I don't think it's very meaningful, a single search from such a position with so much material on board isn't going to solve it.
Over at CDBCN, the evaluation of the Grob is at -143cp and it has been trending down. For every single potential draw line that get suggested, just focusing it by going deeper ends up finding a bust. The sheer amount of lines that white can throw at black, however is practically limitless.
The critical test of 1. g4 d5 2. c4 is Bxg4, not e5, according to CDBCN.
I don't see ChessDBCN as meaningful here. Based on http://talkchess.com/forum3/viewtopic.php?t=71764 it's only depth 22. Even if it's deeper now, I looked at these positions much deeper and also with 7-piece TBs and 450 GB RAM. I'll check 2... Bxg4 but it didn't look as good.
The leaf evals are depth 22, but the tree is massive. As you know, a depth 80 search with SF doesn't mean looking at depth 79 all children, then depth 78 all their children... There is a lot of reductions on all moves that look bad. Searching a million nodes with depth 22 leaves can be quite competitive.
ChessDB has also the advantage of persistency, while you lose your massive hash if you stop Stockfish.
I wouldn't expect my analysis to be useful for popular openings but 1. g4 d5 2. c4 is not exactly common, so I expect that it wasn't analyzed nearly as deeply. But I'm not certain on what the apparent depth number displayed by ChessDB.cn interface under the notes column means exactly and how to compare it to the depth reached by engines. Or how it does pruning since it's clearly not analyzing all moves. Also, I'm analyzing with SF NNUE and I'd guess ChessDB.cn used older versions of SF.
ChessDB's SF sometimes gets updated, it's currently using a NNUE-based SF version. I'm not sure it makes a massive difference for the task of busting the Grob, though.
Let us know what your Stockfish search suggests in the Bxg4 line as a defense for white.
Also, while the DB wouldn't automatically try to analyze the Grob a lot because 1. g4 is bad, this line has seen massive user-driven exploration. 1. g4 has much better exploration than say 1. f3.
mmt wrote: ↑Sat Sep 26, 2020 3:55 pm
I wouldn't expect my analysis to be useful for popular openings but 1. g4 d5 2. c4 is not exactly common, so I expect that it wasn't analyzed nearly as deeply. But I'm not certain on what the apparent depth number displayed by ChessDB.cn interface under the notes column means exactly and how to compare it to the depth reached by engines. Or how it does pruning since it's clearly not analyzing all moves.
Certainly, earlier this year, it did not look like it was useful for this opening. We'd want some sort of stability in its scores before we can begin to take it seriously.
Alayan wrote: ↑Sat Jan 25, 2020 7:38 pm
So, white needs to deviate from this line somewhere. ChessDBCN hasn't found something convincing yet, its score after 7. ... Ne7 has moved to -1.30 after I fed it the refutations to its preferred line.
I managed to get the score up to -0.90 by feeding it my refutations to its black's main attacks.
Looks like you can just pick a side and refute its lines and make its score whatever you want.
Alayan wrote: ↑Thu Jan 30, 2020 4:22 pm
I can confirm you had it pushed down over -1.00. I pushed it back to -1.30 by feeding it counter-analysis.
But it likes Rb5 after Na4 in the h5 line, while my homefish in the end preferred Rb8 (Rb5 was the top move for a few depths). It could be refuted back and forth for a while I think.
The only true values of a chess position are draw and mate in X. A score like 150cp is always incorrect from a chess truth perspective, it's only a useful approximation for humans and engines to denote a position that's very difficult to hold, maybe losing by force, maybe not. So of course it can't be stable.
If you feed the DB with a way to draw against its best attack, or to win against its best defence, you'll move the score one way or another. But it already contains a lot of data. And remember, you don't need a single drawing/winning line, but many, against the different options available to the other side.
So far, no drawing attempt has stuck, there is always a way to push the evaluation back to a high cp score either by going deeper in the line or by switching to another attacking move. A few days ago, The evaluation after 1. g4 was 140cp, now it's 150cp. I've used the live AI to play games from the leaf positions of the best-drawing attempts, with the resulting positions automatically added for DB evaluation, and it has consistently allowed to grow black's eval, cp by cp.