Saturday, July 23, 2016

Removal of the guard

I hope you don't get bored by all those positions and analysis. I actually find it very exiting! I feel we are making steady progress, albeit slowly.

My list with positions where I want to learn to see the solution in  stead of just to calculate it, becomes slowly shorter. Meaning that it is indeed possible to replace calculation by seeing. I'm in no hurry, and I take my time to grasp every detail of a position, trying to devise some rules that can be applied in other positions. I'm not interested in the very position itself, because most details are accidental, and hence limited to only that single position. I'm only interested in the details of the combination. What makes that the combination actually works? Why does it gain wood? I make little changes to the combination, trying to keep everything else the same. I add or remove pieces, and see how that influences the outcome of the combination, while ignoring the effects on the position. I change the move order and see what difference that makes. I use Stockfish to check my conclusions.

The following position is about removal of the guard.

Diagram 1 White to move
r1qr2k1/pp2ppbp/1np3p1/4P3/2b1PP2/2N1B3/PPQ1B1PP/3R1RK1 w - - 1 1
Solution

The black knight protects the bishop. I realized that removal of the guard by 1.Bxb6 only works because it threatens the black rook on d8. It is a multi purpose move. A capture plus a threat.

Let us see what the effect is of removing the possibility of this threat. Does the combination still work? In order to find out, I removed the black rook on d8. In order to keep the material in balance, I removed the white rook on d1 too. We get the following position:

Diagram 2 White to move
r1q3k1/pp2ppbp/1np3p1/4P3/2b1PP2/2N1B3/PPQ1B1PP/5RK1 w - - 1 1

Now the combination doesn't work any more. Since 1.Bxb6 doesn't threaten the rook, black is no longer obliged to take back on b6 first. Black has time to get rid of his problem bishop by playing 1. ... Bxe2
This leads to the following
Rule: removal of the guard only works when it is a capture which gains a tempo. 
 In this case, by a threat.
This rule probably can be generalized:
Rule: look for captures which are accompanied by a threat.

I asked myself whether it is necessary to threaten a piece of higher value. So I decided to replace the black rook by a knight. Adding a white knight on b1 for material balance.  I saw that 1. ... Bxe2 threatens the rook on f1.  For clarity I decided to remove the remaining rooks too:

Diagram 3 White to move

2qn2k1/pp2ppbp/1np3p1/4P3/2b1PP2/2N1B3/PPQ1B1PP/1N4K1 w - - 1 1

The combination still works. This leads to the following
Rule: look for captures with follow up captures

What happens when black has a follow up move too? I replaced the white knight from b2 to f1:

Diagram 4 White to move


2qn2k1/pp2ppbp/1np3p1/4P3/2b1PP2/2N1B3/PPQ1B1PP/5NK1 w - - 1 1

Now the combination no longer works. If both sides have equal follow up moves, removal of the guard doesn't work.
Rule: you need one follow up capture more than your opponent

Giving black more than one follow up possibilities doesn't change the outcome. The one who started with the first capture can decide to break off the series of captures whenever he wants.


Rule: only the one who starts the series of captures has the chance to win a piece. He wins a piece when the opponent runs out of captures.

Does the move order make any difference? In diagram 1 there are 3 possible captures:
  • 1.Bxb6
  • 1.Rxd8+
  • 1.Bxc4
The last capture shouldn't be considered as first, since it is a capture without a follow up threat.
What if  white plays 1.Rxd8+ Qxd8:

Diagram 5 after 1.Rxd8+ Qxd8 white to move


r2q2k1/pp2ppbp/1np3p1/4P3/2b1PP2/2N1B3/PPQ1B1PP/5RK1 w - - 1 1

If white now captures the knight on b6 with 2.Bxb6, then black can take back with 2. ... Qxb6+ and the combination doesn't work due to the check. But that should be considered an accidental feature of the position. If the king had been on h1, the combination would still work!

What does this tell us? We have to look for captures with an additional punch first. The winning tactical theme here is removal of the guard. Worries include the possibility for the now unprotected black bishop to counter attack. It becomes a desperado.

Other captures than 1.Bxb6 that have an additional bite, are in general equally well playable (=1.Rxd8+) without changing the outcome of the main tactical theme (removal of the guard). In essence, such captures with extra bite are postponement moves. They postpone the execution of the tactical theme by one move, since they require immediate action. You can play an infinite amount of postponement moves, but the main problem for black remains.

It is important to grasp the main ideas of the position, to reduce you calculations. But you still need to calculate every line. Since there might be a capture with a counter bite for your opponent as well that you might overlook. But calculations with a mind that is not already overloaded, is usually no problem.

Removal of the guard

Thursday, July 21, 2016

Rules

This is the second or the third time I encountered this position. I devoted quite some time to it in the past. I might even have posted about it, I'm not quite sure. If so, I apologize that I post about it again. My excuse is, that I have apparently learned nothing from it, since I got the solution wrong. Again.
I think it is an important position, and I think it should be possible to see the solution at once. Maybe I can devise a rule or two during the process of learning how to see the solution.

White to move
r1bq1rk1/pp2ppbp/2np2p1/8/2PNP1n1/2N1B3/PP2BPPP/R2Q1RK1 w - - 1 1
solution

The last move of black was 1. .. Ng4. The black knight is hanging. The problem is, that if white takes it, the white queen becomes overworked. If white wants to stay ahead, he needs an extra tempo. Where does this extra tempo come from? After 1. ... Ng4 2.Bxg4 Bxg4 the white queen is hanging. If the queen takes on g4, black will get his piece back by capturing the knight on d4. White can gain a tempo by 3.Nxc6 in stead of 3.Qxg4. This power move accomplishes the following:
  • White gets rid of his problem piece
  • It captures a piece
  • The black queen is under attack
Rule: use your problem piece to capture
Rule: choose the capture that poses a new threat

This might give you the impression that you can change the move order. Why doesn't that work?
1. ... Ng4 2. Nxc6 is answered by blacks power move 2. ... Nxe3
Now it is black who abides the two rules above:
  • He gets rid of his problem piece
  • He captures a piece
  • He threatens the white queen
Which leads to a new
Rule: capture the piece of your opponent that can capture a piece of you with tempo first

If black takes the knight first 1. ... Ng4 2.Bxg4 Nxd4 then the rule "use your problem piece to capture" applies. 3.Bxc8 gets rid of your problem piece.


Tuesday, July 19, 2016

When to count?

I changed DGF into DIG, since that put the tasks in the right order, and it is easier to remember than DFG. DIG = domination square, invasion square, guard. So now we get M WIMP DIG.
I always used invasion square (or convergence square) in stead of focal point in the past, btw.

When I tested the system we are trying to develop against practice, I noticed that we might count at the wrong moment. I started with counting right away when I look for domination squares. The reason for this is that I want to prune squares which there might be a contact, but which are not interesting. But look what happens in this position when you start counting:

black to move
 5r1k/pp6/3bprqp/3p2p1/1P1P4/P2QPNPP/6K1/3R1R2 b - - 1 1
solution
Domination squares: f3, f1, d3, g3
Invasion square g4

If I count the situation on f1, then it is attacked twice and defended four times, so it scores minus two. But that score is irrelevant, since with two power moves, I can harass 3 of the 4 defenders. For instance: 1. ... g4 2.Nh4 Qxd3 3.Rxd3 winning the white rook at f1. This means it makes no sense to count before you consider the defenders and how to harass them. You can't prune anything based on counting for domination. Without counting for domination, you don't know whether you dominate a square. That makes the term "domination square" nonsensical. Maybe "contact squares" is a better term. That frees the D from Domination, which I can then use for Defender, the term that I am used to, in stead of Guard. Which reads M WIMP CID, which is a bit of a pity, since DIG is easier to remember than CID. But hey, it is a work under construction! We need to automate it anyway, so it is just a pair of training wheels.

Monday, July 18, 2016

See more, calculate less

My database with failures at CT has grown to 135 positions. After carefully studying them, I found 62 positions to be "easy". That is, I know them well enough to not make the same mistakes again in the future. I can now see the solution of them without any calculation or effort. 9 positions are too complex. Meaning that I cannot find the solution without serious calculation. I don't expect to see their solution any time soon without considerable study first. That leaves me with a selection of 64 positions which have a simple solution (with hindsight, after studying them well), but where I don't see the solution without calculation. I think I should be able to see the solution, though. I'm going to use these 64 position solely for exercising to see the solution in stead of calculating it.

I expect to need a few hours per position. What makes it even more difficult is that such exercise is definitely out of my comfort zone, and I constantly try to escape it by checking my facebook, twitter, the news, writing blogposts like this, etcetera. I now open five positions at the same time, and if I want to escape from one, I do so by going to the next position. That somewhat works.

For you to know what I'm talking about, I will give you a few example positions.

Diagram 1 White to move
 2r3k1/1R5p/p1q1rnp1/3p1p2/P7/2Q2B2/1P4PP/2R3K1 w - - 0 1
solution
As said, what I try to do is to see the solution without any calculation. I already have done that with a few other positions, and it is possible. I focus on the story of the initiative. How do I find forcing moves, and how does that translate to the gain of wood? Which squares am I dominating, who are the defenders and how do I harass them, what are the focal points? In short: M WIMP DGF?

Diagram 2 black to move
 4r3/p1k3pp/2p2p2/8/PrRnP3/1P2R1K1/3N2PP/8 b - - 0 1
solution
Diagram 3 black to move
5r1k/3B2pp/3P2n1/pp2p3/6P1/PP5P/1BPNpr2/3R1RK1 b - - 0 1
solution
Diagram 4 white to move
4R3/r4kn1/5P1B/2pp1qQ1/1p5P/3b2P1/5P1K/8 w - - 1 1
solution

I expect that what I learn from these exercises will transfer to other positions as well. I have mastered a few positions this way, and it seems that my intuition has grown. It happens more often that I can't resist to make a certain move, which happens to be the right one. In the past most of the time such moves were wrong. So it might work.

The fact that I have written this very post, already indicates that I'm still struggling with distractions though.


Tuesday, July 12, 2016

M WIMP DGF

I have been testing the three counting methods from the previous post. It turns out that only the first method is viable in practice. Counting dominated squares, or "interactions on squares" as Robert put it, is in itself not enough to solve every puzzle. A few extra characteristics must be recognized before you can figure out the whole combination. To remember them, I use the mnemonic above "M WIMP DGF"

M-Material balance
You can't do without evaluating the material balance.

WIMP
You must know what the position is about.
  • W-Wood. Gaining wood
  • I-Invasion
  • M-Mate
  • P-Promotion 
It is quite possible that the theme of the position is about more than one subject. WM - you try to M-mate, and the opponent can only prevent that by giving up W-wood, for instance.

D-Domination
It turns out that finding the squares you dominate is essential to understand any combination.

G-Guard
In many positions it is key to find the defenders, and the defenders of the defenders, since "removal of the guard" is the main issue in the position. You must be able to see the (chain of) guards.

F-Focal points
If you focus on domination, it turns out that you will miss some important squares. Invasion square, attacking square and the like. Focal points comprise these squares.

When I use the mnemonic above, I usually am able to find the solution. Given the importance of domination, it is probably a good idea to isolate that as a separate exercise.

The other two counting methods, counting tempo's and counting the value of obligations, are not practical. Counting tempo's, or obligations, is in fact counting the amount of dominated pieces. You must learn to see the effect on tempo's of a move. That should be treated as an isolated exercise too. Counting tempi might be of help initially, but the sooner you change to seeing the effect of a move, the better. It is really move related. We must see whether moves are single tempo or duple tempo. After the exercise of seeing the effect of a move on the obligations has been fully mastered, we will have a look at the third counting method, counting the value of the obligations. Hopefully it isn't necessary by then any more.

WIMP

Monday, July 11, 2016

Counting from -/-1 to 3

Robert has commented on the previous post, and he inspired me to this new post based on the position in his comment.

White to move
 r3r1k1/2qb1pbp/n1p2np1/4p1B1/1P2P3/P1N2N2/2Q1BPPP/3R1RK1 w - - 0 21

There are 3 separate subjects to count:
  • Domination
  • Tempo's
  • Value of obligation 
Domination
First, you count for domination. I don't know whether "domination" is the best term for it, but it seems appropriate. You count from -/-1 0 +1 (-/- is minus sign). -/-1 = you don't own the square. 0 = balanced, +1 = you own the square (and the piece upon it). Both a6 and f6 are balanced, but since the pieces on it are both defenders, black is weakened elsewhere if you capture them. Counting for domination gives you a sense of which targets play a role.


Tempo's
Second, you count the tempo change caused by a move. You count 0, 1, 2. 0 = no forcing move, 1 = single tempo move (places one obligation on the shoulders of your opponent), 2 = duple tempo move (places two obligations on your opponents shoulders). 21.Bxa6 is a single tempo move. Black is obliged to take back a piece of the same value. Sooner or later. With a postponement move, black can postpone his obligations, but there will come a time that he has to comply with his obligations. In this position there are no postponement moves, so black must take back on a6 immediately. 22.Bxf6 is a single tempo move, and black is forced to take back immediately.
23.Nd5 is a duple tempo move which attacks two undefended pieces. Black needs a duple defense move to save them both. There are two of these duple defensive moves: 23. ... Qd8 and 23. ... Qd6
Both move has as drawback that the black queen is now overworked, it cannot defend both bishops on f6 and d7 at the same time. White finishes with the duple offensive move 24.Nxf6+, and since black hasn't an appropriate duple defensive move, he looses a piece.

Counting the tempo change of a move is a simplification of the tempo counting system. It is based on the assumption, that the obligations for both sides are in balance. If you play a game, you know that for sure, since if it were not, you would have gained or lost a piece already. If for a puzzle the obligations are not in balance, you simply gain the wood you dominate. Only if you have to analyze a position that is not your own game or a puzzle, you need an absolute obligation counting system. Since such system is rather complex, we should avoid it when solving puzzles. The relative tempo counting system I propose here should be sufficient. At least, that is what we have to test.

You must realize that you can have obligations already. This means, that if you attack with a piece that was previously under attack itself, you gain two tempi. One defensive, and one offensive. We must develop a sense for this.

Value of obligation
If you capture a rook, your opponent is obliged to take one or more piece(s) back of equal value. It doesn't suffice for him to take only a minor piece back. I don't like counting with numbers like 3, 5 and 9, since I easily overload my poor short term memory with it. So I'm going to experiment with Minor piece = 1, rook = 2 and queen = 3. I'm interested in the mechanism of the combination. Once I have unearthed the mechanism, it is soon enough to calculate the exact value for the gains and losses on both sides. When I don't have a clear sight on the inner workings of the combination, it suffices to know that if my rook is taken, I need at least a rook or two minor pieces back. It is just a precaution for overloading my STM.

Anyone who can count from minus 1 to plus 3 should be able to use these 3 counting systems.


Sunday, July 10, 2016

Two tempi gain wood

I have selected a problem set where the first move is a capture. The second C of CCT. Categorizing them and investigating them is an ongoing task, rather technical, so there is not much to write about. But from time to time, a little nugget is unearthed. We already know that a duple attack gains wood by the effect that two targets are attacked with one move, while the opponent cannot save two targets with one move (usually). The same principle works with captures. If you can gain two tempi with one move, it gains wood (usually). Unless your opponent has a duple tempo move as an answer. See the following diagram.

White to move
r2qr1k1/ppp2ppp/2nb4/8/3PB1b1/2P1BN1P/P4PP1/R2Q1RK1 w - - 1 1
Solution

As you see, both a white and a black bishop are hanging. Taking the black bishop is just a trading off of the bishops. But 1.Bxc6 gains two tempi. It attacks the black rook, and it prevents the bishop from being taken on e4. So even one tempo is a defensive one! The move trades off a problem piece of white with tempo. It is important to develop a sense for those two tempo moves. Both offensive and defensive.