Crazyhouse tournaments and rating list

Evert · Post by **Evert** » Fri Nov 11, 2016 5:35 pm

Guenther wrote:
Evert wrote:
Guenther wrote: I found a new issue concerning underpromotions in WB. Somehow it seems I
cannot get an underpromotion when I edit or want to play a move that
promotes. WB always assumes a Q promotion in the pgn and on the board.
(promote to Q always is set to false of course)

I changed my font to WB default (= no font), but still I always get a Q?
Normally there should be a dialogue for selecting the piece?

I changed the promotion piece from Q to N in a pgn file I had and the right piece (special promoted N) appeared.
Thus it seems the missing underpromotion only concerns Human players?
N at least should be possible as it is even more needed than in standard chess.
The promotion dialog has been gone a while. You now need to make a sort of gesture with the mouse to under promote: pick up the pawn and drag it backward to select the promotion piece.
Hmmm thanks. Somehow this doesn't work as expected though.
I get an hint that I should move the pawn backwards to underpromote.
The funny thing is each time I get a different piece and seldomly the one I wanted? Can you exactly describe how to do it?

Even when reading the xboard manual I don't understand it?

But when you drag such a
pawn backwards first, its identity will start to cycle through the other avail-
able pieces. This will continue until you start to move it forward; at which
point the identity of the piece will be fixed, so that you can safely put it down
on the promotion square.

You need to pull it backward until it shows the piece you want, then move it to the destination square. I actually find it rather inonvenient in combination with a trackpad.

Guenther · Post by **Guenther** » Fri Nov 11, 2016 5:37 pm

Evert wrote:
Guenther wrote:
Evert wrote:
Guenther wrote: I found a new issue concerning underpromotions in WB. Somehow it seems I
cannot get an underpromotion when I edit or want to play a move that
promotes. WB always assumes a Q promotion in the pgn and on the board.
(promote to Q always is set to false of course)

I changed my font to WB default (= no font), but still I always get a Q?
Normally there should be a dialogue for selecting the piece?

I changed the promotion piece from Q to N in a pgn file I had and the right piece (special promoted N) appeared.
Thus it seems the missing underpromotion only concerns Human players?
N at least should be possible as it is even more needed than in standard chess.
The promotion dialog has been gone a while. You now need to make a sort of gesture with the mouse to under promote: pick up the pawn and drag it backward to select the promotion piece.
Hmmm thanks. Somehow this doesn't work as expected though.
I get an hint that I should move the pawn backwards to underpromote.
The funny thing is each time I get a different piece and seldomly the one I wanted? Can you exactly describe how to do it?

Even when reading the xboard manual I don't understand it?

But when you drag such a
pawn backwards first, its identity will start to cycle through the other avail-
able pieces. This will continue until you start to move it forward; at which
point the identity of the piece will be fixed, so that you can safely put it down
on the promotion square.
You need to pull it backward until it shows the piece you want, then move it to the destination square. I actually find it rather inonvenient in combination with a trackpad.

But it does not change here when pulling it backward only when I promote, thus I don't know what I will get ;-)

hgm · Post by **hgm** » Fri Nov 11, 2016 5:49 pm

That is strange. Are you using click-click moving or drag-drop moving? With drag-drop the piece you are draggig should change Q->N->B->R->Q... approximately every half square you draw it backwards. (Although in Crazyhouse the order could be different.) With click-click you can make a non-stationary second click, pressing the mouse button on the square, and then moving the mouse vertically to see the identity of the promotion piece change in place. (And then release when you see the piece you want.)

You do have 'Animate dragging' switched on? As soon as you pick upa second-rank Pawn it should already morph into a Queen (indicating you will get a Queen if you move it forward).

BTW, the old promotion popup still exists; a persistent option -sweepPromotions true|false decides whether you get the popup or select the piece by mouse movement.

Guenther · Post by **Guenther** » Fri Nov 11, 2016 6:30 pm

hgm wrote:That is strange. Are you using click-click moving or drag-drop moving? With drag-drop the piece you are draggig should change Q->N->B->R->Q... approximately every half square you draw it backwards. (Although in Crazyhouse the order could be different.) With click-click you can make a non-stationary second click, pressing the mouse button on the square, and then moving the mouse vertically to see the identity of the promotion piece change in place. (And then release when you see the piece you want.)

You do have 'Animate dragging' switched on? As soon as you pick upa second-rank Pawn it should already morph into a Queen (indicating you will get a Queen if you move it forward).

BTW, the old promotion popup still exists; a persistent option -sweepPromotions true|false decides whether you get the popup or select the piece by mouse movement.

Thanks HG. It was the dragging animation setting that was off.
I didn't know that it was necessary for this. It works now.

I guess I need to read the full manual now after all the years.

hgm · Post by **hgm** » Sat Nov 12, 2016 12:51 pm

Ferdy wrote:
hgm wrote:I now uploaded a version of CrazyWa (0.1.0) that should be almost free of crashing problems, to the usual place ( http://hgm.nubati.net/dropper.exe ).
This is good no more crashes.

First gauntlet result:
https://sites.google.com/site/zhassocia ... tour-nr-27

Games download, look for CCVA Tour Games\ccva_tour_nr_27.pgn:
https://sites.google.com/site/zhassociation/download

Blitz rating update:
https://sites.google.com/site/zhassocia ... list/blitz

Second gauntlet is in progress:
https://sites.google.com/site/zhassocia ... tour-nr-28

I am disappointed that there seems to be a regression of 100 Elo compared to the (buggy) 0.0.8. My testing method must be totally flawed. It seems I will have to rethink everything.

In the mean time, I applied the fixes that solved the crashing problems to 0.0.8, and replaced the asserts for some rare problematic conditions (such as hash-key collisions or depth overflow) by silent handling. I uploaded this 'non-crashing 0.0.8' to http://hgm.nubati.net/dropper8.exe . Unlike the original 0.0.8 this one also has repetition detection.

One problem is that 0.0.8 contains a known bugs, where it does apply LMR reduction to hash probes that are below the depth where LMR should be supplied. But if I fix that, it seems to play worse. To counter that I can lower the threshold depth for LMR, but now I am not sure about anything anymore. So I just left this bug in.

Guenther · Post by **Guenther** » Sat Nov 12, 2016 2:16 pm

hgm wrote:
Ferdy wrote:
hgm wrote:I now uploaded a version of CrazyWa (0.1.0) that should be almost free of crashing problems, to the usual place ( http://hgm.nubati.net/dropper.exe ).
This is good no more crashes.

First gauntlet result:
https://sites.google.com/site/zhassocia ... tour-nr-27

Games download, look for CCVA Tour Games\ccva_tour_nr_27.pgn:
https://sites.google.com/site/zhassociation/download

Blitz rating update:
https://sites.google.com/site/zhassocia ... list/blitz

Second gauntlet is in progress:
https://sites.google.com/site/zhassocia ... tour-nr-28
I am disappointed that there seems to be a regression of 100 Elo compared to the (buggy) 0.0.8. My testing method must be totally flawed. It seems I will have to rethink everything.

In the mean time, I applied the fixes that solved the crashing problems to 0.0.8, and replaced the asserts for some rare problematic conditions (such as hash-key collisions or depth overflow) by silent handling. I uploaded this 'non-crashing 0.0.8' to http://hgm.nubati.net/dropper8.exe . Unlike the original 0.0.8 this one also has repetition detection.

One problem is that 0.0.8 contains a known bugs, where it does apply LMR reduction to hash probes that are below the depth where LMR should be supplied. But if I fix that, it seems to play worse. To counter that I can lower the threshold depth for LMR, but now I am not sure about anything anymore. So I just left this bug in.

Somehow I could it crash in a Human game. After I lost a few games I finally could win one game with White.
Just when I wanted to make the final mating move it obviously stopped working and WB did not respond anymore.
Note that WB was fully reactive again after closing CrazyWas' process.

I have no debug but I can reproduce it each and every time, when I load
that game nearly to the end and let CrazyWa again play the Black side.

Code: Select all

&#91;Event "RWBC"&#93;
&#91;Site "CAPPUCCINO"&#93;
&#91;Date "2016.11.12"&#93;
&#91;Round "-"&#93;
&#91;White "Guenther Simon"&#93;
&#91;Black "CrazyWa 0.0.8fix"&#93;
&#91;Result "1-0"&#93;
&#91;TimeControl "40/300"&#93;
&#91;Variant "crazyhouse"&#93;
&#91;Annotator "1... +0.01"&#93;

1. d4 e6 &#123;+0.01/11 12&#125; 2. Bf4 Nf6 &#123;+0.30/11 11&#125; 3. Nf3 Nc6 &#123;+0.03/9 5&#125; 4.
e3 Nd5 &#123;+0.03/9 5&#125; 5. Bg3 h5 &#123;-0.13/9 12&#125; 6. h4 Bb4+ &#123;-0.25/8 6&#125; 7. Nbd2
Bxd2+ &#123;-0.10/8 4&#125; 8. Nxd2 N@f5 &#123;+0.05/9 6&#125; 9. B@f4 Nxf4 &#123;+0.13/10 9&#125; 10.
Bxf4 Nxh4 &#123;+0.05/10 8&#125; 11. N@b5 B@a5 &#123;-0.33/9 12&#125; 12. Rh2 g5 &#123;+3.01/9 12&#125;
13. Bg3 a6 &#123;+4.07/8 6&#125; 14. Nxc7+ Bxc7 &#123;+0.44/10 5&#125; 15. Bxc7 Qxc7
&#123;-0.09/9 7&#125; 16. Rxh4 gxh4 &#123;-1.00/9 31&#125; 17. P@g7 R@g8 &#123;-1.01/7 4&#125; 18. N@f6+
Ke7 &#123;-3.86/9 22&#125; 19. gxh8=Q Rxh8 &#123;-3.73/8 6&#125; 20. B@g5 Qa5 &#123;-5.85/7 5&#125; 21.
Qxh5 Qxd2+ &#123;-7.78/6 4&#125; 22. Kxd2 B@b4+ &#123;-15.58/6 18&#125; 23. P@c3 N@g6
&#123;-11.73/5 14&#125; 24. Qxh8 Bxc3+ &#123;-149.95/9 2.7&#125; 25. bxc3 N@b1+ &#123;-149.96/14 3&#125;
26. Rxb1 Nxh8 &#123;-149.97/55 2.7&#125; 27. R@e8+ Kd6 &#123;-149.98/8 0.3&#125; 28. Q@c5+ Kc7
&#123;-149.99/4 0.2&#125; 29. B@d6#
&#123;Xboard adjudication&#58; Checkmate&#125; 1-0

BTW somehow I thought WB could save the thinking time of the Human player, but may be my memory here is wrong?

Screenshot:
You can see I cannot get the move right on the final square, it somehow
hovers above that square and at the top there is a message in German, which means 'no response'.

hgm · Post by **hgm** » Sat Nov 12, 2016 5:45 pm

Guenther wrote:Somehow I could it crash in a Human game. After I lost a few games I finally could win one game with White.
Just when I wanted to make the final mating move it obviously stopped working and WB did not respond anymore.
Note that WB was fully reactive again after closing CrazyWas' process.

I have no debug but I can reproduce it each and every time, when I load
that game nearly to the end and let CrazyWa again play the Black side.

Strange. I tried loading the game upto and including Q@c6+, and then set Machine Black. CrazyWa just moves Kc7, and I can play B@d6 without any problems. The only reason I can imagine why WinBoard would become unresponsive is that CrazyWa is flooding it with input. Because WinBoard adjudicates the mate, it has detected it even before the move goes to the engine. In such a case it should 'put the engine in 'force' mode before sending it the final move.

Do you use 'undo' to reproduce this behavior, or do you just select part pf the PGN text upto the desired move and paste that? Perhaps 'undo' is buggy, didn't really test that. In any case, since it doesn't happen for me, if it reproduces for you it would be useful if you could post the end of a debug log for such an event, to see if CrazyWa somehow goes haywire in response to some of the final commands.

Guenther · Post by **Guenther** » Sat Nov 12, 2016 6:19 pm

hgm wrote:
Guenther wrote:Somehow I could it crash in a Human game. After I lost a few games I finally could win one game with White.
Just when I wanted to make the final mating move it obviously stopped working and WB did not respond anymore.
Note that WB was fully reactive again after closing CrazyWas' process.

I have no debug but I can reproduce it each and every time, when I load
that game nearly to the end and let CrazyWa again play the Black side.
Strange. I tried loading the game upto and including Q@c6+, and then set Machine Black. CrazyWa just moves Kc7, and I can play B@d6 without any problems. The only reason I can imagine why WinBoard would become unresponsive is that CrazyWa is flooding it with input. Because WinBoard adjudicates the mate, it has detected it even before the move goes to the engine. In such a case it should 'put the engine in 'force' mode before sending it the final move.

Do you use 'undo' to reproduce this behavior, or do you just select part pf the PGN text upto the desired move and paste that? Perhaps 'undo' is buggy, didn't really test that. In any case, since it doesn't happen for me, if it reproduces for you it would be useful if you could post the end of a debug log for such an event, to see if CrazyWa somehow goes haywire in response to some of the final commands.

After reading your post I created a debug and from it and the content of
your message I deducted the cause. It seems it is not CrazyWa but WB
which sends a flood of (drop) messages, because I had legality test on!
The move which is shown in the full debug cannot be executed on the board anymore.
Should legality test be set off in crazyhouse always? May be my legal drop count was higher than a certain limit?

Anyhow my thesis could be verified, w/o legality testing there is no problem.
(BTW no undo was used, just went forward in that game and started
engine as Black 2 or 3 moves before the end and made my W moves.)

59760 <first : 0:99:100 iter end, max=99, alpha=-14999 start=-15001 {-14999,-14999}
59760 <first :
59760 <first : store
59760 <first : # 00 23 00 00 00 22 00 00 ff ff ff ff ff fe fe fe ff ff ff ff ff ff
59770 <first : # 20 00 20 3f 00 20 20 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # 00 00 20 00 20 00 00 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # 00 00 00 20 00 00 00 40 ff ff ff fc ff ff ff fe ff ff ff ff ff ff
59770 <first : # 00 00 24 00 00 00 22 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # 40 00 41 5f 40 21 00 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # 00 40 00 40 00 40 00 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # 43 00 42 00 23 00 00 41 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff
59770 <first : # ff ff ff ff ff ff ff ff ff ff ff f0 f0 f0 f0 f0 f0 f0 f0 f0 f0 f0
59770 <first : # ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff dd ff ff ff ff ff ff
59770 <first : 0:99 Hash store 42c69c57d9707ed4, d=99, checker=5a move=d6c7 (45)
59770 <first : # move = 005f7186 f=71,t=86,c=86,fp=5f,tp=5f,mut=5f,vic=00
59770 <first : # evals = -1041,1021
59770 <first : move d6c7
silence
Drop move 19, curr=56, x=11,y=2, p=2
LegalDrop: 2 @ 5,5)
LegalDrop: 2 @ 5,5)
LegalDrop: 48 @ 2,0)
LegalDrop: 48 @ 2,0)

...snipped around 280! lines with more legal drops...

LegalDrop: 48 @ 8,7)
LegalDrop: 44 @ 8,7)
64632 >first : usermove 64632 >first : B@d6
GameEnds(26, Xboard adjudication: Checkmate, 4)
64642 >first : result 1-0 {Xboard adjudication: Checkmate}
64742 >first : quit

hgm · Post by **hgm** » Sat Nov 12, 2016 7:31 pm

I guess WinBoard is doing the mate test in a bit of a stupid way. Originally it tested how many legal board moves there were, and assumed (stale or check)mate if there were none. In games with drops I added an extra test in this case, where the side to move also tries all possible drop moves, and tests if they are legal. (I.e. it performs them, and then generates all moves for the other side to see if one of them captures the King.) Each such a legality test on a drop prints two of these lines (which I apparently added to debug a problem there, but which now are useless and could be deleted). Having 140 possible drops sounds a bit too much in this position (with only a P and a Q to drop), but maybe it performs the mate test twice.

I don't think this can be a problem, though. 140 mate tests should still be done in microseconds;it is equivalent to a not-too-cleverly-written engine doing a search of 140 nodes. And the debug lines are only printed when -debug is on.

The time stamps in the log shows that the move is sent to the engine about 5 sec after the engine made its last move. This sounds reasonable for you having to enter the move by hand. There is no indication in the log of anythig suspect.

Guenther · Post by **Guenther** » Sat Nov 12, 2016 7:58 pm

hgm wrote:I guess WinBoard is doing the mate test in a bit of a stupid way. Originally it tested how many legal board moves there were, and assumed (stale or check)mate if there were none. In games with drops I added an extra test in this case, where the side to move also tries all possible drop moves, and tests if they are legal. (I.e. it performs them, and then generates all moves for the other side to see if one of them captures the King.) Each such a legality test on a drop prints two of these lines (which I apparently added to debug a problem there, but which now are useless and could be deleted). Having 140 possible drops sounds a bit too much in this position (with only a P and a Q to drop), but maybe it performs the mate test twice.

I don't think this can be a problem, though. 140 mate tests should still be done in microseconds;it is equivalent to a not-too-cleverly-written engine doing a search of 140 nodes. And the debug lines are only printed when -debug is on.

The time stamps in the log shows that the move is sent to the engine about 5 sec after the engine made its last move. This sounds reasonable for you having to enter the move by hand. There is no indication in the log of anythig suspect.

The problem is the move never really appears at the board and also doesn't get it to the pgn window, also the result.
The screenshot shows that the UI is not updated and hangs despite what the debug shows. (the screenshot is done minutes after the problem)

Crazyhouse tournaments and rating list

Re: Crazyhouse and underpromotion in WB?

Re: Crazyhouse and underpromotion in WB?

Re: Crazyhouse and underpromotion in WB?

Re: Crazyhouse and underpromotion in WB?

Re: CrazyWa 0.1.0 Blitz tour

Re: CrazyWa 0.1.0 Blitz tour

Re: CrazyWa 0.1.0 Blitz tour

Re: CrazyWa 0.1.0 Blitz tour

Re: CrazyWa 0.1.0 Blitz tour

Re: CrazyWa 0.1.0 Blitz tour