|
< Earlier Kibitzing · PAGE 3 OF 3 ·
Later Kibitzing> |
| Dec-12-17 | | devere: <Tal1949: I am sorry Google- but this game is just ridiculous. Stockfish 8 does not play any of the moves that you claim it does.> Yes, it seems that this "match" is just a publicity stunt by Google to impress gullible people like me. But now even I have figured out the real game they are playing isn't chess. |
|
| Dec-12-17 | | john barleycorn: <devere: ...
this "match" is just a publicity stunt by Google ...> I agree. When things look too good, the chandeliers are too awesome ... you are into a scam. Especially, when Kasparov's face shows up. |
|
| Dec-13-17 | | dannygjk: SF still manages it's time well at various time controls. By the way time management is a non-factor at fixed time/move controls.
The book-Even when Stockfish followed theory in these published games AZ outplayed SF after the opening.
EGTB-Stockfish had lost positions before the EGTB would be of any use.
Hash transposition size-Try it yourself give SF a big hash and see how long it takes SF to see that AZ's sacs were sound.
Based on what I have seen in the published games my theory why AZ outplayed SF is that AZ has vastly superior move ordering. This is supported by the fact that SF was doing 70,000,000 nps while AZ was doing only 80,000 nps. Even if SF has a huge transposition hash table that won't be enough to compensate for much inferior move ordering. Inferior move ordering results in too much time wasted on pointless variations. SF will miss crucial variations because of that. |
|
| Dec-13-17 | | kjr63: Is ..Bc5 and ..Nd7 an opening novelty? |
|
| Dec-13-17 | | dream31: This is absurd!! I could beat Alphazero - Stockfish is playing like at 2000 Elo level. 13. Ncxe5 is like a novice playing. Maybe Stockfish had a virus created by Google? - allegedly. I think alphazero is playing barely 2509 level. |
|
Dec-13-17
 | | Penguincw: Video analysis of this game: https://www.youtube.com/watch?v=-WO.... |
|
| Dec-14-17 | | andrewjsacks: This game is a ridiculous joke. |
|
| Dec-14-17 | | WorstPlayerEver: Freakshow |
|
Dec-14-17
 | | keypusher: I wonder if Go players were as delusional as cg kibitzers seem to be. |
|
Dec-14-17
 | | perfidious: <keypusher: I wonder if Go players were as delusional as cg kibitzers seem to be.> I resemble that remark! |
|
| Dec-15-17 | | dunkenchess: Wrong early sac |
|
| Dec-15-17 | | Scarborough shoal: Computers will crush humans. |
|
| Dec-15-17 | | Scarborough shoal: Stockfish is an i***t. |
|
| Dec-16-17 | | drleper: <Tal1949: I am sorry Google- but this game is just ridiculous. Stockfish 8 does not play any of the moves that you claim it does. 11. Kh1 12. a5 35. Nc4
These are all moves that a 3200 level engine would never play. In fact, 5 out of 28 moves in the middle game (17.8%) are not Stockfish 8 moves, regardless of the time control or lack of opening book. Please do not tamper with the Strength configuration or Contempt- we know what you are doing.> I would guess that these discrepancies depend on the system producing the moves. I've managed to get Stockfish 8 on my system (Intel i7-5820K, running with 4GB hash and 6 threads) to produce 11.Kh1, 12.a5, and 35.Nc4 as the top choices at various times. That said, those moves did change given a bit more time. Interestingly, on move 35 Stockfish 8 initially thinks that 35.Rc1 is good enough for a draw, but if you input that move it soon finds 35...c4! and after backing up, it seems that it really can't decide between 35.Rc1, 35.Ng6, or 35.Nc4 (probably because it's already lost). Most likely stronger hardware would more quickly skip over this illusion that 35.Rc1 provides a draw. |
|
| Jan-10-18 | | sankukay50: Rebel recommends 11...Bc1 after 12'35"
Kh1 has never seen |
|
| Jan-29-18 | | dannygjk: Yes it finds 35...c4 after you enter 35.Rc1 because it's one ply further ahead in the game. |
|
| Jun-14-18 | | ThirdPawn: I find it interesting that AlphaZero plays highly defensive as Black when compared to it playing as White. There are no triple exclaim moves, no sacrifices, no throwing away of pawns, etc, at least from the games released. And positions tend to be very closed in the two games released of it playing with Black pieces. Stockfish on the other hand was quite intidimating and confronting with White pieces. I almost couldn't tell who was who. Perhaps it 'is' White's game... After all, AlphaZero did win 50% of its games as White, but only won 12% of the time with Black pieces. This ratio even held in their 1,200 games. |
|
| Jun-14-18 | | ThirdPawn: In addition, the three wins that stockfish scored against AlphaZero occurred with it playing with White pieces! And like AlphaZero, stockfish had no losses when it played with White pieces. Surely this can't be an anomaly? |
|
| Jul-31-18 | | 1d410: If stockfish played chess perfectly it would give the same move regardless of the system it was on. |
|
| Dec-20-18 | | WorstPlayerEver: I kind of thought why not follow the concept of AO myself? Trap a Queen, sac a pawn. Just be cool. Don't worry for check. 33... Re7
click for larger viewMy idea was this:
33... Re7 34. Qd5+ Kh8 35. Ra1 Qg6 36. Qd8 Qh5 37. Qxf8+ Kh7 38. g4 Qxg4 39. Qf5+ Qxf5 40. exf5 Re2 41. c3 Rc2 42. d4 cxd4 43. cxd4 Bxf5 
click for larger viewStill have to analyze this, but I thought the idea was funny. SFcg 6 secs still plays 33... Qxe5 ~~~~~~~~
33... Re7 34. Qxc5 Rxe4 35. Qd5+ Re6 36. g4 Qc6 37. Qxc6 Rxc6 
click for larger viewFigured out White can't trap the Bishop in this position. I just sacced a fourth pawn. Who cares. Is that brilliant? Of course not, it's kind of average. The White pieces are not coordinated well. A child could see that. The Black Bishops secure the draw. |
|
| Dec-20-18 | | WorstPlayerEver: Oops see I am playing for AO now. |
|
| Dec-20-18 | | WorstPlayerEver: Don't worry, take a move back then, just kidding. 32... Rd7
click for larger view |
|
| Dec-20-18 | | WorstPlayerEver: After 23... Rb4, I thought the White Queen needed a little more spave, so:
click for larger view24. Bxd4 cxd4 25. Qxd4 a5 26. f5 Qc6 27. f6 gxf6 28. Rxf6 Rxc4 29. dxc4 Qxe4 30. Qf2 
click for larger view |
|
| Dec-20-18 | | WorstPlayerEver: Anyway, it's all about spave. You have to create spave. |
|
May-29-23
 | | kingscrusher: It seems Black was actually slightly better out of the opening with the bishop pair. Old stockfish had hand-crafter heuristics and liked denting the opponent's pawn structure. It gave black the bishop pair and it turns out Black is actually already better before the "Symptom" of Nxe5 happened.
Stockfish 15 in 2023 gives Ncxe5 as one of the best moves for White: 4: Stockfish (Computer) - AlphaZero (Computer) 0-1 0.0, AlphaZero - Stockfish 2017 click for larger view
Analysis by Stockfish 15 - 3 threads max:
1. ⩱ (-0.50): 13.Ncxe5 fxe5 14.Nxe5 Re7 15.f4 Re8 16.Be3 c5 17.Qf2 Be7 18.f5 Nd4 19.Qd2 Bh4 20.Bf4 Qe7 21.Qd1 Nc6 22.Ng4 Kh8 23.Ne3 2. ⩱ (-0.56): 13.Ne3 b5 14.Ng1 g6 15.Ne2 Qd6 16.f3 c5 17.Qh4 Nd4 18.Nxd4 cxd4 19.Ng4 h5 20.Nf2 Qc6 21.f4 Qxc2 22.Qg3 Kh7 23.Qf3 3. ⩱ (-0.58): 13.Nfxe5 fxe5 14.Nxe5 Re7 15.f4 Re8 16.Be3 c5 17.Qf2 Be7 18.f5 Nd4 19.Bf4 Bh4 20.Qd2 b5 21.Ng4 b4 22.Qd1 Bg5 23.c3 Bxf4 24.Rxf4 (Gavriel, 29.05.2023) So in other words we seem to be witnessing a "Symptom" - that Ncxe5 is one of the best moves. The root cause is reflected in the opening choice which gave White essentially nothing - and is also one of the reasons the Berlin Defence shot up in popularity in recent years. These variations where black has the bishop pair for pawn structure or even with K in center and Queen's off tend to be fine for black or maybe even slightly better. This game is an example of evolution of chess engines - and later Stockfish embracing NNUE. "If you can't beat them - join them" |
|
 |
|
|
< Earlier Kibitzing · PAGE 3 OF 3 ·
Later Kibitzing> |
Kibitzing
