Hi all,
I will continue my tradition of answering questions put in private on the blog, without giving the source of the question to cause no embarrassment. I do this because I know that a lot of people will find the question and answer interesting.
Question:
Hi Jacob,
I am searching for a (Quality Chess) White repertoire after
- d4 d5
- c4 c6
- Nf3 Nf6
- Nc3 e6
- Bg5 h6
- Bxf6
I only found 6. Bh4 which Schandorff calls the “Anti-Moscow-Gambit”, but no matter which line I analyse afterwards with a computer (Stockfish 8 engine), even at a very high depth, there is nothing better than 0.00 according to the computer. The Quality chess book from 2012 by Schandorff (White repertoire Queen’s Gambit) recommends 6. Bh4 dxc4 7. e4 g5 8. Bg3 b5 9. Be2 Bb7 10. h4 (which according to the computer is the best continuation after 6. Bh4). But in his book from 2015 (Black repertoire The Semi-Slav) Schandorff says and I quote: “10. h4 used to be a respectable main line, but in the last couple of years White has scored miserably after 10. … g4 11.Ne5 Nbd7! [again the best continuation according to the computer] so it seems that the White players no longer trust this continuation.” If 10. h4 is the best move in that situation (and the computer also says it is the best), but only gives 0.00 as evaluation, then there must have been a bad move for White before. But which one?
Now I may have an idea which one: The computer says that 6. Bxf6 (something between 0.15 and 0.30) is better than 6. Bh4 (0.00). But even in 2015 (Black repertoire The Semi-Slav) Schandorff says “6. Bh4!” is the critical move (see Schandorff, The Semi-Slav, 2015, page 57 on the right side right under the diagram). He then gives a line leading to 10. h4 saying the things I mentioned above. But how can “6. Bh4!” be called the best/critical move for White if afterwards he gives a line leading to saying “White has scored miserably. White players no longer trust this continuation.” without giving better white moves inbetween? In the end it is White, and White’s aim is to keep a slight advantage out of the opening. You can’t say 6. Bh4! is the best/critical continuation and then giving a line that leads to no advantage for White (according to the computer) and then say “Oops sorry, White scored miserably here, White players don’t trust this continuation anymore.” That is contradictory to me.
Could you explain this contradiction?
Where do I find a good White repertoire after 6. Bxf6 (which seems to be the main line and best continuation)? I only found Black repertoires after 6. Bxf6 so far.
Best regards
Answer:
There are three central question and I will answer them in turn.
1) No, we do not have a repertoire with 6.Bxf6 in the Semi-Slav and I am unsure if we have anything planned there. We cannot cover everything.
2) The word “critical” does not necessarily mean best. It means the most dangerous for the evaluation of the line. In this case, if a clear path for an advantage for White was to be found against this opening, which we have to remember is in the top 3 of Black openings after 1.d4 in the 21st Century, it would be found after 6.Bh4. Is it best? Who knows. This is part of what the opening is about, exploring different ideas and interpretations of the endless amount of possibilities. I know that I would personally be more afraid of being hit with a killer novelty after 6.Bh4 than after 6.Bxf6. But in the end I think the two moves are about style more than evaluation.
3) Using the engine at move 6 to decide between two main lines. This is not giving you a good understanding of the position in such an analysed area. Yes, sometimes the engine can come with great discoveries. But do a mind check, do you think Anand will play 6.Bh4 if he knows that the engine prefers 6.Bxf6. Well, he does this. As do other top players. He also takes on f6 at times. Or plays 1.e4. The engine is not useful this early.
Take this example:
Clicking on the image should make it bigger.
At move 15 in the opening, we see that the natural 15.h4 is third and gives no edge. Now see what Negi gives in his book:
1. e4 e6 2. d4 d5 3. Nc3 Bb4 4. e5 c5 5. a3 Bxc3+ 6. bxc3 Ne7 7. Qg4 O-O 8. Bd3 f5 9. exf6 Rxf6 10. Bg5 Rf7 11. Qh5 h6 12. Nf3 Nbc6 13. Bh7+ Kxh7 14. Qxf7 hxg5 15. h4! g4 16. h5 Qg8 17. Ng5+ Kh8 18. Qg6 Nd8 19. c4 Bd7 20. cxd5 exd5 21. Kf1 Bf5 22. Re1 Qf8 23. Rxe7 Bxg6 24. hxg6+ Kg8 25. Re1!!
You can check this yourself. The engine comes very late to the fact that White has an advantage here. It was human operation that led it there, not raw data power. This is just one example out of millions. The engine is a great tool for analysing, but do not switch off your own brain please :-).
Interesting. Your example reinforces the Negi books are outstanding and repay careful study.
The article makes me wonder if one of you – perhaps Nikos- had any thoughts on computer set ups similar to the recent chessbase web page article on this topic.
A computer may be easily better than any human during a single game, but the mass of all opening research that has been produced over decades by humans and computers together is FAR better than what any single computer can produce.
Let’s be honest: to start talking about computer says 0.00 at move 5 in an absolute mainline or doesn’t like 5 xx (which has been played countless times by professionals) and gives -0.10 is just crass.
Where are the questions regarding endgame technique and plan formulation to win endgames?
James
If you are named Aronian , you can play Bxf6 and crush Ding . More seriously , even the strongest engines miss interesting moves after only 6 plies . My experience is to let Komodo/ or Houdini or SF run at least 30 plies with at least 5 variations (unless forced cases ) and look at the final PVs to challenge the assessment ( which is very frequently wrong if you re-challenge the eval with the engine to move in the final PV )
Another nice exampe is the KID position presented by Nikos in “Thinking inside the Box”. In that position the best move is Nf6-e8, a nobrainer for experienced KID players. However, the state-of-the art engines have big problems to retreat a knight, therefore they do not propose it as first choice. Reason appears to be the hard-wired evaluation functions made by humans (well, grandmasters). Anyway, I wonder what neural networks will do to those rather “mechanical” limitations… What alpha-go and its successors did to the game GO was rather terrifying.
But most of the time, would you not benefit more from an improvement of general skills and technique?@Pinpon
@Frank van Tellingen
It’s the way I use engines , not the way i study chess ?
Hi Jacob. What are some ways to get a training partner? I used to have some but all of them have stopped playing…
@Fizz
Ask around here perhaps?
I will add to the items already said, and especially Jacob’s number 3 in the original response. Anybody that trusts computer analysis numbers are highly mistaking in the game of chess. Computers are excellent, Superb, SPECTACULAR, at finding very long, forced combinations or multi-layer trees of forced combinations to win material or mate the King. However, computers are still to this day bad at opening and endgame evaluation. So much so that they need additional opening books and table bases added to them. I have still seen computers this decade give K+R+N vs K+R as plus 3! It’s either going to be +99 because there is an instant mate for White, or it’s drawn! +3 is the worst possible assessment. Openings are just as bad when it comes to computer evaluation. If everyone used a computer’s numerical evaluation of a position 6 moves into a book line, nobody would ever play the King’s Indian, Grunfeld, or Modern Defense ever again because computers rate White’s position so highly because of the grab of the center. Unlike a static opening like the Exchange QGD, there are lines in these openings where a computer will say +1 with Black to move, and once Black makes the best move, it’s suddenly -2. Computers don’t understand complex openings until it gets into the middlegame where long-winded tactics and deep positional evaluations take over! I will never trust what a computer has to say with in the first 10 to 20 moves unless a player has gone completely out of…
(cont) book and the computer is able to immediately show the tactical bust for the opposing player!
I still play a little correspondence chess where computers are allowed. I often let the machine run two or three programs for 24 or more hours on a position. Often the programs don’t agree. Also, a careful review of positions after only 2 – 3 moves results in ideas the computer did not place in the top 3. That requires running the computers again for 24 hours and so forth. Even after that it is hard to say if the best moves are found. Here is an ongoing game, albeit, we are a number of moves beyond this position that is an example. 1 d4 Nf6 2 c4 e6 3 g3 d5 4 Nf3 dxc4 5 Bg2 b5 6 a4 c6 7 axb5 cxb5 8 Ne5 Nd5 9 0-0 a6 (Departing from Avrukh’s analysis in GMR1A) 10 b3 cxb3 11 Qxb3 f6. Already choosing where to move Ne5 is difficult. The computer really likes White. But, many lines end up in strange material balances where any result is possible. Picking the first computer move in this type of a game will likely lead to a loss. Therefore, while risky, my opponent’s opening choice is astute and a reasonable gamble. Chess is hard. I am looking for opening choices I understand rather than positions that are chaotic and I run the chance of losing to a younger player who can out memorize me. Analyzing positions like those above are fun, the ideas are entertaining, but, they are not really practical in OTB chess.
Patrick wrote: “I have still seen computers this decade give K+R+N vs K+R as plus 3! It’s either going to be +99 because there is an instant mate for White, or it’s drawn! +3 is the worst possible assessment.”
+3 means the computer doesn’t know if it is 0 or +99. That’s still informative, in this case it means the computer doesn’t have access to a tablebase and the decision is beyond its horizon. As a human, I know that means drawn, I don’t care whether the computer also “knows” this.
So why do you use a computer, if you already “know”? Or how do you find out, when you can believe the computer, and when you have to know yourself?
@ Thomas:
I agree. I would argue that if the computer can’t give a definitive assessment because it’s beyond its horizon, it’s just a game and the player who makes the last error loses. I don’t see a difference with computer assessmens in the opening or middle game.
@Thomas
He knows that if the computer says +3, it is drawn. He doesn’t know beforehand if the computer will say +3 or +99.
+3 is always drawn? In any position? Or only in some? Which are these? If I have to know all these by heart, I would never need an engine. What about let’s say B+B vs. N ?
If +3 in some endings is not useful, what about +3 in some openings? And if the computer is only helpful in +99-positions I wouldn’t need one either. I think most of the times I would win such a position myself.
@Thomas
All pawnless endings (including B+B vs N), endings with few pawns that are simple enough that you’d expect the computer to find a real solution (e.g., R+P vs R), no openings.
I have failed to win many +99 positions (e.g. in rook endings), and an engine would certainly have helped me in those situations, but you are likely a stronger player than I am.
I still don’t see the point. If a certain position is beyond the computer’s horizon (which is the case for most positions, whether in the opening, middle game, or endgame), all it can do is assess the position as good as it can. The examples given are trivial, since table bases have solved all postitions below a certain number of men (if I remember correctly 7). Of course a human can assess a certain position as a draw or as a win for either colour (regardless the computer’s verdict), but it all depends on the position. So what’s the point?
@Ray
To me the main point is that you can’t play chess without your brain. The computer is a very useful tool, but you have to know how to use it. You have to judge the position and have to make your own evaluation of the computer’s evaluation – is it helpful or not.
So there’s no way to improve your play just with the computer and without your own hard work. The computer is a tool, not a solution.
I always use engines linked to tablebases. This means there is no problem with five, or even six, man positions.
Sorry I was not available to answer the question put to me. Let me clarify: R+N vs R is sometimes a draw, sometimes a win. As Patrick indicated, 0 or +99. He then stated that +3 is the “worst” assessment. But it’s not even bad! Unless the hardware is very weak, +3 means 0.
In order to know this, you must first know (a) something about R+N vs R, and (b) something about engine evaluations. Patrick clearly knows both of these things. I was pointing out to him that +3 is not misinformation, it is correct from the engine’s point of view, and more than enough from the strong human’s point of view. But if the human knows neither (a) nor (b), then I agree +3 is very unhelpful.
To answer Thomas’s question: The reason I use a computer is because my recognition of the +99 positions, while pretty good, is not perfect. With tablebases, the computer’s recognition is infallible, and without tablebases, still much better than mine. The computer is yet another tool for “falsification”. It’s another way to analyze, post-mortem, my in-game thinking. I could be serenely thinking draw, draw, I know R+N vs R is a draw, and the computer suddenly lights up with +99 — there’s a way to win the R. Should have seen that during the game, now that it is pointed out to me maybe next time I will.
The further comments about computer use are strange to read, because I agree with all the points made, yet somehow the authors are finding a way to disagree with each other.
@An Ordinary Chessplayer
even the most basic knowledge of chess should help you understand +3 =0. You just need to know that a bare knight can’t mate. So the +3 is actually very helpful in that it tells you that it isn’t one of those +99 positions.
Imagine you are playing a cc game with no engine support and you got into or can choose to exchange into a rn vs r endgame. If you really feel an engine evaluation is unhelpful then would you really let your opponent switch his “really weak” engine on and you don’t ? I doubt it.. in other words the evaluation is extremely helpful