Novelty discovered in the Berlin Defense

Life got a bit more interesting today, with a 7th-move novelty just invented.

Now, if you were Black, fighting Stockfish 10 in this position, what kind of move would you be anticipating? Mainline 7.d4, surely, is an option. 

What follows here, though, is Stockfish's second-best offering, and it was bizarre enough to be worth a real look--

It's interesting....I don't know if it will ever get played, but you defnitely do get some interesting positions from this sacrifice

Working on figuring what some kind of "mainline" idea would be here. My Stockfish prefers declining the gambit so far, currently sitting at a search depth of 32+ on position after 7. b4.

According to SF, Black's best bet at getting some kind of advantage is 7 ... a6.

From a position like that I'd say Black has the easier game.

If Black accepts the gambit? Well, things do look rather muddy.

Updating lines a bit as we go along. Some interesting tactics if White plays 9. Nc3 in Nxb4 lines. Black plays 9 ... f5 and practically invites a Qh5+ attack.

Doesn't look nearly as solid as just playing 7. d4, lol. But it's worth noting that it's a playable gambit.

If you put Stockfish or engine choice reference, plz always add, depth, nodes or duration of analysis (e.g, 1 billion nodes, which is approx 3 mins of analysis). The reason is that engines never evaluate this position on your board,  they evaluate the final position of the principal variation at the end of search horizon/time limit(so depth 20 and depth 40 may differ a lot)

I could not reproduce, your b4 move at 600 million nodes, depth 33/51.

Thanks for the reply! You're correct that the difference between depth can make a huge difference. 

Since you requested it, here's what we have:

Stockfish 10, contempt maximum, current depth 41. I've let it idle on this position for a few hours now. For hardware I'm running an i7 quad at 2 cores. I would love to set it to 8 but I've got other things to do on this PC as well, and the fan gets obnoxiously loud once hyperthreading kicks in.

@Pfren

That is correct, SF prefers 7.d4. Probably any human or engine would also agree.

My curiosity here lies in the fact that 7. b4 is more playable than expected, and has no known lines so we're basically charting some new territory here, though it's quite obvious this isn't the most objectively strong idea. 

What's your engine's contempt setting? Default? 

I've set it to max to force Stockfish to avoid more drawish lines. That may or may not be the main difference here. 

One other note, in the screenshot Arena's just showing you the mainline analysis. You can set it to show second and third-best moves as well, which currently look to be 7. b4 and 7. Nc3 depending on depth

"8.c4 just cannot be good- it weakens the d4 square for practically nothing." 

8. d4  and 8. a3 are apparently the alternatives. But it looks like Stockfish just plays c4 a few moves later, for example 

I guess the main issue between c4 early vs later is that Black gets to solve its issues a move or two faster?

What would you recommend over 8. c4? I know you have a lot of experience beating up people in correspondence. 

Looking at position move 8 now.

Current depth 35, it looks like the engine did reconsider. New line:

Edit: jury's still out. At a depth of 39 move 8, 8. c4 again becomes the engine's "preferred" idea, but only by a very small increment.

Slowly mapping this out here. Hopefully if anyone actually tries to play this position they'll have a couple of ideas to use in advance.

Depth 44, 8. d4 is once again the preference. Current suggested "mainline" is now ...

8. d4 and 8. a3 are consolidated lines, transposing into each other.

Couple days later, we're partway through our first-ever "match" on this opening. Search depth per move varies, but is between 30-45 on each move.

It doesn't look like White will get anything more than an equal-ish position so far. Other than the surprise value of a move 7 novelty then, without seeing anything more useful come out of this opening it might not have much bite to it. Without 7 ... Nxb4 the 7 ... a6 lines allow Black to castle and develop without too much interference, so White does have to ask what's the point?

f6 gg

If you could elaborate on that statement, it would be a lot more helpful

In the database I use there are no games with 6...f6 (hence it's not surprising that your move is a 'novelty'). 6.Bf1 appears to be a rare move order, and when it shows up black players always play 6...Be7 transposing to the mainline of the Re1+Bf1 variation, unless you have some other improvement here.

I use a hybrid approach. Chessbase has an openings database, which does show winning statistics etc. It also has LiveBook, which is essentially crowd-sourced engine openings; whenever you set infinite analysis on a position, the mainline recommendation is saved, and Livebook shows the top 3 analyses by depth reached and engine strength. 

On move 6 it looks like the database has more games played from 6 ... Be7, but for whatever reason the engines think 6 ... f6 is a slight preference. 

The odds are it just transposes back into 6 ... Be7 f6 lines later, so it's not really reinventing much, I think?

I personally try to use whichever recommendation is objectively most reliable, as both database statistics & engine recommendations have their own flaws. It's fun to discover where they disagree with each other.

Game 1 continues. It's looking very likely that White will lose, which isn't a very promising start for this opening. Once it's over I'll have to try and find where any likely mistakes happened.

Somewhat ironically it's White that ended up with too passive of a position. Noticeably, White also never recovered from the loss of the e4 pawn. 

Game finished.  I have no doubt that Black can convert this position, so I'm manually resigning it. Time for the tougher part, looking for where things went wrong.

https://youtu.be/-AdSN6CpuWQ