Monday, August 04, 2008

KQ vs. KR Mating Tutorial, Part 1 - Philidor's Position

I think the time has come to try putting it all into a tutorial. Or a series of posts, as the ground to cover is just too much for a single post. It would no doubt fill a small book, and I'm not gonna go that deep into it. Just the essentials which will allow you to figure the rest out, just like I did.

The amount of different types of positions you need to be able to play is quite extensive, so I can't possibly cover all of them, nor even know all of them thoroughly. But I'll go through the essential positions, trying to point out what I found to be important and practical. The rest you'll need to fill in by yourself, by training these positions over and over and over against an engine. The reason is because, even if I managed to write it all down, it wouldn't be much use to anyone. KQkr mate is not something you can learn by reading some kind of directions, no matter how complete, it's far too complicated for that. Instead you need to get your hands dirty, work on the typical procedures until you know them like the back of your hands. Build up intuition for what kind of candidates to look for, as well as pattern recognition for the various little tricks there are, and become able to jump from technique to another without missing a beat. Which is very typical for this mate, as you'll always need no jump between 3, 5 or more different techniques to get where you want to.

This is not an easy mate, and will require quite a lot of effort to learn. But I firmly believe anyone can learn KQkr mate, it just takes time & elbow grease. It took me 3-4 weeks to get to this point, and a few hundred repetitions mating from the essential positions. It's not gonna happen quickly, and you'll need to rehash everything multiple times.

So why should you do all this work? It's not like you'll run into KQkr mate on regular bases, far from it. - Well. Just like KNB mate, it's the side-effects that count, not the mate itself. The journey is more important than the destination. - It teaches you a lot of piece coordination, how to really operate the mighty queen, and how to defend with the rook. During these weeks the constant KQkr drilling has made a tremendous impact on my board vision, as the drills require you to be aware of the whole board, diagonals, individual squares and the colour of them. It's been like a veil had been lifted from my eyes, I can see a lot more in open positions than I used to. It's definitely worth it, from the practical point of view.

Okay, enough babbling, let's get down and dirty. The first position to learn is Philidor's position, which is a forced mate from as early as 1777:

This is what you're aiming for, and also where you'll end up if you opponent plays the best defence. - What's noteworthy in this position, is that if it's black to move, he's in zugzwang and will drop a rook (I'll get back to that in a minute). But, if it's white to move, as it often happens, you need to lose a tempo. The way to do that is to triangulate with checks 1.Qe5+ (extremely typical check in other KQkr positions as well, get used to looking for these), 1...Ka7 2.Qa1+ Kb8 3.Qa5 (1...Ka8 2.Qa1+ Ra7 gets mated with typical 3.Qh8#). Now it's the same exact position, but black to move.

black's options: 3...Kc8 loses directly to 4.Qa6 pinning the rook and mating. So only rook moves are possible, but every single one of them drops the rook and gets black mated. Let's see how:

3...Re7 4.Qd8+ forks the rook.
3...Rf7 4.Qe5+ Ka7 5.Qe3+ Kb8 (Black king has to return or get mated at once, another typical check from centre) 6.Qd8+ forks the rook.
3...Rg7 4.Qe5+ forks the rook.
3...Rh7 4.Qe5+ Ka8 5.Qa1+ Kb8 6.Qb1+ forks the rook. (This is the main way to move the queen from wrong colored diagonal to right one. Learn it, you'll be using it a lot.)

That took care of the 7th rank, now let's see the other escape direction, the b-file:

3...Rb3 4.Qe5+ Ka7 5.Qg7+ (the good ol' diagonal swap) Ka8 6.Qg8+ forks the rook, or mates if black tries Rb8 to interpose.
3...Rb2 4.Qe5+ forks the rook
3...Rb1, the best defence, falls to diagonal swap, what else. 4.Qe5+ Ka7 5.Qd4+ (work out why that was needed) Ka8 6.Qh8+ Ka7 7.Qh7+ Kb8 8.Qxb1+ Kc8 9.Qb7+ Kd8 10.Qd7#

That's it, black has no other sane moves, Game Over.

The things to take home from the Philidor's position, are the importance of the typical check from the centre and the diagonal swap. Drill these lines against Fritz until you can blitz them as fast as you can move your mouse, to the point where you'll be constantly visualizing several moves ahead, and your hand just goes through the motions. Notice that in none of black's tries the king escapes, instead you can always mate directly after winning the rook. So if you're forced to mate the lone king by chasing it accross the board, you're doing something wrong. In a real game you'll likely need most of the 50 moves available, so it's important that you can mate efficiently.

I hope I didn't screw up the notation, and that the diagram actually shows.

RHP: 1988, 298 games, +204, =17, -77. 1985, 7 games +7, =0 , -0.

Friday, August 01, 2008

Mating with KQ vs. KR

During the last weeks I've been training the intimidating Q vs. R mate, and yes, it's every bit as hard as it's said to be. Maybe even more so.

So why is it so hard? Well, to begin with, there isn't any clear cut single technique to it like in the KNB for example, which btw. seems like a walk in the park in comparison. Instead you need to learn a bunch of typical positions, learn the ins & outs of them, then recognize similarities on the board and jump from one technique to another, in a way which seems quite random. And with a bit of luck, you'll manage to fumble your way into a forced winning of the rook.

Basically, you'll need to drill it over and over with fritz, build up a huge repository of typical maneuvers, and sort of feel your way through it. I think I've got it down fairly well now, as I can win it from random positions against tablebase most of the times. But I still don't know half of the time how I got into a 'familiar position', or why I suddenly dropped off from the beaten path. I just make moves which seem 'reasonable', and somehow end into some recognizable position eventually. So there's still lots of ground to cover, but I'm starting to get there. In slow games I think I can pull it off, but blitz is still out of the question, at least against tablebase. Agains fritz without tablebase I can blitz it, but mostly because fritz plays it very similarly every time.

I'll try to write some kind of a real tutorial at some point, but there's so much material to cover it'll take some time. And also my knowledge about it isn't really focused enough to put it into words yet either.

But anyway, here's an example against the ICC KQkr bot:

[Event "ICC"]
[Date "2008.08.03"]
[White "wormwood"]
[Black "KQkr"]
[Result "1-0"]
[WhiteElo "1329"]
[BlackElo "2200"]
[TimeControl "300+12"]
[FEN "8/7k/8/8/6Q1/8/r1K5/8 w - - 0 1"]

1. Kb3 Ra6 2. Qd7+ Kh6 3. Kc4 Rf6 4. Kd5 Kg5 5. Ke5 Rf2 6. Qg7+ Kh4 7. Qg6 Rf3 8. Ke4 Rf2 9. Qd6 Kg4 10. Qd1+ Kg3 11. Qg1+ Rg2 12. Qe3+ Kh2 13. Kf4 Rg7 14. Qe5 Rg3 15. Qd6 Kh3 16. Qe6+ Kh2 17. Qe5 Rg2 18. Kf3+ Kg1 19. Qa1+ Kh2 20. Qe1 Ra2 21. Qe5+ Kg1 22. Qd4+ Kh1 23. Qh8+ Kg1 24. Qg8+ Kh2 25. Qxa2+ Kh3 26. Qg2+ Kh4 27. Qg4# 1-0

(2008.08.03 - Updated the above example to a new, more streamlined one. This one I managed in a little over a minute against a tablebase. Getting pretty close to being able to blitz it.)

The basic idea is to force the defender in zugzwang, so that he'll have to separate the rook from the king. After which the rook should drop fairly soon. Checks are usually just tools to get into a zugzwang, and the most important (and hardest to see) moves come without a check. Up to a point where you start looking for non-checking moves for candidates rather than those familiar checks. But naturally there are some extremely important checks as well.

I doubt any of this makes much sense without diagrams and specific variations, but hopefully I'll get to that some day.