1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. b4 Bxb4 5. c3 Ba5 6. Ba3 d6 7. d4 exd4 8. O-O dxc3 9. Qd3 Nf6 10. Nxc3 O-O 11. Rad1 Bg4 12. h3 Bxf3 13. Qxf3 Ne5 14. Qe2 Bxc3 15. Bc1 Nxc4 16. Qxc4 Be5 17. f4 d5 18. exd5 Bd6 19. f5 Re8 20. Bg5 h6 21. Bh4 g5 22. fxg6 fxg6 23. Rxf6 g5 24. Rg6+ Kh7 25. Qd3 gxh4 26. Rxd6+ Kg7 27. Qg6+ Kf8 28. Rf1+ Ke7 29. Rf7# 1-0
This is their most recent ongoing head-to-head: https://www.chess.com/events/2021-tcec-20-superfinal
Current result: 9 draws, one win with Stockfish as White, and one win with Leela as White. Drawn.
There is also this one from a couple of years ago: https://www.chess.com/news/view/computer-chess-championship-... “Lc0 defeated Stockfish in their head-to-head match, four wins to three”. Stockfish did got more wins against the other computers, so won the round-robin, but in head-to-head games Leela was ahead of Stockfish.
stockfish finished +9 over 100 games.
I don't have any stock in those 2 engines, so I don't care which one is better than the other. At the end as a poor chess player it won't change anything :) It's actually interesting to compare how those two software are evolving and how they got here.
Stockfish is much older. And it took it a lot of hand tuning to reach its current level. It is (or was) full of carefully tested heuristic to give a direction to the computation. It would be very difficult to build an engine like stockfish in a short span.
Leela got there very very quickly. Even if it was not able to win in October, the fact that it got competitive and forced the field to adopt drastic changes in such a short period of time is impressive. It seems to be a good example of how sometimes no using the "best" solution could still be a win. Getting good results after a few months against something that required 10 years of work.
Right now, Stockfish is winning in the current TETC, but only by one point (one more win than Leela). https://www.chess.com/events/2021-tcec-20-superfinal Stockfish 12 27.5 - Leela 26.5
One interesting thing to see would be how low-rated humans make different mistakes than Leela does with an early training set. How closely are we modeling how humans learn to play Chess with Leela?
Another thought: Leela, against weaker computers, draws a lot more than Stockfish. While Leela beats Stockfish in head to head competitions, in round robins, Stockfish wins against weaker computer programs more than Leela does.
I believe this is because Stockfish will play very aggressively to try and create a weakness in game against a lower rated computer, while Leela will “see” that trying to create that weakness will weaken Leela’s own position. The trick to winning Chess is not to make the “perfect” move for a given position, but to play the move that is most likely to make one’s opponent make a mistake and weaken their position.
Now, if Maia were trained against Stockfish moves instead of human moves, I wonder if we could make a training set that results in play a little less passive than Leela’s play.
(I’m also curious how Maia at various rating levels would defend as Black against the Compromised defense of the Evans Gambit — that’s 1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. b4 Bxb4 5. c3 Ba5 6. d6 exd4 7. O-O dxc3 — where Black has three pawns and white has a very strong, probably winning, attack. It’s a weak opening for Black, who shouldn’t be so greedy, but I’m studying right now how it’s played to see how White wins with a strong attack on Black’s king. I’m right now downloading maia1 — Maia at 1100 — games from openingtree.com.)