TCEC relaunched! New rules, Divisions ++ MUST READ!

[Martin Thoresen]

http://www.tcec-chess.org

The major changes are quoted below from my Rules page, but there are also other small changes found on that same page. Also read the News page for more specific info about the next event (Division 3).

General Information

Each game presented on this site is played with a time control of 40 moves in 100 minutes, the next 20 moves in 50 minutes, then finally 20 minutes for the rest plus 10 seconds added per move at this last time control. If an engine loses on time for various reasons the result will stand - hence the game will not be replayed. The openings are always limited to 12 moves / 24 plies when using either a pgn file or a neutral opening book.

Engine Configuration
Each engine is allowed to use up to ~4 GB of hash and is allowed to use up to all six cores of the processor, if this is supported. Pondering or "permanent brain" is not allowed in TCEC so it is always disabled. If an engine executable supports the pop_cnt instructions it will be preferred over the non-pop_cnt executable, the same applies to 64-bit builds being preferred over 32-bit builds. It is not allowed to change the engine configuration or parameters, so the default settings for each engine are always used. In general, most commercial and free engines are available in TCEC.

The TCEC Division System
There are currently 3 seperate TCEC Divisions: 1, 2 and 3 where 1 is the "Premier" Division. Each Division consists of 8 engines. In each season, 56 games will be played per Division, which means a double round robin where everyone of the 8 participating engines play each other twice. The openings are made so that they are random per engine pair in rounds 1-7. In rounds 8-14 the colors are reversed, with the same openings and pairs, so all in all this means that each engine plays both black and white of the same position versus the same opponent.

If an engine currently scheduled to play in any Division has been released in a new version, this new version takes the place of the old version, but only if the Division hasn't started yet. So during a Division being played, no engine upgrades are allowed. When a season ends, meaning that all 3 Divisions and the Elite Match have been played, the top three engines are promoted to the Division above, and the bottom three engines are demoted to play in the Division below for the next season. In Division 3, the three bottom engines will be replaced by three new engines (or previous engines which has been released in a newer version) so they are out of the TCEC system for the coming season. If necessary, tiebreaks can be used to determine promotions and demotions. The first tiebreak critera is the Sonneborn-Berger. If still a tie, the greatest number of wins decides. In case of still being tied, the direct encounter between the tied engines is the last tiebreak.

TCEC Elite Matches
After all the Divisions has finished their group play, engine number 1 and 2 from Division 1 will meet in an "Elite Match" of 40 games. If there are several engines placed as number 1 or 2 in Division 1, then the same tiebreaks described above are used. The match is played with 20 different openings so that each engine plays both black and white of the same position. In the case of a drawn match, 8 rapid games with a time control of 25+10 are played. If still tied, then 8 blitz games with a time control of 4+2 are played. The winner of this match becomes the "TCEC Season X Champion" and this ends the current season.

I hope this will be a success, without all the fans I wouldn't really bother.

[IGarcia]

Hi Martin,

In general looks very good. Personally, I'm interested in 1st division and for sure 2nd division games.

Will be a long wait while all 3 div finishes. Maybe using only 2 divisions will speed up. Counting with external rating (SWCR, CCRL, CEGT or IPON) to determine witch engines will go up into div 2 for replacing the three bottom engines.

Is just a suggestion. Maybe its good to make seasons a bit longer (by keeping division 3) to give extra time to programmers to come with new engines version, or complete new engines when the new season starts.

Regards.

[Denzwell]

I also think that two divisions are sufficient. The third division may be less interesting people

[Andres Valverde]

I also think that two divisions are sufficient. The third division may be less interesting people

Perhaps programmers would like a 4th division though

[M ANSARI]

Actually I think that we already have plenty of testing for 2nd and 3rd division engines. I think you should stick to 1st division top engines at LTC games. I know that I personally became interested in your site because of the super high quality of the games as they are played on very strong hardware without corruption of games due to time blunders.

If I would recommend something, that would be instead of adding a second division, to maybe expand the size of the tourney or matches to include more participants. It would be great to include Naum 4.2, DS, Hiarcs, Shredder, Zappa and maybe even Fritz (if you can work around lack of UCI). What I would hate to see is division 1 with 6 Rybka based engines playing each other. Maybe choose the strongest clone and Rybka 4 and add the other mentioned participants (also nice to keep DJunior).

[Martin Thoresen]

Actually I think that we already have plenty of testing for 2nd and 3rd division engines. I think you should stick to 1st division top engines at LTC games. I know that I personally became interested in your site because of the super high quality of the games as they are played on very strong hardware without corruption of games due to time blunders.

If I would recommend something, that would be instead of adding a second division, to maybe expand the size of the tourney or matches to include more participants. It would be great to include Naum 4.2, DS, Hiarcs, Shredder, Zappa and maybe even Fritz (if you can work around lack of UCI). What I would hate to see is division 1 with 6 Rybka based engines playing each other. Maybe choose the strongest clone and Rybka 4 and add the other mentioned participants (also nice to keep DJunior).

Thanks!

There will not be hundreds of "clones", I am sticking to Ivanhoe
and Houdini in regards to the so called IPP engines.

Zappa is not included because it's obsolete (but still strong, by all
means). Of course when Rondo appears it will find its way into TCEC.

All the other engines you mentioned, besides Fritz (which is not going to
enter TCEC unless they release the engine as UCI), are already included.

I think it's fun with variety and 3 Divisions should keep that alive.

[Martin Thoresen]

I also think that two divisions are sufficient. The third division may be less interesting people

Personally I find it interesting, there are a few engines I really enjoy (like Gaviota for instance).

I want to applaud engine authors that continue to develop their own
engine, step by step even if it's far away from Rybka strength.

I do see your point, of course.

[Martin Thoresen]

Hi Martin,

In general looks very good. Personally, I'm interested in 1st division and for sure 2nd division games.

Will be a long wait while all 3 div finishes. Maybe using only 2 divisions will speed up. Counting with external rating (SWCR, CCRL, CEGT or IPON) to determine witch engines will go up into div 2 for replacing the three bottom engines.

Is just a suggestion. Maybe its good to make seasons a bit longer (by keeping division 3) to give extra time to programmers to come with new engines version, or complete new engines when the new season starts.

Regards.

One Division, 56 games, should take something like 1 1/2 to 2 weeks.
So all in all, perhaps 1 1/2 month for the 3 Divisions plus the Elite Match
to finish. I don't think this is too bad, considering the rather sporadic
setup I've had until now.

[beram]

hi Martin,

Do it as you like it best for yourself.
personally I am more interested in games between the best programs.
And I agree that two of the best from Ipp family is enough.

Kind regards Bram

[Matthias Gemuh]

A swiss tournament with 30 engines and alternating colours can be exciting.
Maybe even with a time gradient from slow games at top of table to fast games at bottom.

I am addicted to swiss.


Matthias.