By Assaf Wool from My TCEC Experience
This season’s Superfinal was unusual in many ways, (check out my statistics report to see some numbers). The draw rate was 71% which is rather low for a superfinal. This was partly due to the fact that there were many (7) game pairs with two white wins. Perhaps the openings were a bit too biased for this season’s finalists? There were 7 games which ended with mate on the board (all Stockfish wins), inconceivable for a superfinal. The games were unusually long even for a superfinal with a NN engine (median of 96 moves per game).
Draw rate, wins
The three most common game termination causes were:
35% – SyzygyTB
35% – TCEC draw rule
12% – TCEC win rule
There were no crashes in the stage. Seven games ended with mate on the board, weird for a superfinal.
Much higher than usual even for a superfinal, the graph is also very unusual. In many games Leela had long shuffle periods, even when it was winning. There were 46 games longer than 100 moves (17 over 150 moves), the longest was 211 moves (Leela – Stockfish, game 58, draw).
Time per game (hours)
The time control was 90 minutes, shorter than usual for a superfinal.
There were variable length book openings in this stage chosen by Jeroen. The first letter of the ECO codes was distributed as follows:
The engines had almost no freedom to choose the opening variant, 90% of the game pairs repeated the same ECO code, 88% repeated the opening variant. In 94% of the game pairs the ECO first letter was repeated twice. The exceptions were a few very short book sequences of length under 5.
Reverse pairs, wins
Reverse pairs, same moves
Pairs of reverse games diverged very quickly, 28% diverged immediately out of book, 66% of the pairs diverged at most after 1 move. The longest repeated sequence of moves was 10 plys (Stockfish – Leela games 79 and 80 that had a 30-ply book opening, Ruy Lopez: closed, Flohr-Zaitsev system, two draws)