CEGT Blitz 40/4 started

[Heinz van Kempen]

Hi all ,

this is a first general answer to all the many contributions here.

First of all, thanks for giving your opinions (more are of course still welcome). For the CEGT testers it was surprising that there is so much interest in Blitz for comparison. Of course like also in our team opinions vary a lot.

First to explain the situation: CEGT really is a promising project and should continue over years. But for the moment we have only around ten fast machines. This means when using more than three machines for Blitz the usual tests and tournaments with 40/40 would be reduced too much and those longer games should always be our main goal.

Blitz anyway is also nice and allows to have more engines and versions, even betas and privates in.

So we had a vote from all CEGT testers and in principle all want to add a Blitz list. There was also a majority for a time control 40/4 adapted, what would then be 40/3 repeated on fast machines. Anyway one of our testers did not accept the vote, because there was no discussion before and opinions were only given with the vote. His proposal is to use somethig like 40/8 or comparable. It is easy to calculate that such a time control already coming close to rapid chess with the available ressources would lead to test again only the elite commercials and a few more amateurs. It is easy to calculate how long a game will last with 40/8 repeated for the many games we have with more than 100 moves.

I resume a few advantages for a time control 40/4 repeated and for Blitz in general:

a) 40/4 is exactly 1/10 of our usual timecontrol

b) we can include more engines, versions and betas and do not have to neglect our 40/40 games too much

c) it offers a quick estimation if engines might be also strong enough for the longer timecontrol

d) 40/4 is well comparable to the 40/40 time scheme and we already have some vast Blitz rating lists with Fischer time control for example the YABRL from Robert Allgeuer and the list from CEGT tester Michael Koppel with 6+4 and 7+4 on faster Athlon?s. I would anyway not say that x+y is worse with a decent increment not causing the engine to blunder in endgames. What I do not like personally is anything like 5+0, because of the time trouble in endgames leading to severe blunders and there are more games with exceed of time limits.

Best Regards
Heinz

[Robert Allgeuer]

Having different rating lists under different conditions is interesting, because it allows to compare and to draw conclusions which conditions favour which engine.

I agree that in x+0 time problems in the endgame will overly influence results, which I do not think is desirable. Hence x+0 is almost certainly not a good choice. Otherwise x/y or x+y is a matter of taste, IMHO.

What I personnally like about x+2 or similar is that it limits the time used for some of these extra long drawn games (200 or even 300 moves long)at least to a certain extent. We all know them: 49 bishop moves, then one single pawn advanced and another 49 bishop moves and so on. A nuisance to watch.

Robert

[Heinz van Kempen]

Hi all ,

Alessandro was so kind to offer his help for starting a Blitz rating list. We agreed to use 40/4 adapted what would mean 40/3 on faster computers like Alessandro has. I will use an Athlon64 3500+ and other CEGT testers will also surely help as soon as they are back from holidays. Much easier it would be if two or three people would still come to help us.

Anyway it will of course need weeks or even months to build a rating list with a lot of games for many engines.

We agreed to start with the best amateurs and run the tournaments as long as all have 500 games. The better group then has to face the commercials and the weaker ones will play against engines that are close to CEGT level and or newer versions or betas.

So here is a provisional startlist taken from CEGT 40/40. Updates are still possible until Wednesday midnight and we will start on Thursday then. For those updates I can already tell that they will also be included with 40/40 then. Charles tests the new versions with longer time control and Ralf will help him. I also think that Michael will prefer to add matches, as 40/4 is almost bullet for him . I will also ask Chris, Rainer and Uschi what they prefer (Blitz or longer time control), anyway we should leave most computers for playing the 40/40 tournaments.

Code: Select all


1 Fruit 2.1                      : 2713   12  12  2063    61.4 %   2633   35.0 %
2 List 512                       : 2629   15  15  1439    49.5 %   2633   33.9 %
3 Pro Deo 1.1                    : 2628   16  16  1338    49.3 %   2633   30.6
4 Scorpio 1.3                    : 2617  115 112    29    62.1 %   2531   27.6 %
5 Spike 0.9a                     : 2612   14  14  1416    51.3 %   2603   36.7 %
6 SOS 5 for Arena                : 2600   15  15  1343    45.5 %   2631   35.1
7 Pharaon 3.3                    : 2600   21  21   704    55.3 %   2563   33.8 %
8 Pseudo 0.7c                    : 2598   17  17  1025    48.2 %   2611   36.6 %
9 SlowChess Blitz WV             : 2593   16  16  1223    47.6 %   2610   36.6 %
10 Aristarch 4.50                 : 2593   16  16  1297    44.6 %   2630   31.8 %
11 Glaurung 0.2.4                 : 2590   22  22   610    54.4 %   2559   35.7 %
12 Naum 1.8 ?                   : 2590   55  54    89    54.5 %   2559   43.8 %
13 Scorpio 1.3                    : 2577   22  22   639    53.1 %   2556   34.6 %
14 Zappa 1.0                      : 2571   15  15  1314    45.9 %   2600   34.9 %
15 Thinker 4.7a                   : 2562   15  15  1265    45.8 %   2592   38.3 %
16 Delfi 4.5                      : 2555   16  16  1242    41.9 %   2612   32.0 %
17 Green Light Chess 3.01.2.2     : 2553   19  19   798    48.4 %   2564   36.1 %
18 Little Goliath Evolution       : 2550   54  54    97    49.0 %   2557   40.2 %
19 ETChess 01.05.05               : 2545   57  57    89    47.2 %   2564   38.2 %
20 AnMon5.53                      : 2535   22  22   673    48.7 %   2544   29.1 %
21 The Baron 1.6.1                : 2532   26  26   502    48.7 %   2541   29.3 %
22 Crafty Cito 1.2                : 2530   24  24   523    48.2 %   2543   34.4 %
23 Ufim 7.00                      : 2527   61  62    76    46.1 %   2555   39.5 %
24 Zarkov 4.75                    : 2515   55  55    99    43.4 %   2561   36.4 %
25 Movei 00.8.310                 : 2503   19  19   879    41.1 %   2565   33.8 %
26 Yace 0.99.87                   : 2502   22  22   679    36.5 %   2598   30.3 %
27 Amyan 1.597                    : 2481   21  21   709    40.3 %   2550   33.7 %
28 Pepito v1.59                   : 2475   40  41   182    38.2 %   2558   37.9 %
29 Petir 2.75
30 WildCat 4
31 Jonny 2.75
32 free spot



Some notes:

-Anaconda is postponed, because Alessandro wants to use Arena GUI

-Zarkov and E.T.Chess...CEGT testers voted last time by majority to include also privates and betas in Blitz. Those versions were given to me for the ATL tournaments

-Scorpio: it would be fine if a bugfix for Scorpio 1.4 would be available that does not crash. Otherwise we can use Scorpio 1.3

-Naum 1.8 is announced (WBEC Forum). If it will not be available until Thursday morning it will be postponed. I will write an email to Alex to ask

-for Ufim the new version will be used as matches with 40/40 are already started

-same for Petir 2.75 where matches 40/40 start on Wednesday

- Wild Cat and Jonny are included. The version tested in Blitz will then also be tested with 40/40. Updates are of course especially here welcome, if the authors have something stronger

Best Regards
Heinz

[Uri Blass]

I think that I will not send a new version not because I have no new version but because I am interested in comparison with older version.

I am afraid that if I send now a new version then it means that 310 will not get tested.

My opinion about testing at long time control is that it should not be done if the author is not interested in it.
Note that I have no objection to testing new version at long time control if I know that it does not mean delay in testing newer versions.

Uri

[Heinz van Kempen]

Hi Uri ,

starting with version 00.8.310 is just fine because it is the only Movei in CEGT so far and testers are also interested in exact comparison. You can add then a new version for Blitz in the second stage. Due to our voting for Blitz there will be a maximum of four versions or betas tested yearly. Of course there will be exceptions when we are seeing big advances.

You have also the option to send a new version for longer time control as soon as Movei 00_8_310 has 1000 games with longer time control or a bit earlier and just wait until it will be its turn with Blitz. For example there was a new version also for Glaurung and Ufim, before they had 1000 games and the new ones are already tested. Anyway it will be tried to give still games to Ufim 6.00 additionally until it has 1000 games.

Like in AEGT I think we will change some day to change to only testing engines of interested authors, especially when testers should drop out. For the moment we like the variety we have and can afford to test them all and especially the strongest. Additionally the work on some older engines seems to have stopped and no new version will come (for example Pepito). Here we do not hurry to have a lot of games with longer time control and after it has enough tests will be finished.


Best Regards
Heinz

[Igor Korshunov]

Hi Heinz!

I am going to release new version of WildCat very soon. So, you can postpone WildCat if she not be ready in time.

But it is wonder for me your addiction to new versions. From my point of view testing of old versions is not less interesting.

I like such old engines: Amy, Aristarch, Comet, Deep Sjeng, Delfi, Gandalf, GLC, Gromit , LambChop , Leila, List, Nejmet, Pepito, Phalanx, Resp, SmarThink, Tao, TCB, The King, Thinker, WildCat , Yace.

This all exellent engines are above 1 year old. I still very like them.


Best wishes,
Igor

[Heinz van Kempen]

Hi Igor ,

you are correct. All fine engines and there is no one I do not like.

As you know there is only one problem: when including too many engines you will not get enough games for each one. Enough for me would be 1000 per version, a compromise would be 500. So starting with 32 engines will already need a long time as we do not want to use more than two or three machines for Blitz. But patience is needed for any tournament giving some statistical value.

Anyway it is planned to add those engines after the initial tournaments. This means the better ones of this tournament will have to face the commercials and the second half can then play against those you mentioned.

Best Regards
Heinz

[Heinz van Kempen]

Hi all ,

the Blitz tournaments are started by Alessandro and me. Other CEGT testers can of course add tournaments from time to time, but it should not take too much CPU ressources.

Planned is the following:

1. starting with 32 strong amateurs (list below) and run tournaments until all have 500 games each

2. the best 16 will have to face the commercials then for another 500 games each

3. the other 16 will play against new releases and or a bit weaker amateurs

4. new versions will be gauntleted into each group according to their strength. Engines improving can promote from one group to the other and will receive additinal games there

All this will of course last a few months.

Startlist for step 1 (40/4 adapted to 2 Ghz, for most of our faster machines this comes to 40/3, around 64 MB hash for all engines).

Code: Select all


1 Fruit 2.1                      : 2712   12  12  2079    61.3 %   2633   35.0 %
2 List 512                       : 2630   14  15  1460    49.6 %   2633   33.9 %
3 Pro Deo 1.1                    : 2628   15  15  1342    49.3 %   2633   30.7 %
4 Spike 0.9a                     : 2611   14  14  1427    51.3 %   2602   36.7 %
5 SOS 5 for Arena                : 2601   15  15  1358    45.6 %   2631   35.1 %
6 Pseudo 0.7c                    : 2597   17  17  1032    48.2 %   2610   36.5 %
7 Pharaon 3.3                    : 2597   21  21   728    55.2 %   2561   32.8 %
8 Aristarch 4.50                 : 2595   15  15  1332    45.2 %   2629   32.1 %
9 SlowChess Blitz WV             : 2592   15  15  1238    47.5 %   2609   36.7 %
10 Naum 1.8                   
11 Glaurung X
12 Scorpio 1.4                    : 2578   88  87    46    57.6 %   2524   28.3 %
13 Anaconda 2.0.1                 : 2573   16  16  1279    44.4 %   2612   33.9 %
14 Zappa 1.0                      : 2571   15  15  1349    46.0 %   2598   35.1 %
15 Thinker 4.7a                   : 2561   15  15  1286    45.7 %   2591   38.3 %
16 Delfi 4.5                      : 2555   16  16  1261    42.1 %   2611   32.1 %
17 Green Light Chess 3.01.2.2     : 2550   19  19   825    48.2 %   2563   36.0 %
18 Little Goliath Evolution       : 2547   51  51   109    49.1 %   2554   39.4 %
19 ETChess 01.05.05               : 2544   57  57    89    47.2 %   2563   38.2 %
20 Ufim 7.00                      : 2537   43  44   155    49.0 %   2543   37.4 %
21 AnMon5.53                      : 2534   21  21   717    48.8 %   2542   29.0 %
22 The Baron 1.6.1                : 2530   25  25   525    48.6 %   2540   30.1 %
23 Crafty Cito 1.2                : 2528   24  24   540    48.1 %   2541   34.4 %
24 Zarkov 4.75                    : 2510   54  55   103    43.2 %   2557   35.9 %
25 Movei 00.8.310                 : 2505   19  19   902    41.6 %   2564   33.8 %
26 Yace 0.99.87                   : 2500   22  22   689    36.4 %   2596   30.2 %
27 Petir 2.75                     : 2499   51  51   127    43.3 %   2546   29.9 %
28 Amyan 1.597                    : 2483   21  21   736    40.8 %   2548   34.0 %
29 Pepito v1.59                   : 2475   39  39   193    38.3 %   2558   38.3 %
30 Wild Cat 5
31 Jonny 2.82 (beta)
32 free spot



Glaurung will be gauntleted in as soon as the new version is out.

Arena and Fritz GUI are used, as well as different shorter books and also Nunn/Noomen will be used.

Best Regards
Heinz