Cornfed wrote: ↑Tue Jun 30, 2020 4:27 am
For Chessbase Tactical Analysis, using MCTS on multi-pv (even 3 or 4) with 3 or 4 cores conveys no advantage over Regular Komodo at the same # of cores even though the program defaults to multi-PV of 2...and you could probably set it to multi-PV of 3 or 4 before starting it?
MCTS should be better than regular Komodo for this and any feature that relies on MultiPV.
I'll quote Mark Lefter to explain why:
mjlef wrote: ↑Sun Nov 11, 2018 7:07 pm
In regular search mode, when you set say MultiPV = 3, the engine has to effectively search 3 times (one for each move). Of course, the second and third searches should take less time since they progressively have fewer moves to consider. But it does take a lot longer than finding the best move. In MCTS mode, since the program is creating a search tree in memory, with the nodes linked, it does not take extra time to find the second and third best move (well other than figuring out what they are based on visit count and win percentage). So MultiPV is "free" in MCTS mode. Of course, the way MCTS grows the search tree is to visit and expand the best performing moves/nodes more, so lower scoring moves will not have as large a tree under them which you can see by looking at the PVs. I suppose humans do the same thing, search the lower chance of winning moves less deeply than the more promising ones.
mclane wrote: ↑Tue Jun 30, 2020 11:12 pm
my last version is 12.2.
why should i update to 14 ??
More than 400 elo gain in the MTCS version which is the best A/B engine for MultiPV=2 and above, and besides the elo improvement over 12.2 there's a new Armageddon mode which allows you to analyze positions as if one side had to win at all costs.
Cornfed wrote: ↑Tue Jun 30, 2020 4:27 am
For Chessbase Tactical Analysis, using MCTS on multi-pv (even 3 or 4) with 3 or 4 cores conveys no advantage over Regular Komodo at the same # of cores even though the program defaults to multi-PV of 2...and you could probably set it to multi-PV of 3 or 4 before starting it?
MCTS should be better than regular Komodo for this and any feature that relies on MultiPV.
I'll quote Mark Lefter to explain why:
mjlef wrote: ↑Sun Nov 11, 2018 7:07 pm
In regular search mode, when you set say MultiPV = 3, the engine has to effectively search 3 times (one for each move). Of course, the second and third searches should take less time since they progressively have fewer moves to consider. But it does take a lot longer than finding the best move. In MCTS mode, since the program is creating a search tree in memory, with the nodes linked, it does not take extra time to find the second and third best move (well other than figuring out what they are based on visit count and win percentage). So MultiPV is "free" in MCTS mode. Of course, the way MCTS grows the search tree is to visit and expand the best performing moves/nodes more, so lower scoring moves will not have as large a tree under them which you can see by looking at the PVs. I suppose humans do the same thing, search the lower chance of winning moves less deeply than the more promising ones.
Thanks. I'll have to get 14.1 when it comes out and simply compare to see what if any differences there would be.
I am not wondering how well MCTS might perform (vs say Komodo Regular or SF) in the Chessbase Deep Analysis feature. I have not tried that in a good year.
mclane wrote: ↑Tue Jun 30, 2020 11:12 pm
my last version is 12.2.
why should i update to 14 ??
More than 400 elo gain in the MTCS version which is the best A/B engine for MultiPV=2 and above, and besides the elo improvement over 12.2 there's a new Armageddon mode which allows you to analyze positions as if one side had to win at all costs.
Really? More than 400 Elo?
I see only 100 elo in the usual ranking lists.
mclane wrote: ↑Tue Jun 30, 2020 11:12 pm
my last version is 12.2.
why should i update to 14 ??
More than 400 elo gain in the MTCS version which is the best A/B engine for MultiPV=2 and above, and besides the elo improvement over 12.2 there's a new Armageddon mode which allows you to analyze positions as if one side had to win at all costs.
Really? More than 400 Elo?
I see only 100 elo in the usual ranking lists.
Yes, the big jump for Komodo MCTS was between 12.1 and 12.2, roughly 200 elo, roughly another 100 elo since then. Aside from the MCTS elo gain, there are also some new features such as personalities and auto-skill, and moderate elo gain for the standard version.
In the everyman book "understanding the sacrifice" from 2002, in the chapter exchange sacrifices, the author brings up a fine example of the piece constallation RB vs NNP. If the bishop is tampared or just plain bad, often the knight duo is capable to put up some resistance.
Just like bishop and knight with/without a pawn often is capable of putting off some resistance against rook and bishop. Of course bishops need to be on opposite colors.
I think the NNP side should be given some sort of compensation in the material table if tangible.
Piece combinations eligible for material bonus/malus in the material table
RB vs BN + Pawn (bishops of opposite colors) sometimes give the defender some scopes for a draw in the long run.
RB vs NN + Pawn (bishop need to be bad to have effect). If the bishop is blocked and if the knights can get posted firmly, they're able to put up some resistance sometimes. Knight-duo is capable of holding a rook and bishop at bay if much of the board is blocked.
RB vs BB + Pawn (Having BBP when facing RB is clearly worse than facing RN. Nevertheless I'd say BBP is scantly enough compensation against RB. Although a pair of bishop is able to put up a stubborn defense.
RN vs BB+ Pawn (RN vs BBP is often ample compensation but still underdog, ceteris paribus)
[d]6k1/2q2pp1/1p2p2p/2p1n1nP/p1Pr2P1/P2PR1QB/1P5K/3R4 w - - 0 33
Minor piece duo RN vs BN - the weaker side usually require a pawn to meet the requirement as compensation. I can't think of any possible scenarios were this particular duo excels at, maybe there is positions but hard to imagine.
