Thursday, January 29, 2015

Improving at Chess Tactics - train the visual and the logical

There is both a visual and a logical piece to chess, and both need to be trained.  

Beginners need to know basic tactical motifs -- both what they look like (visual), and how they work (logical).  For the visual piece they need to know, for example, that a bunch of pieces in a line could be a tactic (pin, skewer, discovery, fork, etc), pieces on the same color square as the knight can be forked, or what loose pieces look like.  Then they learn the logical piece that generate candidate ideas like "attack pinned pieces", and "loose pieces drop off".  

Once a beginner has these basic concepts down, and has been taught to be aware of/look for (the visual piece) these things in their games, the next piece is the calculation step (or what some people call "looking ahead").

Say a beginner who has been taught to be aware of basic tactical patterns "sees" a bunch of pieces in a line, and thinks it might be a pin.  First of all that beginner needs to stop right there and do a happy dance!  Visually seeing patterns in games is a huge step for beginners.  It is frustrating as a more experienced player to watch beginners play, and you "see" lots of potential (even though you might not have calculated any actual moves), but the beginner is completely clueless as to what opportunities are right in front of him.  As you improve, you learn to see more and more, and hopefully miss less and less.  There is a great quote about Kasaprov's ability to see things:

"I thought I was playing the World Champion - not some 27 - eyed monster who sees everything."  -  (on losing May 1986 match 5-1/2 - 1/2)  -  Tony Miles

Beginners have problems "seeing" things on the board in front of them.  More advanced players have problems "seeing" things several ply into the future -- for example, a series of captures could lead to a mate pattern that would be completely obvious if on the board in front of you, but is much harder to "see" deeper into al ine.  Exact same problem though.  

Ok, so back to the example...our beginner finished his happy dance, and goes back to looking at the pieces in a line.  What now?  This is now the "logical" piece, and where training often fails for beginners.  But this process is simply calculation and every beginner can learn some basic stuff.  The process is to generate a list of candidate moves, calculate each one to the end, evaluate the final position and pick the move that leads to the best evaluation.  It's not actually that hard to do, and there is a very specific logic used to calculate that anyone can learn.

Continuing with the same example, our beginner needs to compile a list of candidate moves.  In order to do that, he needs candidate "ideas".  Luckily he remembers something he heard once: "attack pinned pieces". So he starts to search for moves that attack the pinned piece.  Most beginners can get this far -- and it is at this point they begin to fail.  They find one move that attacks the pinned piece, 
do another happy dance, and make the move...wondering how in the world his opponent is going to reply to this vicious tactical blow.  

And then he gets mated.

The beginner needs to learn to do a few things.  First, identify as many candidate moves as possible based off of the candidate "ideas" you found when looking at the position.  Look wide first, not deep.  You should always try to force yourself to find more than one candidate move.  

Second, calculate each candidate move to the end.  Look for your opponent's refutations first.  I have a blog on how to do line calculation (I taught my young daughter how to do this -- I call it the "logic of forced variations").

Third, evaluate the final position.  Beginners should just stick with a basic material-only evaluation, and as you get more advanced you can add "activity" to your list.  But keeping track of what happens to your pieces and pawns during an exchange is really hard for beginners to do accurately.  Evaluation is a skill that will come with practice, and evaluating future positions can be pretty daunting.  

Finally, having evaluated the final positions you either scrap both candidate moves because they suck, and you have to start all over in looking for candidate ideas, or find you get something good in one (or more) of the candidate moves and make that move.

The visual and logical are indeed very different skills, and beginners should take the time to learn how to calculate variations as part of learning tactics.  

Solving Chess Tactics -- the Logic of Forced Variations (or, only calculate what matters)

In my original blog series on calculation I recommended ways to more efficiently calculate forced variations using the broadly-recommended approach of calculating ALL "checks, captures, and threats".  My original blog series discussed the overall process of calculation (candidate identification, line calculation, and evaluation), with an emphasis on ways to more efficiently calculate specific lines (the second step) such as looking at your most forcing moves first, and searching for your opponent's refutations first.  Many of those ideas are here.

Calculating all checks, captures, and threats all the time is extremely tedious, and in writing that blog series I felt there was something missing.  And I have finally figured out what that something was -- I call it the "logic of forced variations".

This is an extremely simple idea but one I see strong players fail to appreciate (and I have seen annotators give exclams for moves that fall into one of these basic ideas):

When there is a threat on the board, you only need to calculate the five ideas that deal with that threat (capture, block, move, defend, and counterattack).  You do not need to calculate every check, capture, and threat on the board.

Also, if your move creates a threat, you do not need to calculate every check, capture, and threat your opponent can make.  You only need to calculate replies that address your threat.  If your opponent simply ignores your threat, that puts you one step closer to winning the point.

Simple, right?  This is all you ever need to know to efficiently calculate any forced variation.

Ok, maybe a few more details would help.

Let's start with a very simple idea.  The most forcing move possible in chess is a check, and everyone knows there are three (and only three) ways to deal with a regular check:  capture the piece, block the piece, or move your king.  The rules of the game require you to only make one of those three moves.  If you can't, the game is over.

However, if your opponent attacks your queen, the rules of the game do not require you to deal with it.  You can deliver checks, capture pieces, and threaten things, completely ignoring the attack.  It is completely legal to make a move that does not address a threat to your queen.  Yes, you will lose your queen, but the game does not end there (you should probably resign though!).  However, assuming you see the threat you will probably want to do something about it, and those same three move ideas -- capture, block, move -- apply equally here as well, with a few very important additions:  defend, and counterattack.

I have written about this in other blogs (see here, and here), but have done more Stoyko exercises using this approach and I have confirmed for myself that I have not missed any answers due to exclusion (not even considering a candidate idea).  When I did miss moves it was because I didn't "see" the possible move, even though I looked for it.

The basic process is this -- every time your opponent makes a move, or you make a move that threatens something, you:

(1) Find the threat -- actually, find ALL threats -- that the move created.  You must do this every single move, period.  The one time you don't do it, is when you miss your opponent's obvious mate in 1.  Prioritize the list of all threats, and be very specific about what the threat is (eg, "he's threatening to play Rook to a8 mate", and not just "he's threatening mate").

If you can't find any threats, then do not continue with these steps.  This only applies to forced variations, meaning there must be a threat on the board.

If you miss a threat because you looked but didn't see it, go train those types of patterns until you always see it.  If you miss a threat because you didn't look...then look next time.

(2) Make sure it is an actual threat.  Allow the threat to play out in your mind.  Does the threat actually win anything?  If so, what specifically?  Is there some tactic at the end that makes it not work?  Can you make a move that, if the threat is carried out, would undermine the threat?  Add that move to your candidate list.  The "threat" could actually be a mistake, so start by looking at that.

(3) Now that you have found a threat that is both real and specific, you have to deal with it.  Here are the five options you have to deal with the threat:

- capture:  or somehow deal with one/any of the attackers (like a pin to something more valuable, or deflection);
- block: move piece to block the threat, or reposition a piece to block on the next move;
- move: simply move your piece somewhere else, or create space for your piece to escape to (or to allow a piece to block) on the next move;
- defend: simply add a defender to the attacked piece/square
- counter-attack: this can have three distinct goals:
- counter-attack to WIN: a move that creates a new threat that is at least equal to, but hopefully greater than, the threat against you,
- counter-attack to DRAW: a forced series of moves that draws through repetition, or stalemate
- counter-attack to DEFEND: a forced series of moves that allows you to employ any of the four basic defenses above (capture, block, move, defend).  This could simply be giving a queen check that allows you to reposition the queen to defend a square, or say attacking a queen with a rook in order to reposition the rook to block another threat.

(4) Search for all moves that accomplish each of the above candidate ideas.  This will give you a thorough and complete list of candidate moves.  This covers every single conceivable reply possible.  You might not actually "see" the candidate move, even at 1 ply, but just looking for it is the key.

If you don't find a candidate move, but you did look for it, then train the pattern that you missed until you never miss it again.  If you don't even look for the candidate moves, then you'll never find them!  Look next time.

(5) Work through each of the defensive replies, playing out each move one at a time, until you find the move that results in something good for you...or the least bad option.

That's it!  Try this the next time you run through a set of tactical positions.  This process will NOT help you identify vulnerabilities or find possible tactics.  But if you do see a threat on the board, you now know how to work through it...and if you find a way to threaten your opponent, you now know the full range of replies you need to calculate...which is not every single check, capture, threat possible.

For future blog posts I plan on exploring each concept in more detail.  For example, blocking an attack results in a pin, defending results in possible remove the guard tactics, and counter-attacking is what world champions do...