Thursday, November 29, 2007

Chess blog nostalgia

Via the wayback machine, we have access to chess blogs that are long dead. You don't get the whole blog, but at least you get a little taste. My two favorites are Patrick's Chess for Blood and Quandoman.

Funny Fritz

White to move (a position I reached in a game I just played as white):
Pretty much any move that doesn't give away the queen for nothing wins for white. I made a reasonable human move, Qf2+, forcing him to trade Queens, giving me an easily winnable endgame. Fritz, on the other hand, doesn't even rank that move in the top 20! Indeed, it says of that move 'Gets refuted' (and the "refuting" move is Qxf2+).

The lesson: Fritz doesn't play like a person, and Fritz doesn't like to simplify to an easily won endgame.

Note I'm not saying Qf2+ is theoretically best, but I like it because it involves the least amount of thought, follows the old saw "When ahead, make even exchanges, especially of Queens."

P.S. My play is better now that I've stopped worrying about ratings again. The cycle continues.

Tuesday, November 27, 2007

Testing liquid egg product's product

Thanks to Liquid Egg Product for telling us his hack into Chesspublisher's code so you can scroll the annotations on the right. Hopefully it works.

I don't think I handled this ungodly opening (white played 1. b4, I was black) all that well, but I definitely blew it in the middlegame. This was a tough loss largely due to positional blunders that set up tactical opportunities for my opponent. I totally blew it at one point as described in the annotations.

Monday, November 26, 2007

Horse latitudes


Being done with the Circles is weird....Drifting aimlessly...The honeymoon is over. I am somewhat planless, aimless. Plans for after the Circles that I had envisioned at the time don't appeal to me. Partly I just don't feel like working hard at chess right now like I did. Perhaps I need to follow my own advice and relax, not worry about it for a while. I don't want to burn out on chess. Luckily I'm not sick of chess, just sick of chess improvement. Or at least intense crazy chess improvement like the Circles.

I'm not as on fire in my games, likely because I am playing better players now that my rating is higher. Hopefully a chance to make new patterns of mistakes. Or it is because my tactical vision has dulled. My first 20 games after the Circles, I think I lost only one. I was just smokin'. Now my results are a bit more even.

I am working on visualization using Snyder's Unbeatable chess lessons for juniors. It is a very pleasant and helpful move-by-move annotated game collection with diagrams every four ply, give or take two. I cover up the next diagram, follow the moves from the previous, and visualize how the board should look. I then compare my vision to the actual board.

(Incidentally, there are few books with enough diagrams so I can follow along without a board like this. Any suggestions (the only other one I have that is good for following along without a board is Pandolfini's Russian Chess)? More than 8 ply and it starts to become useless, so any books with 6-8 ply per diagram is ideal.)

I have been doing PCT tactics and strategy modules to help me stay sharp and learn some strategy. It's like I got so comfortable doing computer puzzles doing the Circles, it feels like I'm selling my soul when I don't do it. So I go along doing them like a zombie.

I'd really like to get a coach again, but can't afford more than 20 bucks a lesson. Anyone know of any good ICC coaches that are in that range?

Where am I? Who am I? Is that a candy bar? Cool. I think I'll eat it. I think the advice to follow my own advice (link above) is very good. I'll eat that too. Maybe it's OK that I'm improving buffet style for now, just doing whatever the fuck I want, sometimes nothing. Usually a game or some such. That may be enough for now.

Wednesday, November 21, 2007

Tactics: space, time, or spacetime?

Recently as I've been playing gambits and attacking more and becoming less materialistic, the dimensions of time and its cousin activity have become forefront in my mind. This post offers a different (for me) spin on tactics. My argument will be that we can view tactics as moves that will gain material unless your opponent is given extra moves (i.e., given extra tempi or time). As I'll explain, this focus on time seems a useful complement to the usual space-centered approach. In the end I try to fuse the spatial and temporal aspects of tactics....

Beware the following is somewhat theoretical. I'll be the first to admit it may not be useful to many people. However, I'm not guaranteeing it is not useful, so you might read on.

Let me first stipulate that for this post I use the word 'tactic' narrowly to refer to a set of moves that forces a gain in material or game points (so a forced stalemate in a losing position is a tactic as it gains you half a point).

Most approaches to tactical classification use a space-based approach. To pick on a representative sample, Tempo, in his many treatises on the subject, defines tactics as duplo attacks, moves that make two threats which cannot both be met, and which gain wood. This is a spatial concept: you are making threats at two different locations in space. You can point to 'em. But this isn't sufficient: as Tempo pointed out recently it must be that both threats cannot both be addressed at the same time. That is, the opponent would require extra tempi to meet the threats. Only in such cases does a double attack make a real tactical threat.

Next consider traps, which are not double attacks. Traps can be looked at spatially, though in a different way than the spatial view of double attacks: you are attacking a piece that has no escape squares, no place to run. But traps can also be looked at from a temporal point of view: you are attacking a piece and it needs more than one move to get to safety, it has no time to run. If given more tempi, this is a threat that could be met.

Notice the description at the level of tempi is the same for both traps and double attacks: they are both threats the opponent would need extra tempi to address.

Let's consider a third tactic that is neither a double attack nor a trap: leaving a piece free to be captured. This may be the most elementary tactic we all learn--take wood that is given to you for free. It can be looked at in a spatial sense: there is a piece there, and you can take it. It can also be recast in terms of time: the opponent has zero tempi with which to protect wood: if you were to give him extra tempi he would be fine. Hey, wait, that's what I've been saying for all of them!

All of these tactics have one property in common: the defender would need to be given some extra tempi to save material. A simple, general, exceptionless definition of a tactic? Or the violent ramblings of a madman high on Thanksgiving fumes?

Note before everyone lines up to ram a chainsaw down my throat, I am not saying I think the spatial perspective is wrong. It is quite practical and useful. You can point to two attacked pieces. Try to point to time. And now describe time to a scholastic beginner. Time is subtle and weird. Space is concrete and fun.

Finally, most importantly, obviously space and time interact in chess. The trapped piece needs extra tempi because of the spatial characteristics of the position. Indeed, this may be the most general way to characterize all tactics: threats (which depend on the spatial layout of the material) that would require extra tempi to block.

Saturday, November 17, 2007

Some post-circles thoughts and suggestions

It's been a little over a month since I finished the Tactical Circles. I've focused mostly on playing. My ICC rating went up to 1479, and is probably stable at around 1450 right now. Since I started at 950 I am happy with that for now, though my long-term goal is to reach 1500. The first month after the Circles I had one loss in about 20 slow games (all played with players within -100 and +300 rating points at the time). These wins were greatly helped by the tactical vision and confidence I gained with the Circles.

Anyway, I have some thoughts/suggestions for those who have yet to finish the Circles, how to handle yourself when you finish (I think these suggestions are probably not just applicable to people who did the Circles, but what the hey):
1. Play
Play a lot of slow games and analyze them, especially for tactics. As you are fighting to finish the Circles you will not be playing much, you just won't have time. When you do finish, don't worry about your rating. Just jump in and start playing.
2. Thought process
Create and use a thought process during the games. This was MDLM's third step in his training (vision drills, Circles, then thought process work). I was lucky in that I spent countless hours developing a thought process as I worked through the Circles, a process that is practical and feels natural to use.
3. Stay sharp
Use some kind of program to stay tactically sharp. Indeed, I would strongly recommend, as soon as you can stomach it after finishing the Circles, go to one of the tactics servers (I really like Chesstempo), and do 100 problems. This tells you your post-Circles tactical baseline. Then, do tactics there periodically to see how you are doing. If you ever drop too much, practice more, either there or with your original Circles problems so you don't become rusty. I just started doing this this week, unfortunately, and really wish I had done it right when I finished the Circles. My baseline at Chesstempo is around 1600.
4. Warm up
Before games, warm up tactically by going through 8-15 problems from the set of problems you solved for the Circles. This will put you in a tactical mindset before the game and remind you that weak squares don't mean squat if there is major wood to be grabbed.
5. Relax
You won't need convincing. When you finish the Circles, don't start chess again until you really feel like it. And when you start playing, don't freak out if you don't see results right away. Even MDLM didn't see results right after finishing his Circles: he had to use a thought process to try to make his new knowledge manifest itself in games. We all go through a kind of cognitive reshuffling when we learn something new about chess, and there is often a period of equilibration when we actually play worse as the new knowledge becomes properly integrated with our old knowledge. This is part of getting better, so don't fret.

Thursday, November 15, 2007

Chess hangover

Have you ever been an idiot? I have. Last night after my first loss (posted below), I got all raged up and played again. And again. And then I started playing blitz to work on my openings. At 5:45 this morning my wife comes downstairs to go running, and I'm lying there like a heroin addict still playing. She asks matter of factly, "Hi hon, playing a game?" I was like, "Uhhh what the hell am I doing? I'm such an idiot!" She was like "Oh, don't say that" and went of running while I crawled into bed like a little slug.

The aftermath? My rating went from 1470's to less than 1420 in slow games. Zero slow wins out of four. It was one of those nights when I was just tired, not playing well, but stubbornly wouldn't stop playing. Plus, the last two losses were against lower rated players I played to "boost my confidence." Yeah, that was a good idea.

Ugh...remind me not to do that again.

Wednesday, November 14, 2007

Another loss: unsound sacrifice

Ahh, Blunderprone why didn't I study your game more closely?

My second loss in a slow game post-Circles. I didn't play correctly in the opening (Smith Morra, I had white), and then I made an unsound sacrifice which he refuted, quite soundly. My decision was: either mess around on the queenside where all his material is aiming, or sac a Bishop and go in for a King beheading. I chose wrongly.

The Smith-Morra gambit is just a powerful tactical cannon. I started with it about a month ago, and haven't learned to wield it quite yet. There are so many cool little traps and tricks for white, but some of them are counterintuitive.

Pins aren't tactics

Meta-note: Now I think I may not have been all that off. See this post from Tempo where I commented.

Note: I now think I'm likely wrong about this. Addendum added at bottom of post.

This is probably obvious to everyone, but I just wanted to note that
pins aren't tactics. If a pin can't be exploited for material gain, how is it a tactic?

I look at a pin as the creation of a severe positional weakness on the pinned material, limiting its mobility to the line of the pin. It is this positional feature that makes pins a formal invitation to look for tactics. A kind of neon sign inviting you to attack the pinned piece, or grab for free some material that the pinned piece is defending. That must be why pins are included on everyone's list of tactical motifs.

I bring this up because in my previous game I missed an obvious grab of a Knight defended by a pinned pawn.

Now, skewers, those are real tactics.

Note added: I am starting to think I am wrong about this, or at the very least that things are a bit more subtle and complicated than I allowed in the original post. As discussed in the comments, if we define tactics broadly, pace Heisman, as the "science" of piece safety, don't pins sometimes fall into this category? Yes, most tactics involve immediately going up in material via some double attack or trap, but pins are often more subtle, they can be used to protect material, constraining the general flow of the game.

Such pins provide the tactical constraints that often dictate reasonable long term plans you can make. E.g., if I have his Knight pinned to his Queen, I may never capture the Queen, may never get the Knight for free. However, it may stop his Knight from grabbing my pawn on d4: it keeps his Knight immobile because of tactical considerations. So in this case the pin is a tactic because it directly bears on material safety. It protects my pawn that was attacked by his Knight because I have generated an even bigger threat that will be realized if he grabs the pawn. That's the answer to my original question, "If it can't be exploited for material gain, how is it a tactic?"

Katar, Glenn, and Loomis got me thinking I was probably wrong. More arguments in the comments section...

Note also I forgot when I wrote this to mention Wang's interesting post about where tactics ends and strategy begins.

Sunday, November 11, 2007

Thanks Tempo: Alapin French

Thanks to Tempo for showing me an opening that made the French fun for once: the Alapin Gambit. Below is my second time ever with it (I played it once in Blitz last night). I played recklessly, purposely erring on the side of ignoring threats rather than responding to phantoms if it was at all unclear to me. This lead me into a severely down and out situation, getting stomped on the queenside. I managed to miss a couple of obvious one-move tactics near the end but somehow managed to pull out the win.

It actually made it more fun to have only a vague idea of the moves. After I played Nxf3 I had nothing to go on, except I remembered Tempo saying "If he wants to trade his Knight for your dark bishop, let him." So that helped.

Here's the game in all its ugliness. I like the opening and will get a book on it. To the opening pedants out there, yes I know it isn't GM-level sound...but it was fun as hell!

Friday, November 09, 2007

Game, counting redux

Back from vacation, and back to blogging about my post-Circles games. Tonight I got the rust out as black playing the Scandinavian. Game is here with my novice lengthy annotations. I got lucky here. He blundered early. I think he was tired or something. A 1520 player moving quickly likely playing zombie chess plays worse than I do playing slowly. Good to know. There's hope for me yet.

I post the game only because of the elementary counting error I made on move 15. When a sequence of captures goes more than three or so half-moves into the future, I find it oh-so-easy to be lazy and evaluate the position based on a narrow search horizon: if it looks bad in such a horizon, I don't make the move (if it looks good, I analyze through to quiescience). That is an important assymmetry in my thinking I need to be wary of. Oh, I have some excuses. Partly I was a little rusty after almost a week away from chess. Partly I was worried about time (he had twice as much time as me on the clock). But the biggest reason is flat-out laziness, which will kill you in sharp positions where lots of exchanges are possible.

Note to self: move fast in quiet positions so in positions like this you can think through exchanges to quiescience. I took seven minutes thinking about an early move in a quiet position, which put me in a little bit of a time crunch, which made me abandon some of my responsibilities later in the game.

Finally, getting back to what seems like forever ago on the counting problem. I look at counting as necessary but not sufficient for competence in move selection. The most basic question for any candidate capture is 'It is safe? Will I lose material (or gain material) with this set of exchanges?' It is more basic than elementary tactics, for instance, but almost all tactics books include zero such problems. Glenn, in a nice challenge, posted examples where rudimentary counting at a single square is not sufficient. There will be tons of cases like that, of course! But there are cases like that with all elementary tactics. Don't fork his knight and bishop with your pawn if he is going to mate you in the next move. But that doesn't mean you don't need to be able to spot forks!

Sunday, November 04, 2007

Panov attack

Tournament game this week: my first use of the Panov attack is here. Extensive annotations, especially at critical move 12. I'm in San Diego at a conference, so I'm light with the blogging right now.

Thursday, November 01, 2007

Counting rules

Inspired by Tempo. This is a semi-formal treatment of counting, when to start a series of exchanges when a bunch of attackers and defenders are all pointed at a piece on a square. It assumes no in-between moves.

If you don't like technical weird geeky computer stuff, stop reading now before you get annoyed. Chess programmers probably have this stuff down flat. I'm not sure it works perfectly, but I think it does. I'd be curious to see if anyone can find a counterexample.

In a follow-up post in the next few days I'll give more concrete examples.

Definitions
  • A=attacker (to move). There are m attackers.
  • D=defender. There are n defenders of the material D0.
  • Ai is a number, the value of the piece (e.g., if the first attacker is a pawn, then A1=1).
  • A and D are sorted in ascending order, except D0, which can be any value (e.g., it can be a queen). For example, if A has a rook, queen, and pawn, then A1=1, A2=5, A3=9.
  • AxD(i) notation means attacker i takes defender i-1, and DxA(i) means defender Di takes attacker Ai. So, for instance, AxD(1) means A1 takes D0.
  • E(i), the ith exchange value, is the difference between AxD(i) and DxA(i). For example, if AxD(1) is 9 (pawn takes queen) and DxA(1) is 1 (something takes pawn back), then E(1) is +8.
  • Cum_Sum(k) is the cumulative sum of exchange values from E(1) to E(k). k must be less than or equal to m. Let’s define Cum_Sum(0) as zero, the pre-exchange state.
Before formulating the general rule for deciding whether to take D0, let’s consider some examples to familiarize ourselves with the meaning of the abstract notation system. I'm going to drop the subscripts in the rest, as I'm sick of adding them in manually.

Example 1: A1 is 1, A2 is 9 (attacking with pawn and queen). D0 is 9, D1 is 1, D2 is 1 (queen is attacked, and defended with two pawns). AxD(1)=+9 and DxA(1)=-1, so exchange value E(1)= +8. Should white continue? In this instance AxD(2)= +1 and DxA(2)=-9, so E(2)=-8. So the cumulative sum after two sets of exchanges is E(1)+E(2)=0. In sum, Cum_Sum(1)=8 and Cum_Sum(2)=0, so stop after the first capture. We have one suggestion for a rule: stop exchanging when Cum_Sum decreases in value.

Example 2: Attacker values 1 and 5. D0 is 1 and the other defenders are 1 and 9. E(1)=0, E(2)=-4, and Cum_Sum(2)=-4. So don’t go through all the exchanges. You might do the first exchange if it helps you positionally, the deciding factor with even exchange sequences.

Example 3: Attackers have values 3, 3, and 3. Defender has values 1 for D0 with 1 and 9 for the others. E(1)=-2, E(2)=-2, and E(3)=9 (in this case there are more attackers than defenders, so DxA(3)=0). So Cum_Sum(1)=-2, Cum_Sum(2)=-4, and Cum_Sum(3)=+5. Should you do the original exchange? Of course not! D doesn’t have to recapture with his queen at the end. If you were to simply sum the values of all the attackers and defenders, or to count the number of attackers and defenders, you would think you should start the exchange. But defenders are not forced to recapture. This suggests that you should only continue to attack if Cum_Sum is nondecreasing. But this isn’t quite right, as the next example shows.

Example 4: Attackers have values 5 and 9. Attacked piece has value 1, and defender has value 5. E(1)=-4, so Cum_Sum(1)=-4, you have lost material. But in this case, there are no more defenders exist so E(2)=5. Cum_Sum(2)=1, you have won a pawn. This is the real value of having more attackers than defenders: it isn’t that the exchanges in such a case always yield material gain (see Example 3), but that you get to extend your capture sequence one more than your opponent, so can accept a temporary decrease in the Cum_Sum value when your final unanswerable capture will then bump your Cum_Sum back into the black.

So all these examples suggest the following rule for capture sequences.

Capture Rule: Make the move AxD(i) if it will increase Cum_Sum(i). If Cum_Sum(i) would stay the same, then make the exchange if it improves your position in other ways. If your opponent has fewer defenders than you have attackers, you can accept a temporary reduction in Cum_Sum on your second-to-last capture if on your final unanswerable recapture the Cum_Sum goes positive again.

In practice, it isn’t all that bad to start by counting the number of attackers, defenders, and summing all the relevant piece values. This is the standard solution, which is essentially Tempo’s solution. It will often work well when there are fewer defenders than attackers. But then you need to reconsider the sequence of captures to make sure it isn’t like Example 3, that the defender can’t simply stop the capture sequence ahead in material. When you realize this, something like Cum_Sum is required, and helpful as it is pretty simple. It is simple because it only involves addition and subtraction, and you don't have to go through the sequence of moves in your head, only the sequence of arithmetic problems.

It isn't necessary or desired to think in terms of all this notation: the notation is merely required to acheive an appropriate level of generality. In a real game you just have to think in terms of values of actual pieces in front of you, so it is a lot easier. Typically you can just see that he can't stop the exchanges, that you can keep going one more if he has fewer defenders than attackers. So lots of this we already do intuitively without thinking. But if our goal is to make a general mathematical strategy that always works, this or something equivalent is necessary.

I will toy with using it explicitly to see how much it screws up my games.