zaterdag 25 augustus 2018

TCEC 13 superfinal: opening selection

Here are Jeroen Noomen's comments on the opening selection for the TCEC 13 superfinal:

"This are the most important details regarding my selection of the 50 opening lines for the TCEC 13 superfinal:
  • All important openings will be played
  • The chosen lines typically have a length between 3 moves and 13 moves. With a few exceptions: f.e. there is one 3 ply (2 moves) line and one 17 move (34 ply) line
  • Average move length is 7.4 (in the TCEC 12 superfinal it had been 8.18)
  • There are sharp lines in the set, as well as positional lines, more complex lines, lines with opposite castling ("SALC", i.e. short and long castling) and gambits
  • The complexity of a position is one of my main selection criteria, i.e. most pieces are still on the board and there should be a lot of play left, with a reasonable chance for a decisive result
  • In each line one side typically has a slight advantage, with book exits mainly between +0.20 and +0.55 (note: this is with contempt = 0)
  • I have avoided positions that could lead to many exchanges and/or have a high draw rate, as well as positions with symmetrical pawn structures and positions that could lead to easy wins for the side with an advantage

ECO code distribution

  • ECO A: 13
  • ECO B: 12
  • ECO C: 12
  • ECO D: 6
  • ECO E: 7

What did I do differently compared to the previous superfinal selection
There are some important differences regarding my selection method, compared to the TCEC 12 superfinal selection:

  • All lines have undergone a quick check with the 3 top programs (Stockfish, Komodo and Houdini), which means around 5 minutes on a Ryzen 1700X system (16 threads) for each engine on each position;
  • In the quick check the final position of each line is evaluated on a few criteria: score (preferably not 0.00 or above +0.55), no mass exchanges, no easy drawing lines;
  • Please note that the evals in the quick check have been determined by using contempt=0. If an engine uses a contempt other than zero in the superfinal, the book exits will of course be different;
  • Book exits may also be different because TCEC uses much faster hardware compared to the hardware used to do the quick check and because engines are updated and might use different settings;
  • The reason there are less lines selected with ECO codes D and E, is because many lines in these sections are "too easy" for the current top programs. 

My expectations
The TCEC 9 superfinal saw 25% wins, the TCEC 10 superfinal 24% wins, the TCEC 11 superfinal 22% wins and finally the TCEC 12 superfinal 38% wins. It will be very hard to beat the 38% of season 12. I'd like to see a minimum of at least 20% wins. But of course I hope to beat the 25% wins of the TCEC 9 superfinal.

Anything can happen
I have seen quiet openings turn into wins and wild openings turn into draws in my first four superfinal selections. Hence it is my firm view that anything can happen in any line. It is up to the engines to decide how to follow up an opening line and to turn it into an interesting battle (or not). I see myself merely as a facilitator whom picks lines that can lead to interesting and exciting games, but it is surely no guarantee that this will happen :-).

Last but not least: have fun!
TCEC is currently the most important chess engine tournament around. For me it was once again a challenging but interesting job to pick the positions for this superfinal edition. I hope you enjoy the games and we will see lots of exciting chess games!"

3 opmerkingen:

  1. Deze reactie is verwijderd door de auteur.

    BeantwoordenVerwijderen
  2. thanks Jeroen fort putting so much effort in to this. Opening selection is such a viital aspect of computer chess.

    BeantwoordenVerwijderen