The Busted Dutch

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
ouachita
Posts: 448
Joined: Tue Jan 15, 2013 3:33 pm
Location: Ritz-Carlton, NYC
Full name: Bobby Johnson

Re: The Busted Dutch

Post by ouachita » Sat Sep 14, 2019 1:43 am

Ovyron wrote:
Fri Sep 13, 2019 9:12 pm
zullil wrote:
Fri Sep 13, 2019 10:41 am
"Busted" means that you can draw as White? :shock:
I also have a line that equalizes against 2.Bf4
What is it? Nf6?
SIM, PhD, MBA, PE

Uri Blass
Posts: 8530
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: The Busted Dutch

Post by Uri Blass » Sat Sep 14, 2019 1:52 am

Ovyron wrote:
Fri Sep 13, 2019 10:30 pm
Heh, a 0.00 score.

I actually went and tried my mainline against ChessDB and I was quite impressed about the wide coverage it already had of it. It also showed a 0.00 score. Eventually we were at a "this position doesn't exist, click here to request analysis" thing, an after 5 seconds it was still showing 0.00. But I pushed on, and sure enough, eventually I hit on a move it was missing, and it was very satisfying to see the score jump from 0 to -185. By then it had already adjusted its scores, but it still suggested a losing move on a previous position with a score of 0...

So we will see, I have no problems beating people that just let Stockfish reach high depth and play their move against me, but my opponent in this game isn't like that, so I expect he'll play already better than Stockfish MultiPV=3 at depth 52...

I do not see information in the link that you give about depth of analysis.
What is the source of your multipv=3 at depth 52?

I do not see written depth of analysis for every position.

User avatar
Ovyron
Posts: 2333
Joined: Tue Jul 03, 2007 2:30 am

Re: The Busted Dutch

Post by Ovyron » Sat Sep 14, 2019 2:13 am

ouachita wrote:
Sat Sep 14, 2019 1:43 am
What is it? Nf6?
Hehe, if you want to know you'll have to play against me. I promise to play the Dutch even if it means I could never beat you... (at least I'm an IM now, so it wouldn't be a walk in the park...)
Uri Blass wrote:
Sat Sep 14, 2019 1:52 am
What is the source of your multipv=3 at depth 52?
zullil analyzed that position like that, and got 0.00 from Stockfish, at that depth, with that MultiPV. It's funny that I've been looking at those positions since August 21 yet zullil can come and refute my big tree of analysis with a few hours of unassisted output :mrgreen:
Make someone happy today.

User avatar
MikeGL
Posts: 873
Joined: Thu Sep 01, 2011 12:49 pm

Re: The Busted Dutch

Post by MikeGL » Sat Sep 14, 2019 4:11 am

My opinion on this 2.d5 against Dutch is that this move allows black to equalize quickly.
Maybe time to call that 2.d5 move Tsvetkov Attack. But seriously, it was not yet played by very strong players in tournament conditions because it is counterintuitive.

Reason why 2.d5 wasn't played, it violates one of those golden fundamentals in openning.
1. Don't play your Q early in the game.
2. You must develop pieces as quickly as possible (give up pawns [gambit], if need be).
3. Don't push the same pawn twice. (unless this pawn push gains tempo in its previous move or the succeeding move, or is currently attacked and needs to advance)
etc.
etc.

Perhaps those fundamentals were felt by engines on deeper analysis hence considers 2.d5 a bit inferior compared to other alternatives on that specific position.
Claiming that it is a strong move is debatable, but it just tossed away the initiative for white IMO.
I told my wife that a husband is like a fine wine; he gets better with age. The next day, she locked me in the cellar.

User avatar
Ovyron
Posts: 2333
Joined: Tue Jul 03, 2007 2:30 am

Re: The Busted Dutch

Post by Ovyron » Sat Sep 14, 2019 4:46 am

Thanks MikeGL. So where do you rank the move? I claimed that there's at least 16 moves better than 2.d5, someone else claimed that there's at least 20 moves better than 2.d5. Depending on your view the move might be worse than what everyone expected, instead of being anywhere close to best :o
Make someone happy today.

Uri Blass
Posts: 8530
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: The Busted Dutch

Post by Uri Blass » Sat Sep 14, 2019 7:19 am

latest stockfish with default parameters does not see d5 as one of the top 20 moves at depth 30,31:

Here are the evaluations

1)c4 0.90/30 0.79/31 0.85/32
2)b3 0.89/30 0.78/31 0.84/32
3)Bg5 0.86/30 0.99/31 0.84/32
4)g3 0.84/30 0.85/31 0.74/32
5)e3 0.8/30 0.65/31 0.74/32
6)Nf3 0.76/30 0.87/31 0.81/32
7)c3 0.71/30 0.71/31 0.61/32
8)Nh3 0.7/30 0.65/31
9)Bf4 0.68/30 0.64/31
10)Qd3 0.63/30 0.63/31
11)a4 0.61/30 0.58/31
12)Nc3 0.61/30 0.63/31
13)h4 0.59/30 0.67/31
14)b4 0.57/30 0.44/31
15)a3 0.56/30 0.68/31
16)Nd2 0.53/30 0.55/31
17)h3 0.45/30 0.63/31
18)Be3 0.42/30 0.71/31
19)Na3 0.4/30 0.54/31
20)Bd2 0.36/30 0.69/31

User avatar
Ovyron
Posts: 2333
Joined: Tue Jul 03, 2007 2:30 am

Re: The Busted Dutch

Post by Ovyron » Sat Sep 14, 2019 9:51 am

Well, I'll go on record saying that latest Stockfish is insane if it really thinks 2.c4 is 0.85. Maybe Larry is right about some weird scaling happening now.

For contrast here's private Stockfish with learning up to some depth 33. The thing with this version is that, if 2.d5 is best, I can show it the variations and it will learn them, and then 2.d5 will remain at the top, even if I unload the engine and restart my computer. It also uses learning that it has from the positions I've already fed to it, and the price is that it shows really short PVs...

1) 33 86:00 +0.46 2.g3 d6 (11.413.083.882) 2211
1) 33 86:00 +0.46 2.Nf3 Nf6 (11.413.083.882) 2211
3) 33 86:00 +0.43 2.Nh3 Nf6 (11.413.083.882) 2211
4) 33 86:00 +0.37 2.Bg5 (11.413.083.882) 2211
4) 33 86:00 +0.37 2.c4 d6 (11.413.083.882) 2211
6) 33 86:00 +0.36 2.b3 b6 (11.413.083.882) 2211
7) 33 86:00 +0.35 2.e3 g6 (11.413.083.882) 2211
7) 33 86:00 +0.35 2.Bf4 (11.413.083.882) 2211
9) 33 86:00 +0.32 2.h4 d6 (11.413.083.882) 2211
10) 32 86:00 +0.28 2.Nc3 Nf6 (11.413.083.882) 2211
11) 32 86:00 +0.25 2.Nd2 Nf6 3.Ngf3 b6 4.g3 g6 5.Bg2 Bb7 6.O-O Bg7 7.c4 O-O 8.b4 e6 9.Bb2 Nc6 10.a3 Ne7 (11.413.083.882) 2211
12) 32 86:00 +0.24 2.Qd3 e6 3.c4 Nf6 4.Nc3 b6 (11.413.083.882) 2211
13) 32 86:00 +0.23 2.h3 e6 (11.413.083.882) 2211
13) 32 86:00 +0.23 2.c3 Nf6 3.g3 (11.413.083.882) 2211
15) 32 86:00 +0.22 2.a3 Nf6 3.Bf4 e6 4.e3 Be7 (11.413.083.882) 2211
16) 32 86:00 +0.11 2.a4 Nf6 (11.413.083.882) 2211
16) 32 86:00 +0.11 2.d5 e6 3.Nf3 Nf6 4.dxe6 d5 5.g3 Bxe6 6.Bg2 Nc6 7.O-O Ne4 8.Nc3 Be7 9.h4 a6 (11.413.083.882) 2211
16) 32 86:00 +0.11 2.Bd2 Nf6 3.c4 d6 4.Nf3 g6 5.g3 Bg7 6.Bg2 O-O 7.Nc3 Na6 8.O-O (11.413.083.882) 2211
19) 32 86:00 +0.06 2.b4 e6 3.a3 b6 4.c4 Nf6 5.Nf3 c5 6.bxc5 bxc5 7.e3 Be7 8.Bd3 O-O 9.Nc3 Nc6 (11.413.083.882) 2211
20) 32 86:00 +0.05 2.Be3 Nf6 3.c3 b6 4.Nf3 Bb7 5.Bf4 e6 6.e3 (11.413.083.882) 2211
21) 32 86:00 0.00 2.f3 e6 (11.413.083.882) 2211
22) 32 86:00 0.00 2.e4 fxe4 3.Nc3 Nf6 4.f3 d5 5.fxe4 dxe4 6.Bg5 Nc6 7.Bb5 Qd6 8.Nge2 a6 9.Bxc6+ Qxc6 10.d5 Qb6 11.Qd4 Qxb2 12.Rb1 Qxc2 13.Rd1 Qb2 14.Rb1 Qc2 (11.413.083.882) 2211
23) 32 86:00 0.00 2.Na3 e6 (11.413.083.882) 2211
24) 32 86:00 -0.01 2.g4 fxg4 (11.413.083.882) 2211
25) 32 86:00 -0.04 2.Qd2 Nf6 3.g3 d5 4.Bg2 Ne4 5.Qd1 c5 6.e3 g6 7.Ne2 Nc6 8.O-O Bg7 9.f3 Nf6 10.b3 e5 11.dxc5 O-O 12.Nbc3 (11.413.083.882) 2211
26) 32 86:00 -0.07 2.f4 e6 3.c4 Nf6 4.Nf3 Be7 5.g3 d5 6.Nc3 O-O 7.e3 b6 8.Ne5 Bb7 9.Bg2 Nc6 10.O-O Ne4 11.Qc2 Nxe5 12.dxe5 Nxc3 13.Qxc3 c5 14.cxd5 Bxd5 15.Rd1 (11.413.083.882) 2211
27) 32 86:00 -0.54 2.Kd2 e6 3.Ke1 c5 4.e3 Nf6 5.h4 Nc6 6.Nf3 d5 7.c4 cxd4 8.exd4 Bd6 9.Nc3 O-O 10.Be2 (11.413.083.882) 2211

The rest of the moves it claims are losing, so I didn't waste time on them (but it says 2.Bh6?? is the best move that loses.)

Of note:

My move that I claim is best appears at top 4. I claim I can show the refuting moves so that the rest and 2.c4 score 0.36 or less.

Nh3 is surprising, I hadn't seen it before, but appears as third best. On Uri's list this was at top 8.

ouachita's Bf4 appears as top 7 on this list, while it appears as top 9 on the other one.

Finally, 2.d5 comes back with a vengeance, now appearing as top 16 again, tied in score with 2.Bd2 and 2.a4.

Conclusion: Either 2.d5 being best is nonsense, or it's quite the most brilliant chess move ever found.
Make someone happy today.

zullil
Posts: 5556
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: The Busted Dutch

Post by zullil » Sat Sep 14, 2019 12:17 pm

Ovyron wrote:
Fri Sep 13, 2019 10:30 pm
Heh, a 0.00 score.

So we will see, I have no problems beating people that just let Stockfish reach high depth and play their move against me, but my opponent in this game isn't like that, so I expect he'll play already better than Stockfish MultiPV=3 at depth 52...
This is an intriguing claim, that, if true, might reveal certain weaknesses in Stockfish's current search (for example, that it's too narrow, due to too much reduction/pruning). Since your are already playing a current open online game, would you be interested in "forking" that game to a second one, played online here? I'll supply moves for Black, and I'll do nothing more than run Stockfish-dev with MultiPV=1 and 6-man tables. No human assistance of any kind, no forward-backward investigation, etc. Just 20 threads and 64 GB hash.

We can start from here, if you are interested, which Stockfish-dev says is basically even:


OneTrickPony
Posts: 90
Joined: Mon Apr 29, 2013 11:29 pm

Re: The Busted Dutch

Post by OneTrickPony » Sat Sep 14, 2019 2:02 pm

I wish there was more discussion about specific lines. Right now it seems the only way is to play correspondence and hope someone plays an interesting variation vs you.
I will try though: I think 2.Bf4 rules out g6 setups as white gets strong initiative there. The alternatives are either Stonewall (which seems very comfortable for white without g3 played and d6/e6 setup. Anyone interested in some discussion trying to prove equality for black in either of those?

ouachita
Posts: 448
Joined: Tue Jan 15, 2013 3:33 pm
Location: Ritz-Carlton, NYC
Full name: Bobby Johnson

Re: The Busted Dutch

Post by ouachita » Sat Sep 14, 2019 4:02 pm

Ovyron wrote:
Sat Sep 14, 2019 2:13 am
ouachita wrote:
Sat Sep 14, 2019 1:43 am
What is it? Nf6?
Hehe, if you want to know you'll have to play against me.
Then, let's play. Where, when?
SIM, PhD, MBA, PE

Post Reply