Thursday, April 28, 2016

How to speed up logical reasoning

My trial and error is replaced by logical reasoning. I believe that is a necessary step. But logical reasoning is terribly slow. The choice is to loose points due to excessive time usage, or due to erring because of skipping the check for possible refutations. Somehow, the logical reasoning must be sped up.

But how to speed up logical reasoning? The thinking itself cannot be sped up. I have been thinking about this for days, but I didn't found a satisfactory method. The only thing I can think of, is that the logical thinking must be condensed into some kind of knowledge, and that this knowledge must be memorized.

What I'm going to try, is to categorize important positions, based on the characteristics, in the hope to find some knowledge that is worth categorizing. Maybe someone has a better idea?

To err or too late

Wednesday, April 27, 2016

Investigating the initiative II

 Let's see how a more complex situation works:

Diagram 1 Black to move

r2q1rk1/pp2ppbp/2p3p1/4Nb2/PP1Pn2N/2nQ2P1/1B2PPBP/R4RK1 b - - 1 1
solution

[O] = Offense
{D} = Defense
Attacks that are easy to spot as harmless are coloured grey.
Tasks that play a role in the actual combination are coloured red.

White:
Nh4
[O] Attack Bf5
[O] Threatens Nxg6 - Rf8

Bg2
[O] Attack Ne4

Qd3
[O] Attacks Ne4
[O] Attacks Nc3
{D} Defends e2

Bb2
[O] Attack Nc3

Ne5
[O] Threatens Nxg6 - Rf8 
[O] Threatens Nd7 - Rf8
[O] Threatens Nxf7 - Qd8
[O] Threatens Nxc6 - Qd8 

e2
{D} Defends Qd3

d4
{D} Shields Qd3
{D} Defends N

Black:
Bg7
Attacks Ne5

Bf5
{D} Defends Ne4
[O] Attacks indirect Qd3

Ne4
{D} Defends Nc3
[O] Threatens Nxg3 - Rf1
[O] Threatens Nd2 - Rf1
[O] Threatens Nxf2 - Qd3
[O] Threatens Nc5 - Qd3

Qd8
[O] Attacks d4
[O] Skewers Qd3

Nc3
{D} Defends Ne4
[O] Threatens check at Nxe2+
[O] Threatens Nd1 - Bb2
[O] Threatens Nxa4 - Bb2

This list immediately shows the problem with rampant usage of CCT. There are way too much possibilities. Even when I greyed the most unlikely options, the list is way too long. Can we apply logic to the list, so we can filter it? What I mean is, can I define some logic that builds a small list without even taking the redundant issues into consideration?

The first thing is to formulate what the logic should accomplish. When I tried to solve this position, I considered the line 1. ... Nxg3 2.Qxg3 without noticing the knight fork 2. ... Nxe2+, even after 7:50 minutes thinking. I was clearly overwhelmed by the amount of offensive possibilities of white. The logic should help me to focus on the relevant pieces of the combination.
 It is only necessary that the idea behind the combination is revealed. Once I have the right idea, working out the moves and checking them for refutations is relatively easy.

The easy fix I hoped for, the rule of thumb of the previous post ("harass the piece with the most tasks"), is clearly not working here. The most active piece of white is Ne5. But I already greyed its tasks, since it is easy to see that its offensive possibilities are easily refuted. So should I focus on the defensive tasks only? The white piece with the most defensive tasks is pawn d4. Although there are some ideas that almost work, none of them withstands a more thorough investigation. To save time, I played 1. ... Qxd4, based on the knight fork Nxe2+. So I had seen a knight fork, but the wrong one.

The only clue that gives something relevant away is Nc3 threatens Ne2+
And that triggers a whole bunch of ideas that are not new to me, but I never managed to apply in practice, due to too little exact understanding:
  • Start to investigate the attack on the piece with the highest value.
  • Ignore the counter attacks for the moment. I want to find the right idea behind the combination, checking for refutations can be done later.
  • Gaining wood is always accomplished by a trap or a duple attack. Suppose that the piece with the highest value under attack is one of the targets. What is the most likely candidate for the second target? Can the second target be put in place?
  • If not, repeat with the attack on the piece with the next highest value.
With this simple reasoning I would have found the idea behind this combination. I wonder if it has any use beyond this position. I see that I'm way too worried about the responses of my opponent. That makes me investigate way too much possibilities. If I attack the piece with a high value, all counter attacks against my pieces with a lesser value are irrelevant. I just need the idea behind the combination. I must trust myself that I will be able to calculate the refutations later on.

Let's see if the points I have found in this post have any use for other positions. I'll be back.


Tuesday, April 26, 2016

Investigating the initiative

The past month has shown me the importance of applying logic in stead of trial & error. Maintaining the initiative during a combination is probably the most important subject of chess logic. But there are other important subjects too, of course. I plan a series of posts on the initiative. Since I have not the slightest idea where my investigating will lead me, I have no idea how long that series is going to be. I will probably often make use of diagrams with arrows etcetera. Alas doesn't Lucas Chess allow positions without kings, so please don't get distracted by issues in a position not relevant to the subject, like "the remaining endgame is a draw" or something like that.

Let's start with a simple position, the removal of the guard:

Diagram 1 White to move
Pieces perform tasks. A task can be something like "attacking a piece" or "defending a piece". Every task usually consumes a tempo. In general, one attacking move can be answered by one defensive move. If we want to make progress, we will have to find multi purpose moves, which perform multi tasks. If the opponent can't find a move that provides an answer to all attacks, then we will win.

In diagram 1, two of white's pieces attack a black piece (the green arrows). Two of blacks pieces perform a defensive task (the red arrows). Every task takes a tempo. But the tempi are evenly distributed. One move is one tempo for each colour.

1.Nxd6 Nxd6 is an example of two single purpose moves. The white knight executes the attack by capturing, the black knight executes the defense by capturing back. Both pieces disappeared from the board, along with the tasks they were involved in.

1.Bxf7 is a multi purpose move.
If white plays Bxf7, the distribution of tasks has changed. Not only the black knight has disappeared from the board, but also the defensive task it was performing has ceased to exist. Now black is in trouble, since he has to perform two tasks (save the black bishop, take back the white bishop), and he has only one tempo to so. Since there is no such multi purpose move, black looses a piece.

This might indicate a general rule of thumb: "try to harass the opponents piece that performs the most tasks, both offensive and defensive."

Task counting helps in this case to find the right move order. 1.Bxf7 is better than 1.Nxd6.

Hierarchy.
The value of the pieces can change this simple approach. Have a look at diagram 2.

Diagram 2 White to move
The situation is about the same in diagram 1. An important difference is that the black bishop on d6 is attacked by a white rook, a piece with a higher value.

Now 1.Rxd6 Nxd6 is no longer equal due to the higher value of the rook.

All of this is very simple, of course. In a next post I will look if the rule of thumb we found still holds true, and whether counting the tasks still is useful in more complex positions.

Monday, April 25, 2016

Enjoying the exercises

Today it is exactly a month ago that I quitted the salt mines, in favour of working on a thought process. Which conclusions can be drawn?

Board vision.
The salt mines I have mined, were specifically designed for improving board vision. While solving exercises at CT the past month, I time and again asked myself: Is my failure or excessive time usage caused by poor board vision? In less then 1% of the cases I found that poor board vision played a decisive roll. In other words: you can't expect much improvement by salt mining when it comes to blitz problems at CT with a rating around 1700.

Context of the position.
I wanted to create a formal check-list with questions to interrogate the position. That was what I initially understood as being the thought process. In two weeks this check-list developed towards a list of only three questions, which are mainly useful to understand the context of the position:
  • How is the material balance?
  • Is the position about trap, double attack, promotion?
  • Which targets are likely to be going to provide the wood to gain?
Usually these questions can be answered in under 10 seconds.

The real thought process.
The real thought process developed in a quite different direction than just a check-list to get an idea of what the position is about. I had an important revelation about the role of the initiative.
In the past, I had tried to integrate CCT in my game. But since trial & error was my main approach to the game, I found it to be too time consuming. Most CCT can be refuted in an easy way, and to investigate all these moves without any filtering is a daunting task with too little returns for the time invested.

That all changed when I replaced trial & error by logical thinking. Logical thinking provides the filter that CCT needs. You only consider CCT moves which are logical. It is hard to describe how the application of chess logic develops overtime. Chess logic has its own patterns. Analysis of the positions reveal those logical patterns, which were previously overlooked. The unconscious brain works its magic, and the new patterns are assimilated without further ado.

Progress.
It takes time to change more than a decade of old habit of trial & error. But after an initial dip of minus 70, I improved 50 points (reckoned from the initial starting point of 1700) during the past month. Analysis of spilled time shows that the room for improvement is gigantic, and I see no reason why an improvement to master level shouldn't be possible. I'm still talking about the niche of tactics solely, of course. The only uncertain factor is a change in the problems themselves when they become higher rated. But I have done a lot of 2200-2400 rated problems in the past (CT standard mode), and as far as I remember them, most are plain simple after analysis. Above 2400 the problems became more esoteric every now and then. It is my take that between master and grandmaster level the role of board vision will become more profound. When I reach a blitz level of 2000 at CT, I consider the improvement an epic success, and I will resume to play chess again.

Joy.
Initially, doing chess exercises was fun. When you feel you make progress. But when exercising became excessive, initialized by the Knights Errant, fun gradually disappeared. I didn't detest it of course, since I don't do things I detest for a hobby. The exercises themselves were addictive, to a certain degree, but the fun mainly consisted of satisfying the curiosity how the human mind and learning works. I learned a whole bunch of things about these subjects.

With applying chess logic, the joy in exercising and in the positions themselves is back. Which is probably the most important thing that happened the past month.

Saturday, April 23, 2016

Time usage

The good news is that I replaced a lot of clueless trial & error by logical thinking. The bad news is that logical thinking takes an awful lot of time, so that the nett result is an improvement of only 50 points in rating. Hence the hunt for possibilities to simplify logical thinking is open. In the last years the main motive for seeking chess improvement has been narrowing down to the mere curiosity how the mind works. I'm happy to see that enjoyment in the exercises themselves comes back due to adding logical thinking to the positions. So far I found the following causes of possibly uneconomic time usage:
  • decide with which piece to capture
  • decide which escape square is the best
  • decide which order of moves is best
  • decide whether the counter attack is dangerous or not
  • declining a good move 
  • repetition of lines which are not working

Decide with which piece to capture
What I need is a decision model. If the piece to capture is defended, then the piece with the lowest value should be considered first. But with the knight or the bishop? When to use a piece with a higher value like the rook or the queen? If a piece of my own is under attack, it is best to trade it off as first. Etcetera.
Diagram 1 black to move
3rr2k/1p4p1/p4b1p/1PR5/3B3P/1QPq2P1/5P2/1R4K1 b - - 1 1
solution

It took me a minute to decide whether 1. ... Bxd4 or 1. ... Rxd4 is best.
It should be possible to decide that faster, don't you think?


Decide which escape square is the best
I seem to have a special talent to choose the wrong escape square. I need a decision model. Which escape square prevent further checks? Which king move has an attacking value? Which king move defends a crucial piece?

Diagram 2 white to escape check
8/5ppk/2R4p/1R2p3/3q4/3P1PP1/4r1KP/2Q5 w - - 0 2
solution

After winning a rook, the only thing to do is to bring the king into safety. Why must that take an awful lot of time?

Decide which order of moves is best
 Sometimes you perfectly know which moves play a crucial role in the combination, but you need an awful lot of time to put them in the right order.

Diagram 3 white to move
r3r1k1/ppp2pb1/6p1/q2n1nB1/4N3/1P1Q1P1R/P1P1N3/2K4R w - - 0 1
solution

What should I do first:
1.Rh8+
1.Bf6
1.Nf6+
1.Qxd5
??

It takes a lot of time to work out the right move order. Why? Are there ways to improve on that?

Decide whether the counter attack is dangerous or not

Diagram 4 black to move
6qk/pp5p/8/P2P4/B4n2/5P1P/6r1/4RQ1K b - - 1 1
solution

Of course I know which move I want to make. But will the counter attack peter out? It took me 2:20 min more than average to find out. Are there ways to do it faster?

Declining a good move
Sometimes you decide to look not further in a certain line since your gut feeling tells it would not work. That prevents you from calculating a whole bunch of superfluous lines. Grandmasters calculate much less lines than amateurs. But when your gut feeling is wrong, you are toast.

Diagram 5 white to move
R6n/1p2r1pk/p1p4p/3qBP2/6P1/4Q2P/PP5K/8 w - - 1 1
solution

I declined the move 1.Qxh6+ in the first minute of investigation, since I thought the black king could escape. It took me 8 minutes (!) to decline all other possibilities and to have a closer look at the move.

Repetition of lines which are not working
When the lines I investigate don't lead to a solution, I tend to keep grinding the same moves over and over again. They look so tempting, but they do not work. Repeating your thoughts is a bad habit which takes an enormous amount of time. It seems to be a case of mental paralysis. In fact I'm clueless. The thought process is designed to take me out of that paralysis. Besides that, I'm making a list of themes I tend to miss.

Time is not on my side


Wednesday, April 20, 2016

Counting

During solving a position, the following diagram came along.

White to move
1rr5/4pp1k/p2p2p1/3P4/3bnPPp/B6P/PR4BK/2R5 w - - 0 2
solution

It took me about a minute to work out the best sequence to capture. That shows a clear weakness of me, I have the feeling that I should see this kind of sequences immediately. This slows me down in every position with multiple mutual captures. How to improve this?

Something really weird happened just yet. The following position took me 5:58 min to solve:

White to move
 6Q1/pb4r1/1p2N1k1/3p1p2/2P5/7P/1q3P2/2R3K1 w - - 1 1
solution

5:58 min?? Well, I was tired after doing a lot of exercises, but that's a poor reason. It shows a fundamental flaw. The fact that the problem has a blitz rating of 1715 shows that this flaw is pretty common:

Problem Blitz Rating:1714.7
Blitz Av Seconds:00:51
Blitz Attempts:200
Blitz Success Rate:64%

Problem Standard Rating:1517.4
Standard Av Seconds:02:58
Standard Attempts:790
Standard Success Rate:63.04%

Intuitively, I feel that having it correct in, say, under 30 seconds would be much more appropriate. Anyway, it's definitely an area that needs much, much improvement.

Tuesday, April 19, 2016

Examples of errors

Overlooking preliminary moves.

Diagram 1 White to move
Here I overlooked the effect of Ne6, chasing the king into the knight fork (which I missed too).

Overlooking simple mate patterns.

Diagram 2 White to move
Every now and then I miss a simple mate.

Missing an overworked piece.

Diagram 3 White to move
Here I totally missed that after 1.Bxe3 the rook at b8 can't take on b4 without giving up the defense of Rc8.

Underestimating my invasion.

Diagram 4 Black to move
Here I underestimated that the invasion 1. ... Be5 2. Nxe5 Qe1+ leads to mate. I dismissed the move Qe1+ because I had the feeling that white has all sorts of escapes, so I didn't check the move.

Wrong choice between two knight moves.

Diagram 5 Black to move
Here I had to make a choice between 1. ... Ng4 and 1. ... Nf3
Both moves threaten mate with 2. ... Qh2
I chose 1. ... Ng4 because of  the extra knight fork 2. Re2 Nf2+

Hidden behind this, is that I missed the following mate patterns: 1. ... Ng4 2. ... Nf2#
and 1. ... Nf3 2. ... Qh3#

The problem is not of course that the mate patterns are unfamiliar, but that I decide not to look for them. More precise: there is nothing that triggers the decision to look for the mate patterns.

Of course one can argue that I should look better. Of course I would find the patterns then, sooner or later. I can find almost any solution, given enough time. But this is a quest for speed too. Making the right decisions in short enough time. It is about optimizing the search process for the right move. Omitting unnecessary considerations is just as important as considering the right moves, when it comes to speed.

When I encounter the other error categories, I will add them to this post. In general I'm approaching a problem more logically now, in stead of by trial & error. But since logical reasoning takes more time, the improvement in method doesn't show up in the figures just yet. Well, an increasement of 90 points is not too shabby, but that's a new ATH, which is of course a statistical incident. At average, I increased about 50 points.

It is a matter of investigating what is going on in my mind during these mistakes and excessive time usage. I need a good diagnosis before I can think about a remedy. I'm going to count the errors first, so I know which categories occur the most frequent.