Romichess P3k test

[Tony Thomas]

The new version of Romi is finally here. Can it live up to its expectations? The version that was released is not tested by me. If I had a choice I would have told him to release NG9 which scored 237.5/420 games against 14 different common opponents an all time best. Since Mike went ahead and released this version, I decided to do the test to make sure that nothing is broken. I want Romichess to promote to Leo's 3rd division this time around, Romi disappointed me in last two editions.

The Goal

Code: Select all

8 RomiChessNG9              : 2457  420 (+199,= 77,-144), 56.5 %

Danasah 2.85                  :  30 (+ 12,= 10,-  8), 56.7 %
Phalanx Reborn                :  30 (+ 13,=  6,- 11), 53.3 %
Delphil 1.6c                  :  30 (+ 18,=  5,-  7), 68.3 %
Francesca MAD 0.13            :  30 (+ 13,=  7,- 10), 55.0 %
Lime 6.3                      :  30 (+ 25,=  3,-  2), 88.3 %
Djinn 0.925x                  :  30 (+ 15,=  5,- 10), 58.3 %
Zeus 1.28                     :  30 (+ 26,=  1,-  3), 88.3 %
Zappa 1.1                     :  30 (+ 11,=  4,- 15), 43.3 %
GreKo 5.2                     :  30 (+ 16,=  8,-  6), 66.7 %
NanoSzachy 2.7                :  30 (+ 19,=  2,-  9), 66.7 %
Arasan 9.5                    :  30 (+ 10,=  9,- 11), 48.3 %
Horizon_4_3_173               :  30 (+ 17,=  7,-  6), 68.3 %
Naum 2.0                      :  30 (+  3,=  5,- 22), 18.3 %
ChessTiger2007.1 UCI          :  30 (+  1,=  5,- 24), 11.7 %

Current

Code: Select all

25 RomiChessP3k              : 2363   30 (+ 16,=  2,- 12), 56.7 %

NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

[Michael Sherwin]

Hi Tony,

Sorry. I should have just released NG9 as P3k. I always seem to self destruct when up against a deadline. Maybe I should just release NG9 as P3l and then take a rest.

[Tony Thomas]

Hi Tony,

Sorry. I should have just released NG9 as P3k. I always seem to self destruct when up against a deadline. Maybe I should just release NG9 as P3l and then take a rest.

If they are almost the same strength, then you dont even have to worry about a Proto 3L. I am using the JA compile by the way. I am not sure how much speed increase Romi gets from it. Another not so good match. To get a boost in elo I could try the match against chess tiger and Naum but I will save the best for last.

Code: Select all

24 RomiChessP3k              : 2378   60 (+ 40,=  5,- 15), 70.8 %

Lime 6.3                      :  30 (+ 24,=  3,-  3), 85.0 %
NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

[Dr.Wael Deeb]

Hi Tony,
Wouldn't be a good idea to run a couple of 200 games against two reliable opponents with learning switched on so that it gains some experience and then test it in your tournaments

[Graham Banks]

Hi Tony,

Sorry. I should have just released NG9 as P3k. I always seem to self destruct when up against a deadline. Maybe I should just release NG9 as P3l and then take a rest.

Hi Michael,

I've included RomiChess P3k in the Division 6 field for the next series of my Amateur Championship.

Regards, Graham.

[Tony Thomas]

Hi Tony,
Wouldn't be a good idea to run a couple of 200 games against two reliable opponents with learning switched on so that it gains some experience and then test it in your tournaments

Most tournaments Romi going to be competing in are going to be learn off tournaments. So I am pretty much relying on your ability to make a good book. I replace the learn file with the original(yours) learn file after every 30 games to reduce the influence of learning on the pure engine strength.

Here is the current standing, a very decent match against Danasah, second best to date. Comparison after the same matches, current version 63.5/90, previous best version 63/90. So the current version has a 1/2 point advantage.

Code: Select all

14 RomiChessP3k              : 2437   90 (+ 58,= 11,- 21), 70.6 %

Danasah 2.85                  :  30 (+ 18,=  6,-  6), 70.0 %
Lime 6.3                      :  30 (+ 24,=  3,-  3), 85.0 %
NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

[Dr.Wael Deeb]

Hi Tony,
Wouldn't be a good idea to run a couple of 200 games against two reliable opponents with learning switched on so that it gains some experience and then test it in your tournaments

Most tournaments Romi going to be competing in are going to be learn off tournaments. So I am pretty much relying on your ability to make a good book. I replace the learn file with the original(yours) learn file after every 30 games to reduce the influence of learning on the pure engine strength.

Here is the current standing, a very decent match against Danasah, second best to date. Comparison after the same matches, current version 63.5/90, previous best version 63/90. So the current version has a 1/2 point advantage.

Code: Select all

14 RomiChessP3k              : 2437   90 (+ 58,= 11,- 21), 70.6 %

Danasah 2.85                  :  30 (+ 18,=  6,-  6), 70.0 %
Lime 6.3                      :  30 (+ 24,=  3,-  3), 85.0 %
NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

Ok,I see,the opening book is good,you can count on it
BTW,there is a new version of nanoSzachy,v3.1

[Tony Thomas]

Hi Tony,
Wouldn't be a good idea to run a couple of 200 games against two reliable opponents with learning switched on so that it gains some experience and then test it in your tournaments

Most tournaments Romi going to be competing in are going to be learn off tournaments. So I am pretty much relying on your ability to make a good book. I replace the learn file with the original(yours) learn file after every 30 games to reduce the influence of learning on the pure engine strength.

Here is the current standing, a very decent match against Danasah, second best to date. Comparison after the same matches, current version 63.5/90, previous best version 63/90. So the current version has a 1/2 point advantage.

Code: Select all

14 RomiChessP3k              : 2437   90 (+ 58,= 11,- 21), 70.6 %

Danasah 2.85                  :  30 (+ 18,=  6,-  6), 70.0 %
Lime 6.3                      :  30 (+ 24,=  3,-  3), 85.0 %
NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

Ok,I see,the opening book is good,you can count on it
BTW,there is a new version of nanoSzachy,v3.1

There is a new version of most of the engines that I am testing against. The goal of the test is to measure the relative strength. I did intend to add the new version of all the test engines, but with the CPU over heating and all, I havent had the time.

[Michael Sherwin]

Hi Tony,

Sorry. I should have just released NG9 as P3k. I always seem to self destruct when up against a deadline. Maybe I should just release NG9 as P3l and then take a rest.

Hi Michael,

I've included RomiChess P3k in the Division 6 field for the next series of my Amateur Championship.

Regards, Graham.

Hi Graham,

It's about time!

What took you so long?

Just kidding!

This time I was able to disable learning without breaking anything!

[Michael Sherwin]

Hi Tony,
Wouldn't be a good idea to run a couple of 200 games against two reliable opponents with learning switched on so that it gains some experience and then test it in your tournaments

Most tournaments Romi going to be competing in are going to be learn off tournaments. So I am pretty much relying on your ability to make a good book. I replace the learn file with the original(yours) learn file after every 30 games to reduce the influence of learning on the pure engine strength.

Here is the current standing, a very decent match against Danasah, second best to date. Comparison after the same matches, current version 63.5/90, previous best version 63/90. So the current version has a 1/2 point advantage.

Code: Select all

14 RomiChessP3k              : 2437   90 (+ 58,= 11,- 21), 70.6 %

Danasah 2.85                  :  30 (+ 18,=  6,-  6), 70.0 %
Lime 6.3                      :  30 (+ 24,=  3,-  3), 85.0 %
NanoSzachy 2.7                :  30 (+ 16,=  2,- 12), 56.7 %

Okay, will the real Romi please stand up.