TCEC STOCKFISH DECISION

Moderator: Psyck

Post Reply
terrapene
Posts: 69
Joined: Sun May 17, 2015 10:11 am
Real Name: Yes

TCEC STOCKFISH DECISION

Post by terrapene » Sun Oct 11, 2015 7:07 pm

all received votes were "YES", but how much you wanna bet houdart did not respond?
how many actual YES votes were there?


TCEC Press release, Oct 11, 2015

Stockfish has submitted for Stage 3 of TCEC a not fully stable version, which contains a few bugs related to the new “lazy-smp” code. This has prevented the engine to finish two of the four games played so far in the event, instead losing these games on time.

Although the Stockfish team considers this to be a “serious play limiting bug”, it cannot be defined as such by the actual rules of TCEC. The intent of this rule is in cases where the engine a) is hard crashing, meaning it’s unable to make any move, or b) not being able to communicate with the command line interface, meaning it’s unable to make any move. Losses on time because of problems relating to the time management, as in the case of Stockfish, does not fall under this rule.

Furthermore, submitting a not fully tested and bug free version is disrupting the regular flow of the championship. TCEC understands the desire of engines to prove superiority, but the championship is not a testing and debugging ground. This is an important point to make clear.

On the other hand, the inability of Stockfish to complete games is disrupting the other participants performance, expectations, and even the general fun of the audience.

A friendly solution of the case was needed and it was reached with the participation of multiple parties. Anton Mihailov as tournament director and manager of TCEC had a long discussion with the TCEC Operator Martin Thoresen. Possible solutions were analyzed and the constructive public opinion of the TCEC chat was taken into account as well. It was considered that allowing Stockfish to continue would be a compromise with the current rules set and as such it can only be coordinated with all participants in the Stage. A questionnaire was created for the programmers of the other participating engines, asking “Can TCEC be allowed to revert the current version of Stockfish (071015L), playing in Stage 3, to the version that played in Stage 2 (070915)?”. Every participant had the right of veto with a “NO” vote, but in a friendly and exceptional display of sportsmanship all answers received from the engine programmers were “YES”.

With the vote complete and the facts in hand, the situation is resolved as follows:

1. Stockfish will be allowed to revert to the version that played in Stage 2 (070915)
2. For fairness and balance, the revert process will only be complete after Stockfish has played each opponent exactly once i.e. after game 15
3. All results so far will be kept and no games will be replayed
4. In the case of Stockfish qualifying for the Superfinal, it will be allowed to play only with versions taken from the master branch – the latest binaries from the abrok.eu/stockfish site

TCEC wants to thank once again all participating parties in the discussion. The decisions above are not easy ones, but are the best middle ground for the participants and fans alike. These decisions are final.

Post Reply