ChessGUI 0.245f is available

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

Re: ChessGUI 0.245f is available

Post by Martin Thoresen »

Matthias Gemuh wrote: I don't think you earnestly expected me to reply to any of them. Why should I ?
I had a highlighting colour in your chat in Stage 1 of your Season 2 TCEC.
From Stage 2 onwards, you disqualified me from having a green colour in your chat. Even when I specifically asked for a green colour, you refused giving it, pointing me to a rule you had made that permits only engine authors to have a highlighting colour.
I have so far invested dozens of hours tweaking ChessGUI to meet your TCEC demands. Why should I continue doing the dirty work back-stage when a mere green colour in the chat is too much to honour me with ?
I understand that no other GUI is able to replace ChessGUI in TCEC at the
moment, so use it till alternatives become available.

Matthias.
Are you serious? You don't reply because you didn't get a green color in the chat? Wow. Just wow. Even my sister didn't get a green color.

But thanks for clarifying that you feel you are doing the TCEC dirty work. Makes my choice easier.
User avatar
Graham Banks
Posts: 41415
Joined: Sun Feb 26, 2006 10:52 am
Location: Auckland, NZ

Re: ChessGUI 0.245f is available

Post by Graham Banks »

Matthias Gemuh wrote:
Graham Banks wrote: Matthias - have you managed to implement sudden death games for tied knockout matches?
If you haven't, I'd really appreciate it if you could.

The other feature that would be useful would be the evaluation given in brackets after each move in the displayed pgn.

Regards,
Graham.
see pm.
Thanks Matthias. All good. :)
gbanksnz at gmail.com
Modern Times
Posts: 3546
Joined: Thu Jun 07, 2012 11:02 pm

Re: ChessGUI 0.245f is available

Post by Modern Times »

Matthias Gemuh wrote:I have so far invested dozens of hours tweaking ChessGUI to meet your TCEC demands.
Matthias.
You've done an amazing job as the unsung hero behind making TCEC possible with ChessGUI. I don't think there are any other GUIs at the moment that provide all the relevant functionality.

In terms of CCRL 40/40, I'd guess that your GUI provides about 80% of the games every week on average. Sometimes more.
User avatar
hgm
Posts: 27787
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: ChessGUI 0.245f is available

Post by hgm »

Martin Thoresen wrote:1: Will you look to improve the pairing algorithm for round robins?
Just out of curiosity:

What can be bad with a pairing algorithm for round robins, that it can be improved? I would think they either work, or they don't. What makes in your opinion some pairing schemes better than others?
User avatar
Matthias Gemuh
Posts: 3245
Joined: Thu Mar 09, 2006 9:10 am

Re: ChessGUI 0.245f is available

Post by Matthias Gemuh »

Modern Times wrote:
Matthias Gemuh wrote:I have so far invested dozens of hours tweaking ChessGUI to meet your TCEC demands.
Matthias.
You've done an amazing job as the unsung hero behind making TCEC possible with ChessGUI. I don't think there are any other GUIs at the moment that provide all the relevant functionality.

In terms of CCRL 40/40, I'd guess that your GUI provides about 80% of the games every week on average. Sometimes more.
My focus in future will in making ChessGUI more suitable for CCRL indeed.
My engine was quite strong till I added knowledge to it.
http://www.chess.hylogic.de
User avatar
hgm
Posts: 27787
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: ChessGUI 0.245f is available

Post by hgm »

gladius wrote:SF reports scores in centipawns directly from the evaluation. Houdini actually scales it's internal scores into win probabilities. Both are completely valid IMO.
Then Stockfish is compliant (and it seems strange to divide its score by 1.9, as shown in the starting post of this thread). But Houdini is at fault.

Winning probabilities are NOT centi-Pawns, so I don't see what would make this 'completely valid'. The UCI standard clearly requires centi-Pawns. It even prints 'score cp XXX' to indicate it is centi-Pawns. It is simply lying.

If it uses win probablilities internally, there is no excuse for not converting them back to centi-Pawns in order to print them. It doesn't even make the effort to normalize the winning probabilities such that for small scores they would coincide with centi-Pawns.

The very fact that there now is demand for GUI options to correct this behavior shows how undesirable it is.
User avatar
Matthias Gemuh
Posts: 3245
Joined: Thu Mar 09, 2006 9:10 am

Re: ChessGUI 0.245f is available

Post by Matthias Gemuh »

hgm wrote:
gladius wrote:SF reports scores in centipawns directly from the evaluation. Houdini actually scales it's internal scores into win probabilities. Both are completely valid IMO.
Then Stockfish is compliant (and it seems strange to divide its score by 1.9, as shown in the starting post of this thread). But Houdini is at fault.

Winning probabilities are NOT centi-Pawns, so I don't see what would make this 'completely valid'. The UCI standard clearly requires centi-Pawns. It even prints 'score cp XXX' to indicate it is centi-Pawns. It is simply lying.

If it uses win probablilities internally, there is no excuse for not converting them back to centi-Pawns in order to print them. It doesn't even make the effort to normalize the winning probabilities such that for small scores they would coincide with centi-Pawns.

The very fact that there now is demand for GUI options to correct this behavior shows how undesirable it is.
Among Top-10 engines, SF is the only one whose scores lean outside the window.
For that reason alone, many people resent using SF for analysis.
That "winning probabilities" stuff is irrelevant in this discussion.
My engine was quite strong till I added knowledge to it.
http://www.chess.hylogic.de
Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

Re: ChessGUI 0.245f is available

Post by Martin Thoresen »

hgm wrote: Just out of curiosity:

What can be bad with a pairing algorithm for round robins, that it can be improved? I would think they either work, or they don't. What makes in your opinion some pairing schemes better than others?
It's not so much my opinion but some complaints that appeared in the TCEC chat during Stage 2. The current pairing system for single round robins doesn't work if there is an even number of participants, then one engine will get either all black or all white. So Stage 2 had to be changed to an uneven number of participants (from 18 to 19). Even so, some engines got 4-5 whites in a row, then 7-8 blacks then rest whites and vice versa.

Again, for me it's not a big deal. But a proper pairing algorithm with white, black, white black etc. would be the best.
User avatar
Matthias Gemuh
Posts: 3245
Joined: Thu Mar 09, 2006 9:10 am

Re: ChessGUI 0.245f is available

Post by Matthias Gemuh »

Martin Thoresen wrote:... The current pairing system for single round robins doesn't work if there is an even number of participants, then one engine will get either all black or all white. ...
... and that is indeed a bug that I was not aware of (perhaps because hardly anyone run single-round roundrobins) till it popped up in TCEC.
My engine was quite strong till I added knowledge to it.
http://www.chess.hylogic.de
User avatar
hgm
Posts: 27787
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: ChessGUI 0.245f is available

Post by hgm »

Martin Thoresen wrote:It's not so much my opinion but some complaints that appeared in the TCEC chat during Stage 2. The current pairing system for single round robins doesn't work if there is an even number of participants, then one engine will get either all black or all white. So Stage 2 had to be changed to an uneven number of participants (from 18 to 19). Even so, some engines got 4-5 whites in a row, then 7-8 blacks then rest whites and vice versa.
Ah OK, I see. It is the color assignment that is broken.

Well, for whom it is of any interest, there exists a simple pairing algorithm that works for any number of players (and can even be applied in real OTB tourneys without the aid of any computer):

Put all (N) Chess boards on a long table, alternating their orientation. For an odd number of players (2N-1) put only one chair with board N (which isn't really used in that case; that player has the BYE). Now after every round, all players move one chair to their right, wrapping around the end of the table when they reach it.

This will make everyone play against everyone else, because swinging around the table at board N switches them from the odd to the even group. They will also alternate color on every round (because this is how you layed out the boards), even when they swing around the table at board 1. As with an odd number of players there is an even number of rounds, they will all get exactly as many whites and blacks.

Now for an even number of players you use exactly the same system, except you put a second chair at board N. And player nr 2N will ALWAYS remain at board N, moving to the opposite side of the table on every round (so that he strictly alternates color), and the 'one-chair to the right' rule skips the chair he is on, and is applied after he swaps places with his latest opponent. This means all other players now get a white or black (depending on the round) between the game they play on one side of board N-1 and the other (in which they obviously have opposite colors). So they will get two times the same color in a row. But this is the ONLY case where they get two times the same color; in all other rounds they will strictly alternate color.

No turning of boards is necessary. No player has to move more than two chairs to the right ever. (And only two chairs when he moves directly between board N-1 and board N at the opposite side of the table, which each player does only once, except player 2N, who never does it.) In an OTB computer tourney this means you have to move equipment only once every two rounds!

In an engine tourney, where the games of a round are not played simultaneously, you can always play them in the order board 1 -> board N. This will always cause maximum separation between the games of any player; there will never be two games in a row by the same player, (except with <= 6 players); there will always be N-2 other games between games of the same player.