Training Exercise

A couple of posts back I went over part of a game Jim McTigue and I played at the club last Saturday. Analysis from both carbon- and silicon-based life forms went into that post. Along the way we determined that Black should have emerged with a much better position in several lines.

But those advantages are only theoretical unless a player can convert them into wins. So as a training exercise I have been trying something I've been meaning to do for years: play the position out against a strong program.

I had had the idea many years ago while reading books & magazines. Invariably some GM would reel off a long string of analysis and end it with "White is winning" and leave it at that. Many times it was clear enough. But many times it wasn't at all clear, especially not to a mere B-class player such as myself. Later, after playing programs became strong enough, I thought, "If that's a winning position, I should be able to beat Crafty/Fritz/[flavor-of-the-month] most of the time from that point." Now that's a tall order, because even back then Crafty itself was much, much stronger than I. It would certainly punish all my tactical mistakes at the very least.

But that's the whole point of training, isn't it? I've started doing this off and on in recent months, but now I'm going to try an make this a regular feature of my work.

Anyhow, below is a position from the analysis of the game from Saturday. (I believe there's probably at least two other such positions one could use for training from that analysis.) Stockfish 6.64 evaluates this as a -2.71, meaning Black is winning.

White to play, Black to win

5k2/1p1n2p1/p3p3/3pP3/8/3B4/PP3rPP/R2K4 w - - 0 9

So far I'm not doing well. I keep giving up easy draws, except for the one time I conspired to lose! (That can be done by leaving White with both his g- and h-pawns, and then moving the Black king too far into White's queen-side position.)

Give it a try. I'm probably going to try playing the White side when I have the time, to see how the machine handles this winning position. But I'm only going to do that after a couple more tries as Black tonight, after everyone else goes to bed. I'm not finished yet!

Games from the Clubhouse: Frustration Edition (Updated)

Today Jim McTigue and I got into another French Defense. It was a Classical line, I believe, and at some point Jim got in Nd6, but a bit prematurely. Things were going my way when we reached the starting position below. I goofed up, as I'll explain, and then lost my cool and missed a clear win.

Afterwards we got into a prolonged analysis session with Connor and Norm, and perhaps Paul briefly. Ideas were kicked around, holes were found in this idea and that idea, but I was convinced that after ...Rf2 I had a stone cold won position. Then El Supremo plugged the position into his cellphone, and Stockfish began busting one idea after another.* Now came the claim that WHITE was winning.* GRRR.

Anyway, I've run it at home now, and it looks like claims of White's advantage are overblown - Black had an advantage through most of the critical parts, though not always a huge advantage. Still, quite frustrating. I'm putting the analysis in the window below. One thing I didn't add: if White plays 4 Qb4 to cover the d2 square, then Black gets in the cute shot 4...Rcxc2!!

If someone with a faster computer & better analysis comes up with anything, post it in the comments below and I'll add it to the body of the post if it changes anything. Man, what a miss!

* These sentences have been edited.

UPDATE: Paul Leggett has commented; he writes:

I think the challenge with this position is that one is tempted to evaluated statically, based on structure and types of pieces, when the reality is that it is about tactics and development. Statics can be assessed quickly, whereas dynamics often take time to full appreciate.

The idea of ...Rxc2 in response to white's Rf1 move is simple if one looks at what the Bd3 was doing, but we have to train ourselves to think that way (I don't do it naturally, but I am working on it). IM Martin Weteschnik has a method he calls "Status Examination" where recommends looking at what each piece is doing individually. In this case, after Rf1 the Bd3 is overloaded- it is protecting the Rf1 and c2 pawn at the same time. At that point the tactic becomes easy.

That said, I doubt I would have noticed it, either.

I am starting to see a pattern of overloaded defenders in my tournament games and at the club, and I think it may be a neglected tactical motif. We all tend to focus on forks, pins, skewers, etc, but the essence of all tactics is the double attack, and an overloaded defender is highly susceptible to it.

I decided to run the game through the Fritz "full analysis" mode, using Komodo 8 and allocating 5 minutes per move. All of the comments are from Fritz/Komodo, including the comments in English- I added nothing.

It seems that black was clearly better, and that white's only salvation was in the repetition available after black left an escape hatch for white. Here is the link to the game.

The extent of my evaluation was that I was threatening mate in one and was certain I could find a way out of the checks! What I missed, ultimately, was the Qb4+ idea. Pure blindness on my part, and another manifestation of my biggest weakness as a player: I keep overlooking ideas for my opponent. (For the record, Komodo largely confirms prior analysis, so that should be the bottom of this particular position. The resulting endgames, though....)

Games from the Clubhouse: Endgame Edition

A few months back Paul and I played a game that became a topic of lengthy debate afterwards, with Garry also contributing. At some point, the following position was reached, though I have no idea how we got here from the game.

Paul - Todd (analysis)

 

Black to move

We spent a lot of time looking at this particular position. I didn't try to recreate the analysis the night I put this in my database several months back, and I'm certainly not going to try to do so now. We eventually settled on 1 ... Qg5 as the best opening salvo for Black.

But then Paul ran it through Komodo 8 and "the cloud" for an entire night. The best move according to those worthies is 1 ... Qg7. White's problems are that Black's h-pawn is too fast, and if White removes the c5 pawn his own King becomes too exposed - and maybe even if White doesn't take the pawn. Evals ran from -1.50 to -1.80.

It seems like these kinds of very tricky/trappy complex endgames come up more often now in the games I see at my level than in the past. Perhaps it's merely a function of where & who I'm playing (and see play) these days, or maybe I'm just better at finding this stuff now. Still, I don't remember seeing so many complicated endgames in the past.

(Am I the only one that thinks of Aristophanes every time "the cloud" gets mentioned? Yes? Nevermind then.)