Crazyhouse tournaments and rating list

[hgm]

Actually I told that wrong; I see now that I was limiting QS depth to10 ply, not 6. I guess it also depends on how sever the problem of search explosion is.

Very deep QS lines tend to be bogus ayway, in drop games. They accumulate pieces in hand to far beyond the point where you can mate through drops, if you don't test for such mates.

There is a trade-off here; the fewer plies you allow, the more potentially interesting moves you can afford to search. It could be better to allow checks in the first 6 ply of QS, and only static evaluation beyond than to allow them only in the first 2 ply, and captures to arbitrary depth.

[Evert]

Actually I told that wrong; I see now that I was limiting QS depth to10 ply, not 6. I guess it also depends on how sever the problem of search explosion is.

Ah, ok, that sounds more reasonable. I have a hard limit of 20 ply for quiescence, which I think I once picked as being "more than enough" for regular chess. I never "tuned" this.

Very deep QS lines tend to be bogus ayway, in drop games. They accumulate pieces in hand to far beyond the point where you can mate through drops, if you don't test for such mates.

Positions with drops are not really quiet, but I don't think there's an easy way to keep track of that.
Some patterns could probably be detected through evaluation and scored, but I don't think that'll help with taking the stand-pat score rather than continuing with pointless captures.
Perhaps a tempo-bonus that depends on the amount of material in-hand makes sense though?

[hgm]

This is effectivelly what I have, because I calculate the King threat only one-sided. The side to move automatically knows which squares next to the enemy King it attacks, as a side effect of move generation. So I only calculate the term for that side, and it is always positive. That puts an enormous value on initiative when both Kings are under heavy attack.

[TonyJH]

Hi Ferdinand,

I just saw a post on the lichess forum that someone is running a human vs Imortal 2.0 match on Sunday. I am not sure if he's aware that there is an improved Imortal version (2.9001), or if you'd be willing to share it for this match?

https://en.lichess.org/forum/team-crazy ... vs-imortal

[Ferdy]

Hi Ferdinand,

I just saw a post on the lichess forum that someone is running a human vs Imortal 2.0 match on Sunday. I am not sure if he's aware that there is an improved Imortal version (2.9001), or if you'd be willing to share it for this match?

https://en.lichess.org/forum/team-crazy ... vs-imortal

I think this is just a test of how a human would fare against a medium strength zh engine.

[hgm]

Average depth on this blitz (TC 3'+2") tour nr 21, when scores are within the window of [-10.0, +10.0] pawn unit.

Code: Select all

 nr                      player  aveDepth  games
  1           Sunsetter 9 32bit     12.92     24
  2     NebiyuAlien 1.45a 32bit     11.40     24
  3           TJchess 1.3 64bit      9.60     24
  4        Imortal 2.9001 32bit      9.35     24
  5         CrazyWa 0.0.4 32bit      9.18    216
  6       Sjaak II 1.3.1a 64bit      8.85     24
  7   KKFChess 2.6.7 beta 32bit      8.20     24
  8            Sjeng 11.2 32bit      7.70     24
  9           TSCP zh 1.1 32bit      7.42     24
 10        Pulsar 2009 9b 32bit      7.07     24

BTW, what's with the 'a' in Nebiyu 1.45a? I have a Nebiyu 1.45, and this is also what I find on Daniel's website. I also have two different 1.46, but I suppose these are private versions I packaged for the UEC Cup. I am not sure what they are; they might be specially optimized for mini-Shogi.

[Ferdy]

Average depth on this blitz (TC 3'+2") tour nr 21, when scores are within the window of [-10.0, +10.0] pawn unit.

Code: Select all

 nr                      player  aveDepth  games
  1           Sunsetter 9 32bit     12.92     24
  2     NebiyuAlien 1.45a 32bit     11.40     24
  3           TJchess 1.3 64bit      9.60     24
  4        Imortal 2.9001 32bit      9.35     24
  5         CrazyWa 0.0.4 32bit      9.18    216
  6       Sjaak II 1.3.1a 64bit      8.85     24
  7   KKFChess 2.6.7 beta 32bit      8.20     24
  8            Sjeng 11.2 32bit      7.70     24
  9           TSCP zh 1.1 32bit      7.42     24
 10        Pulsar 2009 9b 32bit      7.07     24

BTW, what's with the 'a' in Nebiyu 1.45a? I have a Nebiyu 1.45, and this is also what I find on Daniel's website. I also have two different 1.46, but I suppose these are private versions I packaged for the UEC Cup. I am not sure what they are; they might be specially optimized for mini-Shogi.

1.45a is using inbetween, converting its promotions so that the GUI understands it and converting the GUI moves to something that nebiyu understands too.

This is the inbetween ini file.

Code: Select all

[InBetween]
CommandLine :=NebiyuAlien.exe
;Priority := low
;Debug := 3
;Log := logfile.log

[Client2Server]
;GUI to Engine
;white q, r, n promotion
8q := 8U
8n := 8G
8r := 8O
;white bishop promotion
a7a8b := a7a8I
a7b8b := a7a8I
b7a8b := b7a8I
b7b8b := b7b8I
b7c8b := b7c8I
c7b8b := c7b8I
c7c8b := c7c8I
c7d8b := c7d8I
d7c8b := d7c8I
d7d8b := d7d8I
d7e8b := d7e8I
e7d8b := e7d8I
e7e8b := e7e8I
e7f8b := e7f8I
f7e8b := f7e8I
f7f8b := f7f8I
f7g8b := f7g8I
g7f8b := g7f8I
g7g8b := g7g8I
g7h8b := g7h8I
h7g8b := h7g8I
h7h8b := h7h8I
;black q, n and r promotion
1q := 1u
1n := 1g
1r := 1o
;black bishop promotion
a2a1b := a2a1i
a2b1b := a2b1i
b2a1b := b2a1i
b2b1b := b2b1i
b2c1b := b2c1i
c2b1b := c2b1i
c2c1b := c2c1i
c2d1b := c2d1i
d2c1b := d2c1i
d2d1b := d2d1i
d2e1b := d2e1i
e2d1b := e2d1i
e2e1b := e2e1i
e2f1b := e2f1i
f2e1b := f2e1i
f2f1b := f2f1i
f2g1b := f2g1i
g2f1b := g2f1i
g2g1b := g2g1i
g2h1b := g2h1i
h2g1b := h2g1i
h2h1b := h2h1i

[Server2Client]
;Engine to GUI
;drop moves
p@ := P@
n@ := N@
b@ := B@
r@ := R@
q@ := Q@
;black queen promotion
1u := 1q
;black knight promotion
a2a1g := a2a1n
a2b1g := a2b1n
b2a1g := b2a1n
b2b1g := b2b1n
b2c1g := b2c1n
c2b1g := c2b1n
c2c1g := c2c1n
c2d1g := c2d1n
d2c1g := d2c1n
d2d1g := d2d1n
d2e1g := d2e1n
e2d1g := e2d1n
e2e1g := e2e1n
e2f1g := e2f1n
f2e1g := f2e1n
f2f1g := f2f1n
f2g1g := f2g1n
g2f1g := g2f1n
g2g1g := g2g1n
g2h1g := g2h1n
h2g1g := h2g1n
h2h1g := h2h1n
;black rook promotion
1o := 1r
;black bishop promotion
1i := 1b
;white queen promotion
8U := 8q
;white knight promotion
8G := 8n
;white rook promotion
8O := 8r
;white bishop promotion
8I := 8b

[Server2Server]


[Client2Client]

You can download the whole pack here, look for nebiyualien.
https://sites.google.com/site/zhassociation/download

I am not familiar with 1.46.

[hgm]

I uploaded a new version of CrazyWa (0.0.7). I my tests it consistently beats Sjeng by ~58%, but I guess this is mainly due to the defective time management of the latter. (It hardly ever uses more than 25 sec for a game, often as little as 10, against CrazyWa's 1.5 min.)

Thanks for the InBetween file. I was already wondering why Nebiyu lost so many games.

[Ferdy]

I uploaded a new version of CrazyWa (0.0.7). I my tests it consistently beats Sjeng by ~58%, but I guess this is mainly due to the defective time management of the latter. (It hardly ever uses more than 25 sec for a game, often as little as 10, against CrazyWa's 1.5 min.)

Thanks for the InBetween file. I was already wondering why Nebiyu lost so many games.

Thanks for the update.
What TC did you use vs Sjeng?

[hgm]

40 moves/min.