06 February 2023

TCEC Swiss 4 Underway; CCC Mystery Matches

The action never stops in the low-stakes world of engine vs. engine chess competitions. Our previous fortnightly post mentioned two winners: LCZero Wins TCEC Cup 11; Stockfish Wins CCC19 Bullet (January 2023; 'For the first time in a long while, an engine other than Stockfish won a TCEC or CCC event'). Here's a summary of that post:-

TCEC: In the 'Cup 11' Final match, LCZero and Stockfish tied the regulation games +1-1=10, then LCZero won the second pair of tiebreak games. In the Bronze match for 'Cup 11' third place, KomodoDragon and Ethereal tied +0-0=12; KomodoDragon won the second pair of tiebreak games. • CCC: In the 'CCC19 Bullet Challenger' match, Dragon beat Lc0 527.0-473.0, and in the Final match Stockfish beat Dragon 616.0-384.0.

Since the action never stops, what's the current situation?

TCEC: After 'Cup 11', the site organized 'Swiss 4'; see TCEC Swiss 4 for the rules. Of the 11 'double rounds', the event is currently in the eighth. Stockfish is leading by a half-point ahead of Ethereal and KomodoDragon, which are ahead of 39 other engines. The event should run for another week. For a summary of the previous Swiss, see Stockfish Wins TCEC Swiss 3 and CCC17 Blitz (July 2022).

CCC: After the 'CCC19 Bullet Final' match, the site held a series of exhibition events. The most important was probably the 'CCC19 Bullet Alt Final' match, where Stockfish beat Lc0 301.0-199.0 in a 500 game match. Extrapolating this to the 1000 games played in the 'Final' match, shows that Lc0 performed somewhat better than Dragon.

After this, the CCC conducted a series of events under the header 'Stockfish Thread Dominance'. The first was titled, 'Stockfish Thread Dominance: Weiss'. What were these mysterious matches? The 'Info' tab for the Weiss match misleadingly says, 'Format: CCC19 Bullet (1|1) will consist of the following Events...', and the PGN for the individual games offers nothing informative beyond 'TimeControl "60+1"', i.e. a one minute (bullet) game with one second per move increment, aka '(1|1)'. Stockfish won 73.5-26.5.

The only real info I could find was from end-January: Does this count as anything groundbreaking? (reddit.com) The post showed a screenshot of the result of 'Stockfish Thread Dominance: Dragon', where Dragon beat Stockfish 70.0-30.0, and continued, 'Obviously Stockfish has lost individual games but a whole match? And it looked pretty one-sided.' The most informative comment to the question answered,

That is why the event was called "Stockfish thread dominance", because it was Stockfish with just 1 thread playing against a bunch of different engines, each one stronger than the last, and giving the opponents the full 250 threads available, essentially giving them 250 times more computing power.

Stockfish vs Weiss: Stockfish won by a landslide (0, 0, 9, 35, 6)
Stockfish vs Black Marlin: Stockfish won by a landslide (0, 0, 2, 41, 7)
[...; eight matches total]

The numbers in parentheses, which always add up to 50, weren't explained. There were nine such events listed in the CCC archive.

As I was writing this post, the site had just finished 'OpenBench Interlude #7 - Top 6'. As for plans:-

!next • Actual last Stockfish dominance match: SF Dev 256-thread vs SF Dev 1-thread [not mentioned in the Reddit thread]. Then we are onto 'CCC20 Blitz : Newcomers'.

For this blog's reports on the three main stages of CCC19 -- Blitz, Rapid, Bullet -- see:-

From this we can deduce that each stage runs for about a month and a half.

[For further information from the various stakeholders in the engine-to-engine events, see the tab 'TCEC/CCC Links' at the top of this page. • NB: Leela = LC0 = LCzero; Dragon = KomodoDragon]

No comments: