bob wrote:What time odds? looked like 40 moves in 5 minutes to me for both sides...
You mean it is not in the PGN tag? I guess this should count as a WinBoard bug, then. I did implement time odds, but I did not alter the way the tag was written. Is there an official format for writing the time-control tag in a time-odds game?
From the times the engines spent on the individual moves, the time odds should be apparent. E.g. my engine thought 1.0 sec on the firs move, Stockfish 30 sec...
bob wrote:What time odds? looked like 40 moves in 5 minutes to me for both sides...
You mean it is not in the PGN tag? I guess this should count as a WinBoard bug, then. I did implement time odds, but I did not alter the way the tag was written. Is there an official format for writing the time-control tag in a time-odds game?
From the times the engines spent on the individual moves, the time odds should be apparent. E.g. my engine thought 1.0 sec on the firs move, Stockfish 30 sec...
Did not look at the actual moves, etc. Sorry... I just glanced at the top, and never thought about time odds...
I made some more trweeks to the eval of the specalized Q-vs-N engine. In particular increasng the passer bonus, as passers for the side playing the Queens proved extremely dangerous, and the major source of wins. So even gaining a Pawn is a sttrict no-no if it providdes the Queens with a passer. As a result, the specially tuned engine now has the upper hand over Daydreamer even with only 6 Knights in 5+0 games. (24+16-0=, but two losses on time by Daydreamer in drawish positions.)
From watching the games, the following pattern emerges:
* losing a Knight against any number of Pawns is always fatal.
* trading 2 Knights versus Q is almost always fatal, unless you are already more than 3 Pawns ahead, and your Pawns are reasonably close together.
* 5 Knights against 2 Queens is a very easy win, even without Pawns.
* when you still have 6 Knights, you have the better chances, no matter how many Pawns you are behind (as long as they are not passers).
* A King can safely march across the board even in the presence of 3 Queens, abandoning any Pawn shelter, when surrounded by the 6 Knights.
I guess these observations show there is still room for improvement of the evaluation: the value of the Knight should be upped even more. Basically trading should be avoided at all cost, except at the 'exchange rate' NP or NPP vs Q. So when faced with the choice to lose one or two Pawns, or trade 2N vs Q, is should still avoid the trade. Hanging on to the 6 Knights it can be confident that it sooner or later gains the Pawns back. As the Knights are slow pieces, gaining the opponent's weakest Pawns at the base of his Pawn chain is usually beyond the horizon, especially when you need to do it in such a way to avoid it will allow 2N vs QP. But it will be inevitibly there.
The more I analyze this, the more I get the feeling that with a bit of strategic insight in the Pawn structure and good vs bad relative Knight positioning, the 6 Knights could be unbeatable after all. In the games where they win, they do so extremely decisively. The Queens are completely strangled. Especially when 4 Knights 'gang up' on the enemy King, there is very little recourse for the Queens. The problem is they cannot do that whitout abandoning their King, if the latter stays near the back-rank. With the enemy King out of range the only targets they can conquer are Pawns, and indeed they are consistently gobbling these up, despite the best efforts of the Queens to defend them. Black only loses when it over-extends itself in its greediness, and misses a tactical shot. There always is very deep tactics here.
There seems little against building in a drive to centralize the King, so that King + Knights will go as a dense pack towards the enemy King, as unstoppable as a land-slide. I will experiment a little bit, to see if I can induce this behavior through the King PST.
I did implement time odds, but I did not alter the way the tag was written. Is there an official format for writing the time-control tag in a time-odds game?