worth more than a minor. It doesn't seem to really matter how much more. The score in Spartan Chess did not measurably suffer when I restricted the Spartans to promote only to Captain!
-
- Posts: 782
- Joined: Wed Mar 08, 2006 9:22 pm
Re: On-line blitz tourney December
hgm wrote:I think that is quite natural: such a passer binds a minor to prevent promotion, making it almost entirely useless for any other purpose. But you want to prevent the side stopping it from sacrificing a minor for the Pawn pre-emptively, because you never know if you will be eventually able to get it cheaper.Evert wrote:As an aside, I find it interesting that the value for a seventh-rank pawn seems to converge on ~minor for most programs.
When you have no minors, opponent passers might actually need to get an extra bonus. Although major pieces that have to stop a passer can still exert more influence over the board than minors do. This assuming the promotion piece will be worth more than a minor. It doesn't seem to really matter how much more. The score in Spartan Chess did not measurably suffer when I restricted the Spartans to promote only to Captain!
Hello H.G.,
Isn't the bonus because the major piece is tied to menial pawn passer duties instead of assisting in checkmating activities elsewhere thus exerting less - rather than more - influence over the rest of the board?
A passed pawn that ties up a rook or even a queen is doing wonderful things versus tying up a bishop...
Please feel free to enlighten my darkness...
Later.