5th Amateur Championship Division 2 (completed)

[Graham Banks]

By the way,Is this a prediction contest and do you count points?It would be fun

It takes longer for my tourneys to finish than yours Swami, so I'm picking that many would lose interest.
However, I could consider it for the next series if enough members express interest.

Regards, Graham.

[Uri Blass]

I plan to release version418 today or tomorrow(

Thanks,Uri. Movei needs to the updated because it has to catch up with lot of engines in Division 1 and 2, Movei will do well.It's just that the current version doesn't play at its level in blitz 1/1.

Regards.

I think that the main advantage of the new version is at long time control thanks to pruning based on hash tables.

possible problem of movei at 1+1 is that it prints too much information and it may cause significant speed reduction under an interface that is not winboard.

Movei prints pv in the following cases:
1)when it find new score for the pv move
2)when it fails high(there may be wrong fail high that after research are rejected)
3)when it fails low
4)at the end of every iteration.

For comparison other engines do not print pv at the end of the iteration and do not print fail high that may be wrong.

I do not know if it is the reason for the bad result that I saw at 1+1 and there may be a different reason(note that 438 also showed bad results at 1+1 in jorge testing).

That problem was not corrected in 438 but was corrected in a future versions when there is a noise parameter that tell movei not to print pv unless it searched more than noise nodes so by telling movei that noise=1000000 you can tell it not to print the pv in 1+1 blitz conditions.

I think that printing pv is not an important problem in 40/4 ccrl or 40/40 ccrl conditions and my guess is that the demage at 40/4 is less than 10 elo but it can be more important problem in faster time control.

Uri

[Kirill Kryukov]

I considered again releasing new version of movei and I plan to release version418 today or tomorrow(I already have slightly newer version but I did not test it enough)

If movei418 is going to play it may be favourite to win this tournament because my estimate is that version 418 is at least 100 elo stronger than version 403 at 40/40(my estimate may be wrong and 57-43 against fruit2.0 at 40/30 on PIV3.01 ghz prove nothing).

Great!

Movei is currently #29 in my Free Single-CPU Blitz selection. I will do my usual amount of testing (1000+ games) for new Movei and I am very curious how high it will rank. Thanks for reconsidering about releasing a free version!

[swami]

By the way,Is this a prediction contest and do you count points?It would be fun

It takes longer for my tourneys to finish than yours Swami, so I'm picking that many would lose interest.
However, I could consider it for the next series if enough members express interest.

Regards, Graham.

I see that you do have a point, I'm sure more members would be interested since it is coming from CCRL!

Someday I may see CCRL organizing prediction contest in atleast blitz 40/4.

[Rubinus]

Movei, Pharaon, Jonny.

[swami]

I plan to release version418 today or tomorrow(

Thanks,Uri. Movei needs to the updated because it has to catch up with lot of engines in Division 1 and 2, Movei will do well.It's just that the current version doesn't play at its level in blitz 1/1.

Regards.

I think that the main advantage of the new version is at long time control thanks to pruning based on hash tables.

possible problem of movei at 1+1 is that it prints too much information and it may cause significant speed reduction under an interface that is not winboard.

Movei prints pv in the following cases:
1)when it find new score for the pv move
2)when it fails high(there may be wrong fail high that after research are rejected)
3)when it fails low
4)at the end of every iteration.

For comparison other engines do not print pv at the end of the iteration and do not print fail high that may be wrong.

I do not know if it is the reason for the bad result that I saw at 1+1 and there may be a different reason(note that 438 also showed bad results at 1+1 in jorge testing).

That problem was not corrected in 438 but was corrected in a future versions when there is a noise parameter that tell movei not to print pv unless it searched more than noise nodes so by telling movei that noise=1000000 you can tell it not to print the pv in 1+1 blitz conditions.

I think that printing pv is not an important problem in 40/4 ccrl or 40/40 ccrl conditions and my guess is that the demage at 40/4 is less than 10 elo but it can be more important problem in faster time control.

Uri

I see. Uri,Thanks for all these information, If you have any Movei version that does well at 1/1 even if it is older then you may considering sending it to me if you wish.

I don't know about most of the programming stuffs you wrote but I learn something.

Let it do well in longer time controls in CCRL, Fast tournament is just for fun and not serious.

[Uri Blass]

I am going to test before sending it to you to be sure that I have not a new bug in the last version.

438 was tested enough by me when the last version was not tested much.

In my tests under fritz interface 438 performed well enough but the same was for 403 and I can imagine that with different computer or different interface things are different.

Uri

[Uri Blass]

I considered again releasing new version of movei and I plan to release version418 today or tomorrow(I already have slightly newer version but I did not test it enough)

You'll need to be quick to get it into this tournament Uri, because once it starts there will be no replacements.

Regards, Graham.

I already sent the new version to Leo with a small readme file
I hope that he is going to release it today.
Note that this is the same version that was tested in the FRC list by Ray.

Uri

[Uri Blass]

I also sent this version to Daniel Charles

http://www.superchessengine.com/

I hope that he will also put the engine at his site for download because I understand that he gives free space for chess engines.

Uri

[swami]

I also sent this version to Daniel Charles

http://www.superchessengine.com/

I hope that he will also put the engine at his site for download because I understand that he gives free space for chess engines.

Uri

Maybe I can host one for you since I have a lot of free space, Michael has agreed that I can make Romi betas available for download.