TCEC SuFi openings edited

[pohl4711]

The TCEC Superfinal openings (Season 9-15) from J.Noomen.
Edited by me. What have I done?

1) deleted all lines 4 plies or shorter (to eliminate some really strange stuff like 1.g4)
2) deleted all lines with non-unique endpositions (with pgn-extract, command: pgn-extract --fuzzydepth 0 -D games.pgn --output unique.pgn)
3) checked all endpositions with Komodo 13.02: 10'/move (!!!), Quadcore, eval-interval [-0.50;+0.60]

(In the pgn-file, the eval and search-depth of Komodo are stored in the Annotator-TAG of each game/line by the pgnscanner-tool. Looks like this:
[Annotator "game 5, depth=39 val=23 filter=[-50/60] status=accepted"].
Means a search depth of 39 and an eval of +0.23 (all evals are in white's point of view).

237 out of 350 lines remaining. All endpositions are unique and not too bad for white or black.

Forthcoming Superfinal seasons openings will be added in the future.

(C) and all Kudos to J.Noomen


I did a testrun of the 237 lines (=474 games)

All games played with 5'+3'' (average game-duration: 15-16 minutes), singlecore, 256MB Hash, no endgame-bases, no ponder, both engines with Contempt=+15, LittleBlitzerGUI

Score is the result of asmFish 170426 - Komodo 10.4. Example:
Score 60.3% – 39.7% means, that asmFish scored 60.3% and Komodo scored 39.7%

Here the results of some of my older testruns, which I did for the development of my SALC and Drawkiller openings:

Each testrun 1000 games:
SF-Framework standard 8 move openings: Score 60.3% – 39.7%, draws: 63.4%
FEOBOS v20 contempt 5: Score 58.7% - 41.3%, draws: 64.1%
HERT 500: Score: 60.6% - 39.4%, draws: 60.4%
4 GM-moves short book: Score 60.5% - 39.5%, draws: 57.1%
SALC V5 full-closed 500 positions: Score 66.5% - 33.5%, draws: 47.7%
Drawkiller (small 500 positions set): Score: 66.4% - 33.6%, draws 30.5%

Here the result of the testrun of the TCEC Superfinal openings:
TCEC Superfinal openings (Season 9-15) edited (237 lines, 474 games): Score: 62.5% - 37.5%, draws: 50.4%

As you can see, the result is better, than the results of the other "classical" openings (SF-8moves, FEOBOS, HERT and 4 GM-moves): lower draw-rate and better Elo-spreading of the result. And not so far away from the result of my SALC openings. Which is really impressive. Of course, Drawkiller is out of reach and in its own dimension...


Download the TCEC Superfinal openings (edited) on my website:
https://www.sp-cc.de

or directly here:
https://www.sp-cc.de/files/tcec_sufi_op ... edited.zip