Cute Chess 0.9.4 released

[pferd]

Cutechess works great but I would like to propose the following changes

I would like to use a time control like 5minutes/game + 3 sec/move. Right now I use 240/300+3 to work around it. But I don't know if both timecontrols are the same to engines.


It would be useful if you could check if all engines are responsive at the start of the tournament. Several tournaments were aborted midway through because a single engine was missing.

Kind regards

[Guenther]

Cutechess works great but I would like to propose the following changes

I would like to use a time control like 5minutes/game + 3 sec/move. Right now I use 240/300+3 to work around it. But I don't know if both timecontrols are the same to engines.


It would be useful if you could check if all engines are responsive at the start of the tournament. Several tournaments were aborted midway through because a single engine was missing.

Kind regards

If you want to use e.g. 5min + 3 sec you have to set move number to zero!
Then it will be changed to 'whole game' and if you leave the tc setting it will be shown as normal 5+3 inc game.

[Pai]

I have installed Cute Chess 0.9.4 from the link to my Ubuntu 15.04 box. It worked great and I enjoyed it a lot. However, I tried an un-installation and re-installation, then it stopped working. It always gives me this error message:

cutechess: error while loading shared libraries: libpcre16.so.0: cannot open shared object file: No such file or directory

It worked great before, so the program must be able to run on my box with right installation. I must have removed something while doing un-installation. I wonder if someone knows Linux much better can give me a hint on how to fix it. I was able to install and run old version with:

sudo apt-get install cutechess

Hope to solve this and enjoy new Cute Chess again.

[Werewolf]

However, I also have come across a bug. Using exactly the same settings I changed the time control to 10 seconds per move (previously it was on blitz) and started a new tournament.

There were instances of both engines having their clocks run down at once, the board would suddenly switch, darting between one game and the next and so on. Very strange. I lowered concurrency to 12 but it did not help. Is "time per move" section working properly?

I just tried it and there are a lot of time forfeits, probably because some engines do not provide time margin for this type of TC. To minimize time forfeit, add time margin of 100 ms. If there are still forfeit, increase it to say 200ms.

Again - thank you. Very helpful.

[hgm]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

[Werewolf]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

What if the time per move was really big, like 5 mins per move, and the delay was 200 ms etc?

[Guenther]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

What if the time per move was really big, like 5 mins per move, and the delay was 200 ms etc?

I wonder why using such a time control at all? I can only imagine fixed time per _each single_ move for epd tests, or auto analysis. Here overstepping time is of no importance.

[Werewolf]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

What if the time per move was really big, like 5 mins per move, and the delay was 200 ms etc?

I wonder why using such a time control at all? I can only imagine fixed time per _each single_ move for epd tests, or auto analysis.

Yes, auto analysis in IDeA. I want to know which is the best IDeA engine.

[Ferdy]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

What if the time per move was really big, like 5 mins per move, and the delay was 200 ms etc?

I wonder why using such a time control at all? I can only imagine fixed time per _each single_ move for epd tests, or auto analysis.

Yes, auto analysis in IDeA. I want to know which is the best IDeA engine.

For IDEA, it does not use,

Code: Select all

go movetime <timems>

instead it uses,

Code: Select all

go infinite

and later

Code: Select all

stop

Also it limits to search certain candidate moves only by,

Code: Select all

go infinite searchmoves <move1> <move2> ... <moven>

then,

Code: Select all

stop

[Werewolf]

Beware that this seems a dubious practice. The engines will still overrun time, but you just instruct the GUI to ignore it. That the engines overrun their time is a strong indication they no longer reliably get CPU service, but are often suspended for stretches as time nearly as long as the margin you need. As this will randomly strike some engines and not their opponents, you cannot expect your test results to mean anything under those conditions. Options like this should only be used if you don't care about the results...

What if the time per move was really big, like 5 mins per move, and the delay was 200 ms etc?

I wonder why using such a time control at all? I can only imagine fixed time per _each single_ move for epd tests, or auto analysis.

Yes, auto analysis in IDeA. I want to know which is the best IDeA engine.

For IDEA, it does not use,

Code: Select all

go movetime <timems>

instead it uses,

Code: Select all

go infinite

and later

Code: Select all

stop

Also it limits to search certain candidate moves only by,

Code: Select all

go infinite searchmoves <move1> <move2> ... <moven>

then,

Code: Select all

stop

[/b]

I'm not sure of the relevance of the bit in bold above. What's your point?

But if I'm trying to establish which engine is best at, say, 5 mins per move, isn't this the only option I've got in cutechess?

I don't really see how 5 mins per move in cutechess is different to 5 mins per move in IDeA I suppose.