Abstract Endgame Thinking and New Years Resolutions
Hey chess fans,
I want to present to you guys an endgame problem that arose from a game I played against Seth Talyansky, an Oregon master, at the Washington Class Championships that I find extremely instructional, both for the actual winning technique, and the abstract thinking required to arrive at the solution.
In this position, it is black to play and win. Explain to me the winning technique for black to convert this to a win, and give me the first move in realizing this plan for black. I really recommend you try and solve this before continuing, because I give a long explanation of how to solve this below. Don't scroll past the board if you don't want to see the solution!
Before I give the actual solution, I want to explain the thought process that you need to solve this problem. First off, I think there are two things we should note about this position:
1) Black is up a very dangerous passed b-pawn. Its promotion is the main goal for black.
2) White's only chances to save the game are based on perpetual check resources, especially along the h7-b1 diagonal.
Number 2 is especially important, because we want to win the game in as clean a fashion as possible. Especially if you imagine black's queen gets a check on that diagonal by playing Qb3 (blockading the pawn) then playing Qb1+, I can't play g6 because of h5, but if my king goes back to g8 or h8, and then he plays Qg6, with threats against d6 and also e8+e4 perpetual checks:
These complications are clearly not in our favor, so we want to keep the complications to a minimum. This also means we must take control of the h7-b1 diagonal ourselves.
So considering the importance of this h7-b1 diagonal, we can use the Ideal Square thinking technique, where we ask ourselves "If I could just pick up my queen, and place it anywhere on the board, where would I put it?" And more specifically, because controlling the h7-b1 diagonal is so important, the list of candidate squares we want to put our queen reduces the candidate squares to only squares on this diagonal: g6, f5, e4, d3, c2, and b1:
Obviously any of those squares will fulfill objective number 2 and prevent any checks from black. But considering we want to push the b-pawn as well, this filters this list of candidate squares to only squares near the b pawn because we want to support the promotion of the b-pawn as well. This means d3, c2, and b1 are the main squares we look at. And now we can go through each square in turn.
First we look at c2. The queen on c2 is probably the ideal square, as it both controls the diagonal and it protects b3, b2, and b1 as the pawn promotes. The problem arises though when we ask ourselves, "where can white put his queen such that it stops us from advancing this b-pawn". And then it becomes obvious, that white can still stop the pawn on b3 if he leaves his queen on a4, pinning the pawn to the queen, and its clear that c2 is not a stable square.
d3 suffers from a similar problem. We ask ourselves, "If black could place his queen anywhere to prevent the advance of this pawn, where would he put it?" And it becomes clear that white has a similar pinning technique by putting the queen on a3. If black steps aside to c2 then white goes back to a4, reaching the impasse we discussed before.
That leaves b1 as the only square left and clearly the best square to be on. On b1, the queen controls the diagonal nicely, and there is no way for black to ever stop the pawn from advancing all the way to b2. I even challenge you to try putting the white queen anywhere on the board in a way that stops black from pushing the pawn to b2. You will quickly find that there are none of the pinning defenses from before.This means we can fast forward our calculation and jump straight to this position with our queen on b1 and pawn on b2, and see if this position is convertible.
And it is pretty obvious now, he can step aside with either Qc2 or Qc1, and once black promotes the pawn, the new queen also conveniently guards the h7-b1 diagonal one last time, preventing any checks and perpetuals from white.
Now we know the winning technique:
1) Get our queen to b1
2) Push our pawn all the way to b2 because there is no configuration for white to stop the pawn to get all the way there.
3) Step our queen aside and promote the pawn.
And so now we work backwards. We know that we have to get our queen to b1, so we have to figure out the best way to get there. I do this by taking the starting square on f2 and the ideal square on b1, and start drawing connected lines which represent paths for the queen to get to b1 like this:
From here we see that one of the paths for the queen to get to b1 is through f1 which is also a check, giving us a free tempo, and therefore the correct first move is 1...Qf1+! (Qf5+ leads to the same thing) 2. Kh2 and 2....Qb1!, reaching the ideal square and the rest of the winning technique we already know.
I like this problem for the schematic thinking required to solve it. The solution combines prophylaxis (stoppping checks on the diagonal) with our own goals (promoting the b-pawn), to come up with the perfect square that does both jobs (b1), and then working backwards to figure out the best path for our queen to get to b1 (through a check on f1).
I think a common mistake to this problem, and the reason why I failed to see the solution is that I arrived at this problem and immediately start calculating moves. The problem with calculation here is that the number of legal queen moves is way too large, and unless we prune the number down with more abstract thinking, the problem is too hard to solve by calculation.
Instead, if you refrain from calculating and piece together the joint objectives of controlling the h7-b1 diagonal and promoting the b-pawn, this reduces the candidate moves significantly, from which we see that d3, c2, and b1 are the only real logical moves from which we can achieve both objectives. Once we realize that, then it becomes a matter of merely comparing the three squares, and we quickly see that only on b1 can I push my b-pawn freely to b2, and then stepping aside with my queen to allow promotion is a simple win. We used schematic thinking to whittle down the candidate squares to a more tractable number, and only then do we begin calculating how to reach the ideal square on b1.
Furthermore, I think this problem is difficult because we are naturally averted from putting our queen on b1. Normally we learn to put rooks behind passed pawns, so we would never think about putting the queen in front of the passed pawn. But general principles are always of secondary importance to the actual position, and vague endgame principles can not outweigh the necessities dictated by the specifics of the position.
If you didn't see the solution, don't feel too bad. The solution looks obvious now, especially with the way I've explained it to myself, but practically I think a lot of players would have botched this technique up. This was confirmed by the fact that I asked multiple other 2200+ masters to solve this problem, and not a single person managed to solve it. All of the random queen moves they gave me fail to the simple Qb3 to Qb1+ to Qg6 defense from white.
Just for completeness, I include the full game against Talyansky with my annotations here:
I want to end this post with some of my new year's resolution for 2017, and some changes I'm going to make with my approach to making progress with my rating.
Originally my plan for improvement was to try and lessen the experience gap that I have versus a lot of the other Washington state masters. In my life time, I've only played 90 tournaments, which compared to other masters in the area is very little as the average number of tournaments that all the 2000+ players play at is well over 150.
But I was discussing my plan to improve by "playing out" of my slump with one of my roommates and he pointed out that the physical number of games that I can play per year is limited. Even under the most grueling schedule where I play one tournament every two weeks (which is impossible), and we say that each tournament involves four games, that means 4x2 = 8 games per month * 12 = 96 games/a year at MOST.
96 games is not very many games, considering I've logged tens of thousands of online blitz experience on various chess websites, and from a time perspective I can not afford to be playing a high volume of games, as each tournament can often taking up my entire weekend. Furthermore, these masters in the area have logged over 60+ tournaments more than I have. If I average 10 tournaments a year, it will take my 6(!!!) years to match their level of experience, and considering everyone else is still playing actively, I will never catch up to them experience wise.
In 2017, efficiency will be key. I will only attend events where I am guaranteed 2000+ rated opposition every round, because those are the most educational games for me. This means in Washington, the next tournament I'll play in is probably the Washington State Championships in February, and I'll try to go to more of the Master Series tournaments. I might even travel to a few of the major US tournaments depending on my schedule.
So along with the idea of playing less chess, but focusing on the quality tournaments, I think I'm going to cut down on the amount of blitz that I play online, and shift more of that time to studying chess. Specifically I think I'm going to take more time to study positional chess and endgames, as I think those are by far the weakest points in my game, and probably of most modern players.
A third thing I will do is I will start working with a GM coach, something I have never done in my life. I originally got to 2200 purely through self-instruction, using a combination of books, engine analysis, and self analysis to improve, but I think I've hit a serious wall in my play that self-study can not cover alone. He's already made a number of instructive observations about my play, which I think are educational enough that I'll try to cover some of them in a future post.
Happy New Years,