Zenmastur wrote: ↑
Tue Jan 14, 2020 8:05 am
I think there is an flaw in your reasoning. You assume that with your meager hardware you can find the drawing line. I pretty sure this isn't the case. Even with very good hardware this will be a difficult task. Mainly due to the number of moves it takes to convert an advantage in the opening to the same or better advantage in a solvable endgame. The sub-tree that has to be stored in the TT is way to large for a normal sized TT. So, if it is a draw, you will need a very large TT in order to find it and be somewhat sure that your analysis is accurate. Back when SF could have 1 TB of TT I think it was possible. But IIRC this has been reduced to at most 128Gb of TT. This will make if very difficult as I'm not sure a drawing sub-tree will fit in a 128Gb TT. Trying to fit it into a MUCH MUCH smaller TT is futile I think.
Huh? Storing analysis in a TT became obsolete since engines with learning appeared. Critter Session File, Shredder 12, Rybka 3 Persistent Hash, Jeremy Bernstein's Stockfish learning version and others are great examples of how little you actually need to store on a hard disk and how quick is it to retrieve them to solve a position to a draw.
The only problem is outdated eval, so storage isn't the problem, it's when the old software insists that a drawn variation is lost and the tree is too big to show it all the drawn lines so it can learn the right score. So the solution is to implement learning in modern engines.
Unfortunately* the people that have been doing so have kept their efforts private, because they don't want their learning code to be seen, so only a few people get to enjoy this kind of learning.
(*or fortunately, because if everyone had it then it's as if nobody had it
Interestingly learning is so powerful that in this game Jeremy Bernstein's learning Stockfish (which is public) was not outperformed by learning Stockfish 9 (which is private) by much. Like it was almost half as good (you'd get the same thing after analyzing about twice the positions, with the difference in ELO I had expected you'd need like 10 times the positions.) I was also surprised that learning Stockfish 10 (private) wasn't useful in this game, it threw the towel too early, but couldn't see the difference between drawn lines (say, 2.80 position that is actually drawn) and lost lines (which... turned out to be all of them.)
If I saw a good defense I would give it a go. The truth is I don't. They all look losing to me. It's just a matter of time to convert the opening advantage to a solvable endgame.
After looking for several hours at this almost daily since October I'm in the same boat
So perhaps concluding 1.g4 is lost by force is logical. The thing is that for whatever line you choose to investigate, it can be made to look worse than the others, eventually. One can find a draw and go back and see that black had a much better move that is crushing, it looks hopeless!