Choosing between 1. d4 and 1. Nf3

Hmm... AlphaZ or Hans Berliner... Anyway you know what RJF said...

Rather face English game than any of that Albin, Benko, Budapest, Benoni crap. Some might disagree. 

We dont know statistics of A0 but we have statistics of Leela ' about 60 million games ( Current Leela is much stronger than since Jan 2019 version anyway)

The statistics are like that! 

Against 1. e4 , she found 1.....e5 with Berlin for black in which she find very drawish for white. 

Against 1. d4, she plays 1.....d5 QGD .

1. e4 and 1. d4 have very close winning statistics but she mostly prefer 1. d4

This is the Leela net's move choice which is linked with statistics  of her training games. (She is currently beating 33.5-28.5 against Stockfish 10 dev version in Tcec) 

These are her statistics in first move of all possible first 20 moves.  ( output of policy head + value head statistics)

 1. 54.57 % winning for 1. d4

2. 54.78% winning for 1. e4

20. g4??  27.76 % she thinks the worst first move.

Don't care anymore... what she thinks? Give the Lion Morphy or Capablanca and a good kings gambit? Classic, romantic Chess.

Lion has spoken. 

Anyway, who build these silly super engine things? People. Who move their piece? People. Super engine are so stupid... they cannot switch themselves on?!

My friend, do you have games or Leela's recommendation in symmetrical English, variant 2.Nf3 Nc6 3.Nc3 d6?

I switched from 1.e4 to the more subtle closed games and I must say that delaying d4 catches out more players. So starting with Nf3 and c4 is less cooperative and a real weapon.

Pfrens 2. d3 / 3. d4 is one of my pet lines against f5.

Of course , there are tons of Leela games. 60 millions + games for T30  and 60 million + for T40 in her training server (which is not practically easy to find) 

Other ways are

1. Install Leela engine in your computer

http://blog.lczero.org/2018/09/guide-setting-up-leela-on-chess-gui.html

2. Install leela opening book 

(https://www.chess.com/blog/drmrboss/how-to-use-deeply-computer-analyzed-opening-books)

3. Search Leela games from TCEC . or CCCC

Actually, the Stonewall is playable, as is the classical.  There are a couple of possibilities for the Stonewall:

1.Nf3 f5 2.c4 Nf6 3.g3 (If White doesn't Fiantchetto, the Stonewall is nonsense, and this goes for 1.d4 as well, as Black can Fianchetto his LSB after 3...e6 and 4...b6/5...Bb7) 3...e6 4.Bg2 d5 5.O-O and now:

A) 5...c6 6.d3 Be7 (stronger than 6...Bd6 in this case), though I don't have the same opinion of this line as other Dutch advocates do)

B) 5...Bd6! (the line I definitely agree with) 6.Nc3 (6.d3 dxc4 7.dxc4 O-O is likely to transpose) O-O 7.d3 dxc4! 8.dxc4 Nc6, going for ...e5.

Black can also play the classical with e6 and d6.

Not much White can do about that... if somebody is determined to play KID or dutch... so they will.

Indeed, that's what Kramnik said. I struggle against Stonewall regularly.

Many thanks for the tips! It will help me a lot, especially your article about book openings.

About Lc0's victory at TCEC, I saw your comment on another topic saying that Leela has overtaken her "father" Alphazero. That's right? Lc0 is already stronger than A0?

Play c4 when responding to c5 and later play d4

e4 all the way. d4 is kind of boring. I only use it against someone if I know they will play the sicilian.

I hate the Sicilian.

True. My repertoire:

KID (main)

Grunfeld 

Nimzo/Queen's Indian

Semi-Slav

QGA

Leningrad Dutch

You might want to consider narrowing your repertoire.

As the expression goes: "Jack of all trades, master of none."

Also, regarding the OP's question:

I like 1.Nf3 because I find it more flexible.

One doesn't have to play a Queen's Gambit setup with 1.Nf3. You could adjust, depending on what black plays.

The d-pawn doesn't have go to d4. Likewise, the c-pawn doesn't have to go to c4. (Some players seem to think that d4+c4 must be played at all times.)

With 1.Nf3, white can go into other dynamic structures, like the KIA, the English, or the Colle-Zuckertort ... setups that might be less familiar for your opponent.

Sometimes it's better to stray off the beaten path ... to pull your opponent out of their main prep and into yours.

Pros of 1.d4 

- Easy transposition to a London or to a QG or to a Blackmar-Diemer or to an Accelerated London

- Puts a pawn in the centre

- Opens diagonal for White's DSB

Cons of 1.d4 

-  Easily countered with 1.c5 and if 2.c4 then 2.d5 and it's now a QGD : Austrian Variation which cannot be transposed out of in any favourable way for White

- Easily countered by delaying d7-d5 forever 

- Easily transposed away from a London even after 1.c6!

- Very little if no flexibility of transpositions after 1.d4! so you're opponents can lock you in their prep and close all the doors 

Pros of 1.Nf3 

- Easy transposition to an Italian , a London , a Reti Gambit , a Tennison Gambit , a Reti Opening : Old Indian Attack and to a QGD of some nature maybe even to an Sicilian Defence  : French Variation and even transposes to a Slav Defence : Modern , Bonet Gambit  

- Keeps transpositional avenues and parades open 

- More flexibility regarding opening choice

- Doesn't commit center pawns early leaving the doors open 

- Harder to get stuck in your opponents prep

Cons of 1.Nf3 

- 1.e5! the Ross Gambit can be used a neutralisation transpositional avenue for Black 

- Most transpositions can be refuted if not played correctly

- Need to be careful and engarde based on what opponent plays 

After all this , I'd say 1.Nf3! is better than 1.d4!

For example, d4 makes f5 really strong. So how to play the Dutch against Nf3 and c4? White has always d3+e4.