Leela Chess 10171 demonstrates a stunning theoretical novelty within Ruy Lopez Open – vs Gull

Leela Chess 10171 demonstrates a stunning theoretical novelty within Ruy Lopez Open – vs Gull

59 thoughts on “Leela Chess 10171 demonstrates a stunning theoretical novelty within Ruy Lopez Open – vs Gull

  1. Replayable game with nicely indented variations: https://www.chessworld.net/chessclubs/ltpgnviewer32/ltpgnboard.asp?GameID=5023634&v=Hr5Zeix2gJs

  2. We all are just as excited as you KC. This is a scientific discovery. My mind is blown. Its like pawns are more valuable then just 1 point it is like leela gives pawns a 2.5 evaluation

  3. a very interesting statistic for all leela games would be, percentage-wise how many moves by leela were expected by the engine.

  4. It seems that Leela counts pawns as valuable pieces. The further along they are to queening the more valuable they are. Leela is not sacrificing a piece, it is exchanging a knight for a deep passed pawn. Deeper pawns have more pressure and more value.

  5. Another great Leela innovation. KC, if you ever feeling creative I would love to see, in the background behind you, a royal-looking study full of leather bound books. Seriously. And then perhaps you could wear a cloak and have a pipe! Plus, you would get Anna Rudolph's attention and therefore more chess lovers following your channel.

  6. Leela start to look now more like an evil angry man with a lot of rage wait … Where did i see this … 😮 Omg leela is becoming a Tal 😮

  7. I manage to install Leela and she think white was better up to 20.Na5? I think she just play against weak engine for now I make her play vs latest SF developer and she lose convincingly.

  8. You can actually use Leela Chess Zero to perform an analysis of any position, so you don't have to remain in the dark about what her plans were. I'm not actively following the project, so I don't know the details of how to set it up for such analysis, but I'm sure you could find some documentation about it, or perhaps better yet just ask someone in the forums or on the project to point out the best way how to do it. In fact, it might make for some interesting videos if you went into the process a bit, and showed some interesting results from such analysis, especially as compared with traditional engine analysis. 😊

  9. Great video, again. It seems that 12…Qg6 may be a novelty, which "deserves attention" (as they say in the books) at least. 12…0-0 was played by Bjarne Winkel against David W Coyne in an IECC game in 2003. Both players were rated under 2000 (175 ECF). White won.

  10. Very weird game. To me, it never looked "visually crushing", as Leelas positions often do, until all of a sudden white was forced to sack the queen

  11. exciting idea, will be great to see Leela come up with the ideas and Stockfish or GMs examine them thoroughly. Thanks KC keep up the good work

  12. kingcrusher others have mentioned it : u really should stick to the game line more at 1st. Variation analysis should come afterward. I lose track of the game every time and it just feels like variation analysis. this significantly detracts from the enjoyment of the vid.
    sincerely

  13. Sorry if I missed it, but what was the time control? Inspiring chess and very interesting analysis, but for better understanding of these games probably a deper analysis is needed (I know that in these short videos it is hard to do that).

  14. My problem with anything other than the mainline is that they have a purely materialistic approach to determine if it's better or not, and this game proves my point! No engine would have sacced a knight just to get an agressive chain of center pawns, it would have been considered as "bad" while leela clearly showed otherwise.

    The better option clearly is to ask leela what she would have done instead of asking other traditional engines.

  15. KC: you wanted some testnet vs main net games. I have one from a match where I'm playing through the TCEC s12 superfinal opening. Here ID512 show's T161 how it's done in a King's Indian. https://lichess.org/Cj3A7stU

  16. Awesome! lol. What a nightmare would be to lose with white a piece up against a lower rated player after f5! Leela plays like a child prodigy.

  17. I have explored this variation for decades and iirc Ng5 is better than Nxd4. My comp is in the shop right now but I will look into it again.

  18. This is genius! I use to play the Muzio Gambit in my King's Gambit days, and in those lines you generally sacrifice a bishop and a knight for control over the f- file. Black has managed to control the f-file with sacrificing just one knight. It is exciting and weird to see Leela do this. I wonder how she would play the King's Gambit or Vienna Game.

  19. Even if this was actually sound opening I don't think it can be successfully used by humans. One small mistake can lead to quick downfall. I'm all for gambits but this seems unplayable from human perspective. Although I am looking forward to more exploration in this position cause who doesn't love such madness?

  20. Yasser Seirawan lectured on this opening variation five years ago: https://www.youtube.com/watch?v=qejc9ISNMBU&list=PLVWaFpMwtaGiVZ77NhhvGGGzvF7oFSWcA&index=8#t=8m

  21. By the time Leela moves the pawn to F5 it seems Leela and Stockfish 9 agree that black is down about 1 pawn.

  22. What kind of world is this – Trump is the president of the US and we have a computer program that plays really strong coffee house chess!

  23. Can you possibly just explain the game and the great moves without going deep into variations? I feel loke every chess channel does that and it just takes a lot longer to get through a video. You could even just show the full game and then part 2 halfway through would be where you could go into depth about the variations.

    I feel like you get hard on variations lmao.

  24. such a noobist comentator, boring video lots of bla bla and no showing top engine evaluation, arrow moves etc… shit channel

  25. This just goes to show you that Artificial Intelligence has figured the game of chess out better then ordinary chess engines! Alpha Zero did something similar to Stockfish and ended up destroying Stockfish in the end! The major difference between Artificial Intelligence and regular chess engines is that chess engines for the most part are one extremely materialistic. And two that they place a stronger value on pieces then on pawns. Artificial Intelligence is more concerned with quality when it comes to pieces whereas regular chess engines are more concerned with quantity.

  26. Maybe you could setup some simulations starting from the critical position? If you play 1000 30 sec games, it will take ~17 hours to completion

  27. so far as I can tell 10.Nxc6 bxc6 11.nc3 be7 12.Nxe4 fxe4 13 Rxe4 gives white plenty without an answer by black 10….dxc6 turns out worse for black

  28. Anatoli Karpow needs to analyze this as he used to be one of if not the strongest Spanish player.
    This is crazy Kasparow preperation type of shit XD

Leave a Reply

Your email address will not be published. Required fields are marked *