Is e4 significantly better than d4?

[OneTrickPony]

Black has many very solid options after 1.d4. In Ragozin it's really tough to get anything. The battle ground seems to be the endgame after: cxd5/Bg5 where black goes h6/0-0/Bf5. The resulting endgame is better for white and unpleasant to play for humans as black but it's hard to imagine it's not a relatively easy draw for top engines.
In Be7 QGD the Bf4, Nbd7, c5 system seems to be critical. Again white is better but black seem to be able to defend by going to slightly worse endgames which as above are unpleasant for humans but likely an easy draw for engines.
There is also semi-tarrash, Vienna (just go Nf6 in the gambit line) and semi-slav (the most complicated one and even though it's hard to find an edge, white has many promising tries.

If white can get more in those than it gets in Berlin or anti-Marshall or d3 Spanish (maybe better option than the main line) or Italian is an open question. My view is that it's probably easier to make an opening book which will never lose even with current software/hardware against 1.d4 than it is against 1.e4. On the other hand I like 1.d4 more for human play because black gets little to no counterplay in the critical lines while in 1.e4 if you want to get something you need to allow pretty complicated position.

Btw, I think 1.e4 book with up to date theory written by a very competent (in both chess and analysis) player is long overdue. The market is flooded with hacks. I don't buy chess books anymore (Leela + 2080ti produces much better results in a few minutes than even the most respected books) but I will make an exception for this one (as long as I can somehow buy an electronic version). Best of luck Larry, I am curious to see some of the recommendations mainly against serious but not the very main openings (Caro/French/Janish Gambit if you're recommending Spanish).

[Ovyron]

My view is that it's probably easier to make an opening book which will never lose even with current software/hardware against 1.d4 than it is against 1.e4.

But that contradicts your previous statement. From the get-go you don't even know what moves would you use to neutralize 1.d4. The first question is if you should play 1...Nf6 or 1...d5. This is a really though one, and it comes down to... if you only care about drawing (then 1...Nf6 is best) or if you'd also want to win as black if white is careless (then 1...d5 is best.) It is hard to make an opening book to neutralize 1.d4 because there's just so many options you'd not know where to start or what variation to choose or what's the best for this.

With 1.e4 it is easy. You can pretend the Sicilian doesn't exist, because the material needed in the Sicilian (in number of refuted lines) may be bigger than for 1.d4 and the Spanish put together ( - even just the Smith–Morra Gambit is scary to build for!)

Then you always play 1...e5, and against the Spanish you play the Marshall-Counter Attack, or you have to refute d3 and a4 and that's it. People have already built books that refute the Italian in this manner (play Bc5 before Nf6 because 4. Ng5 is a nightmare), though the lines remain private and uncorked, at least people have shown the concept works.

If neutralizing 1.d4 was easier we could talk about a "mainline" (its equivalent to the Marshall-Counter Attack) and a few other variations that would need to be refuted in case white doesn't follow it. Maybe it exists, and once we know which one is it it turns out it's easier to refute 1.d4 than 1.e4. But for all I know 1.c4 is already better than 1.e4, at the games at the top (best hardware and years per game time control) where people don't want their games to just be a theoretical battle that comes out of book in a drawish position, they even play it instead of 1.d4.

[OneTrickPony]

Well, there are many options after 1.d4 but if you're making a book with the goal to not lose you just choose one. For example you can play the Ragozin and focus on drawing the critical endgame having very deep lines there. Against 1.e4 you get 3 major options white can choose to begin with: Italian, anti-Marshall, d3 Spanish and it's more difficult to simplify the position n in those even if those positions have lower evals than the Ragozin endgame.
Trying to get winning chances against 1.d4 is imo hopeless and no time should be spent on it if we are talking about computer assisted play.

I disagree about refuting the Italian. I've seen the claims, I haven't seen anything convincing. Super GMs seem to not agree and they have access to pretty serious hardware. I also don't agree that 4.Ng5 is a nightmare. I think it's an easy draw at least and you get chances to win as well if you want that.
As to Marshall a4 is not easy to refute, the endgames arising from that are imo more difficult to hold than stuff you get from Berlin or Bf4 QGD.

As to 1.c4, currently it looks like bleak for white in d5/Bc5 line. I am sure it's not the end of the story but the ball is definitely in white's court there.

I guess that's why chess is still interesting at least below correspondence with long time controls level. You never know who understands things better and whose methods are more efficient. I don't play cc and I don't (yet) have access to really serious hardware but I am decently strong otb (FM, haven't played for 20 years, weak GM level rating in online blitz). I also do engine development for living (not chess though) so I think I am not completely clueless. I may be underestimating people who like to post wild claims (like Italian being refuted) though.

[Ovyron]

I guess that's why chess is still interesting at least below correspondence with long time controls level.

All my claims have been at correspondence time controls. I don't find anything else interesting, because top players can defeat ANYTHING below this level. Like, it doesn't make sense about what GM humans play, because anyone with a computer and Stockfish-dev can produce moves that beat them, from any variation (even at Knight odds...), but I want to know the chess truth, and in this case answering the question "Is e4 significantly better than d4?" should refer to the strongest level we can play currently (we can't play perfectly, which makes correspondence time controls interesting still, you've gotta do your best to play the best moves against someone and lose a game to know why.)

>Trying to get winning chances against 1.d4 is imo hopeless and no time should be spent on it if we are talking about computer assisted play.

With people being really poor at assisting computers this still can be worthwhile. Then again, after you see the moves played by those people it might turn out that you could have beaten them with anything anyway. So perhaps you're right, and in this case, 1...Nf6 is best for a draw, and we prune the entire Queen's Gambit from it (whatever defense black chooses, it should not include d5). I also think that there's a big chunk of dangerous openings that are very difficult, because white can come up with new ways to attack black and who knows how big the tree on them could go, so I suggest the defense black chooses also doesn't include g6 and the fianchetto 'd Bishop.

>I disagree about refuting the Italian. I've seen the claims, I haven't seen anything convincing.

I have. In the critical variations engines show some 0.50 advantage to white but it's black with more chances of winning (more than white in the opening position.) The thing is that it's the kind of position where if black plays correctly it can keep its edge until engines finally agree that it does (sometimes too late, so that black has already won), but where playing correctly as black is very hard, it requires at least some 1 year/game time controls, so while the edge may be enough to win games against engines at blitz because they misplay the white side by itself, it may hold no value for humans, they'd probably get outplayed by white, because knowing what to do as black resembles those +100 move mates in 7men tablebase positions, where some Bishop move is a win, and another is drawn, but no human can discern why.

> I also don't agree that 4.Ng5 is a nightmare.

It is in the sense that the tree required to equalize against white would be bigger than the one against Bc5! I don't even know if it'd be bigger than the tree required to equalize against the whole Spanish, at least it might be if we truncate the Marshall Counter-Gambit tree. Spending 60% of your time of refuting 1.e4 on 4.Ng5 sounds like a nightmare, so just playing Bc5 before Nf6 sounds like the best proposition.

>As to Marshall a4 is not easy to refute, the endgames arising from that are imo more difficult to hold than stuff you get from Berlin or Bf4 QGD.

I think I got close to refuting a4 back in the times of Critter 1.6 that had a Session File, at least the analysis held against Houdini 6. I now realized this may be a personal thing, with my database and knowledge of the variations, it'd be easier for me to build a draw-book for 1.e4 than for 1.d4, but that doesn't mean it's true for everyone.

>As to 1.c4, currently it looks like bleak for white in d5/Bc5 line. I am sure it's not the end of the story but the ball is definitely in white's court there.

Right, I guess 1.c4 is just a move to take advantage of people that don't have their theory straightened out. The games I've won with it have been against people that didn't play main theory...

> I don't play cc and I don't (yet) have access to really serious hardware

But you don't need to have really serious hardware to contend nowadays, engines are strong enough that as long as you have at least some 4core and a good understanding of the openings you should be playing, you'd have bridged the difference in hardware. Now, more than ever the methods you use to analyze moves and select them after analysis is what's most important.

To start and check your level I suggest you check what is the highest Stockfish depth that you can beat. Beating Stockfish Depth 1 is trivially easy. Beating Stockfish Depth 2 is slightly harder. Eventually you'll find some depth that you'll struggle to beat, at this point the openings you choose become significant, and you'll probably find it'll be easier to beat Stockfish with 1.d4 than with 1.e4. But the point is, as you become good at beating higher depths, becoming good at beating other people at corr chess comes naturally, because the shocking surprise is, that no matter how highly rated they are, or what titles they have, chances are high that they'll play worse than this Stockfish depth you surpassed, so if you can beat it, you can beat them.

[Laskos]

I don't know, I still seem to trust more Leela than even Larry. I have zero opening theory knowledge, but I am a bit skeptical when people come with some esoteric to me knowledge, while in the openings Leela already at 1 node is positionally probably GM level. Sorry, I think I will stick with my simplistic approach using Leela with the best available net at long analyses.

I went through proposed by Leela line ply by ply to 100 million nodes each ply, and remarkably e4 line was followed all the way to the 8 moves I showed in the diagram in the first post. The same thing, remarkably, happened with the d4 line, this time to 6 moves.

The positions in diagrams do occur in many recent games of top GMs. I let analyze these e4 and d4 lines at the endposition for more than 100 million nodes each. The net is the best JHorthos bignet. Crucially, these trends are hardly visible at say 10 million nodes, the scores stabilize after more than 50-100 million nodes.

Here is the e4 line analyzed after 8 moves, previously checked ply by ply 8-mover, now analyzed to 100+ million nodes:

[d]r1bk1b1r/ppp2ppp/2p5/4Pn2/8/5N2/PPP2PPP/RNB2RK1 w - - 0 9


info depth 27 seldepth 72 time 6867298 nodes 123995548 score cp 5471 hashfull 10
00 nps 18055 tbhits 0 multipv 1 pv b1c3 d8e8 h2h3 h7h5 c1f4 f8b4 c3e2 c8e6 e2d4
f5d4 f3d4 e6d5 a1c1 a8d8 f1d1 b4e7 c2c4 d5e4 f2f3 e7c5 f4e3 e4g6 g2g4 h5g4 h3g4
f7f6 e5f6 g7f6 g1g2 e8f7 d4f5 d8d1 c1d1 g6f5 e3c5 f5e6 g2g3 h8g8 g3f4 a7a5 a2a4
g8e8 d1h1 f7g6 h1e1 e6f7 e1e8 f7e8 c5e7 e8f7 b2b3 b7b5 b3b4 a5b4 a4a5 b4b3 e7a3
f7c4 a5a6 c6c5



Here is the d4 line analyzed after 6 moves, previously checked ply by ply 6-mover, now analyzed to 100+ million nodes:

[d]rnbq1rk1/ppp1bppp/4pn2/3p4/2PP4/5NP1/PP1BPPBP/RN1QK2R w KQ d6 0 7


info depth 25 seldepth 73 time 12398405 nodes 140132643 score cp 5306 hashfull 1
000 nps 11302 tbhits 0 multipv 1 pv d1c2 b8d7 e1g1 c7c6 d2f4 b7b6 b1d2 f6h5 e2e3
c8a6 f1c1 a6b7 c4d5 e6d5 h2h4 c6c5 d2b1 h5f4 e3f4 c5d4 f3d4 a8c8 b1c3 d7f6 c1d1
f8e8 a2a3 g7g6 d1d2 e7c5 a1d1 d8e7 c2b3 f6e4 d2e2 e7f6 c3d5 b7d5 b3d5


=========================================

e4 best line performs 50% better than d4 best line. These are not weird openings.

With e4 line, if it so bad an opening for Black, why very recently it was played as Black by super GMs from Nakamura to Carlsen, Topalov, Karjakin, Anand, etc.? It is a pretty common opening for top GMs, both as White and Black.

With d4 line, if it so bad an opening for White, why very recently it was played as White by super GMs from Nakamura to Carlsen, Kramnik, Gelfand, Aronian, etc.? It is not an uncommon opening for top GMs, both as White and Black.

========================

I know nothing of opening theory, and sorry, I sort of love Leela in the openings, I am very biased trusting it .

[OneTrickPony]

You don't need to trust the tactical problem thing. Just play QGD with white and see how the eval changes once Leela plays the Vienna and realized it can't take two pawns once it's close to that point (Leela won't see 5.e4 as a good move there until you force it and then even when you show her Bxc4 after Bb4 it takes her some time to see that Nxe4 0-0 followed by Bxc3 is very bad once it sees it it goes for 7...Nf6 lines and those produce higher eval for white by the Berlin endgame by significant margin).
The whole analysis from starting point is based on tactical oversight which is easy to verify with Leela itself.
I love running Leela as well. I've run it for many hours until 32gb of RAM run out and I've gone deep in many Spanish and QGD lines
The Berlin endgame try is refuted by Be7/Nh4 plan (instead of Kd8-e8 in your line). It equalizes according to Leela but she can't see it by herself until you actually make the moves with black. It only sees it deep into the position. Again the reason is that it can't understand the tactics (black gives up c7 pawn but organizes Be6/Kf7/Rag8/f5 counterplay which leads to mass exchanges and a draw). Even if you give it 2 hours per move on 2080ti and play the equalizing plan with black you still end up with 0.00.
These days Berlin endgame is played by top GMs mainly as "let's check the prep and if they remember then it's a draw" kind of weapon. If white wants to win it's almost always d3 (or sporadically Re1).

[lkaufman]

I don't know, I still seem to trust more Leela than even Larry. I have zero opening theory knowledge, but I am a bit skeptical when people come with some esoteric to me knowledge, while in the openings Leela already at 1 node is positionally probably GM level. Sorry, I think I will stick with my simplistic approach using Leela with the best available net at long analyses.

I went through proposed by Leela line ply by ply to 100 million nodes each ply, and remarkably e4 line was followed all the way to the 8 moves I showed in the diagram in the first post. The same thing, remarkably, happened with the d4 line, this time to 6 moves.

The positions in diagrams do occur in many recent games of top GMs. I let analyze these e4 and d4 lines at the endposition for more than 100 million nodes each. The net is the best JHorthos bignet. Crucially, these trends are hardly visible at say 10 million nodes, the scores stabilize after more than 50-100 million nodes.

Here is the e4 line analyzed after 8 moves, previously checked ply by ply 8-mover, now analyzed to 100+ million nodes:

[d]r1bk1b1r/ppp2ppp/2p5/4Pn2/8/5N2/PPP2PPP/RNB2RK1 w - - 0 9


info depth 27 seldepth 72 time 6867298 nodes 123995548 score cp 5471 hashfull 10
00 nps 18055 tbhits 0 multipv 1 pv b1c3 d8e8 h2h3 h7h5 c1f4 f8b4 c3e2 c8e6 e2d4
f5d4 f3d4 e6d5 a1c1 a8d8 f1d1 b4e7 c2c4 d5e4 f2f3 e7c5 f4e3 e4g6 g2g4 h5g4 h3g4
f7f6 e5f6 g7f6 g1g2 e8f7 d4f5 d8d1 c1d1 g6f5 e3c5 f5e6 g2g3 h8g8 g3f4 a7a5 a2a4
g8e8 d1h1 f7g6 h1e1 e6f7 e1e8 f7e8 c5e7 e8f7 b2b3 b7b5 b3b4 a5b4 a4a5 b4b3 e7a3
f7c4 a5a6 c6c5



Here is the d4 line analyzed after 6 moves, previously checked ply by ply 6-mover, now analyzed to 100+ million nodes:

[d]rnbq1rk1/ppp1bppp/4pn2/3p4/2PP4/5NP1/PP1BPPBP/RN1QK2R w KQ d6 0 7


info depth 25 seldepth 73 time 12398405 nodes 140132643 score cp 5306 hashfull 1
000 nps 11302 tbhits 0 multipv 1 pv d1c2 b8d7 e1g1 c7c6 d2f4 b7b6 b1d2 f6h5 e2e3
c8a6 f1c1 a6b7 c4d5 e6d5 h2h4 c6c5 d2b1 h5f4 e3f4 c5d4 f3d4 a8c8 b1c3 d7f6 c1d1
f8e8 a2a3 g7g6 d1d2 e7c5 a1d1 d8e7 c2b3 f6e4 d2e2 e7f6 c3d5 b7d5 b3d5


=========================================

e4 best line performs 50% better than d4 best line. These are not weird openings.

With e4 line, if it so bad an opening for Black, why very recently it was played as Black by super GMs from Nakamura to Carlsen, Topalov, Karjakin, Anand, etc.? It is a pretty common opening for top GMs, both as White and Black.

With d4 line, if it so bad an opening for White, why very recently it was played as White by super GMs from Nakamura to Carlsen, Kramnik, Gelfand, Aronian, etc.? It is not an uncommon opening for top GMs, both as White and Black.

========================

I know nothing of opening theory, and sorry, I sort of love Leela in the openings, I am very biased trusting it .

Those are both indeed very good openings for both sides. The problem is that with deep analysis, Leela rejects the g3 move in the d4 opening. I realize now that you probably don't know what the VIenna variation (for example) is, so I suggest that you have Leela analyze the position after 1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3 dxc4 5.e4 Bb4 6.Bxc4 Nxe4 7.0-0 to a very high depth. At first it looks like White has given away one pawn and now a second one for insufficient compensation, but gradually Leela should realize that it's not good to win the second pawn and that there is more than enough compensation for the first one. So the White score should rise to above "par". Then if you step back one ply at a time, looking for better moves for each side, you will probably find that Black can improve on his fourth move, and that this becomes the main line of 1.d4. The score will be very close to what 1.e4 should show if you start with your position after 8 moves and work backward in the same way. The problem is that to see this Vienna variation (with proper eval) given above from the root with Leela's current search might take a trillion nodes or more. Leela's eval is amazingly good, but its search can probably be improved a lot in the future.

[todd]

I don't know, I still seem to trust more Leela than even Larry. I have zero opening theory knowledge, but I am a bit skeptical when people come with some esoteric to me knowledge, while in the openings Leela already at 1 node is positionally probably GM level. Sorry, I think I will stick with my simplistic approach using Leela with the best available net at long analyses.

You are correct to do this in a new position that has never been examined before by computer-assisted humans.

However, when it comes to the chess starting position, humans are not going merely by their intuitions. For the sake of argument, we can even decide to place absolutely zero value on human ability to evaluate positions. We'll stick solely to computer-obtained results.

We have access to extremely deep computer analysis of many leaf positions and results from computer-assisted games where these moves were tried. From there we can work backwards to the starting position to get an idea about what should be played.

The depth of this analysis completely dwarfs any result obtained by running Leela on the start position.

If you would like specific improvements for white on Leela's main lines with 1. c4 and 1. d4 that will result in Leela's evaluations for white increasing for after those moves are played, I can provide them, as I do a lot of analysis work with Leela and SF already and have already worked out the lines. Actually, I see that OneTrickPony and Larry have already provided the lines for 1. d4. I'll dig up the 1. c4 line if anyone's interested.

[OneTrickPony]

If you would like specific improvements for white on Leela's main lines with 1. c4 and 1. d4 that will result in Leela's evaluations for white increasing for after those moves are played, I can provide them

Is there anything you can offer against 1.c4 e5 with the idea: 2.Nc3 Nf6 3.Nf3 Nc6 4.g3 d5 5.cxd5 Nxd5 6.Bg2 Bc5?
I think Leela wants 4.e4 intead of 4.g3 but it's hard to get anything (even Leela eval wise) after 4...Bb4 5.d3 d6 etc.
I was unable to get anything above +0.2 with Leela after 1.c4 e5. It's just equality everywhere.

[todd]

If you would like specific improvements for white on Leela's main lines with 1. c4 and 1. d4 that will result in Leela's evaluations for white increasing for after those moves are played, I can provide them

Is there anything you can offer against 1.c4 e5 with the idea: 2.Nc3 Nf6 3.Nf3 Nc6 4.g3 d5 5.cxd5 Nxd5 6.Bg2 Bc5?
I think Leela wants 4.e4 intead of 4.g3 but it's hard to get anything (even Leela eval wise) after 4...Bb4 5.d3 d6 etc.
I was unable to get anything above +0.2 with Leela after 1.c4 e5. It's just equality everywhere.

I don't have anything for white in that 6...Bc5 line. White can sidestep it with either 2. g3 or 3. g3, and then if black insists on a reverse Dragon then he'll have to play the Nb6 line instead of the Bc5 line. But of course 2. g3 and 3. g3 allow black some good non-Dragon lines that aren't possible after 2. Nc3 Nf6 3. Nf3.

I do have an improvement for white in the line Leela goes for from the start position (posted earlier in this thread) - 1. c4 e5 2. g3 Nf6 3. Bg2 Bc5 and Leela thinks white should play 4. d3 here, because it misses a deep tactical line in one of the Nf3 e4 variations. I don't remember how it goes off the top of my head.

Basically, I think black equalizes everywhere in chess. I'm just trying to show that one can get in trouble by following the lines Leela recommends in the starting position, and therefore simply letting Leela run in the start position as Laskos did isn't the best way to get chess truth. It's a decent start, and it's interesting, but we've gone much, much deeper. You & Larry showed why not to trust Leela's 1. d4 mainline, and similarly its c4 mainline isn't trustworthy. The Berlin endgame is a pretty reasonable 1. e4 mainline, though.