What is the pawn structure.
Help in seeing the board, assess a position quickly.
Sort:
I'm pretty sure computers only use heuristics when the position is quiescent. If there are available moves of a forcing nature such as checks and captures, engines just follow those lines until there are no more. Since this represents a fairly small subset of the possible moves, it makes sense for them to do this. It could be that they have changed. It's been a while since I've had anything to do with computer chess.
Sorry if this is a repeat post. I've searched the forum and found many threads with tips on how to look at a position such as:
http://www.chess.com/forum/view/general/how-to-analyse-position-quickly
http://www.chess.com/forum/view/general/how-do-i-quotlookquot-at-the-boardpieces
and...
http://www.chess.com/article/view/chess-tactics--definitions-and-examples
But all of these tell you WHAT to look for - threats, imbalances, tactical situations, outpost squares, etc. I want to know HOW to SEARCH the board in order to find any of those as fast and efficiently as possible.
As an example, being a math/computer guy, I asked myself, "how would I program a computer to assess the board quickly?" and then I figured I'll train myself to do that method mentally. My first idea:
I. Pieces then squares method
1. Look at each piece.
1.a. Assess current state: is it being attacked, is it being defended, is it defending another piece? other "static" tactical observations.
1.b. Assess possible new squares for that piece: where can it move? can it attack? will it be attacked? will it defend, tactical situations and combinations from that square? etc.
1.c. Repeat until all pieces on the board are done.
2. Squares
2.a. What are good attacking squares for me?
2.b. What are bad squares for me? (want to keep opponent out of them)
2.c. What pieces can get to that square? etc.
3. List candidate moves.
4. Assessment phase of move is done, now calculate move.
I tried a few ordering methods for my "pieces then squares" method to improve efficiency...
1. My pieces first, left to right then top to bottom (why not?), then opponent pieces in same order, then squares in that order. It was slow and I could miss important things if they were on the bottom right since I was tired by then. so next I tried...
2. My pieces, in order of K, Q, B, N, pawns (furthest advanced first, then left to right). Next squares, that attack K, Q, etc. Then repeat for opponent.
Results so far: It's comprehensive, but very slow. I was getting the solutions on tactics trainer, but it took at least a second or two per peice on the board just to do the initial assessment of the board and list of candidate moves (usually about 1 to 2 minutes) so my rating has gone down to 960 because I can almost never get a solution in time. It amazes me that average time is 20 or 30 seconds for some of them. I'm barel half way through my assessment by then.
Worse, this ONLY gets me to a list of candiate moves. I still have a lot of calculations to do for each candidate move and so this doesn't seem sustainable.
My second idea...
II. Tactic pattern search method
I know the human brain tends to be better at pattern matching than brute force, so I thought of going through a board search checking for visual tactical patterns in some priority order. Each pattern has a dependency, such as pieces being on the same rank/file or diagonal with others, etc. But I then I started to think, many patterns depend on the same piece so it feels like I'll be looking at the same piece multiple times. E.g. if my method is...
1. Find Mate threats - look at King, threat squares (N), diagonals and files/ranks open to king, etc.
2. Find pins - look at king (again!), look at queen, etc.
3. Find skewers - look at king (again!), look at queen (again!)
So maybe I should group this by piece to make this more efficient:
1. Look at King
1.a. find mate threats
1.b. find pins and possible pins
1.c. find skewers and possible skewers
etc.
2. Look at Queen
2a. find pins and possibles
2b. find skewers and possibles
etc.
3. Look at diagonals (if B or Q is on board)
- etc. but may involve the King and Queen again!
4. Look at files, then ranks (if R or Q on board)
- etc. but may involve the King and Queen again!
and so on. (BTW, pattern searches for N seems very tough!)
Longer term, I feel like people have got to have figured all this out, but I can't find anything that approaches the methodology at this level of detail. I know the tactics to look for, just not the fastest methods for spotting them.
This is all going to resurface during calculation - I have to go from minutes to seconds for each position to have any chance.
Books, links and other stuff welcome.