KBN versus K

[Alex Holowczak]

Everyone was a bit confused about whether they could just keep trying ad infinitum given that the 50 move rule was hard to impose without an arbiter counting (and I don't think the arbiters were taking much interest - apologies if I'm wrong about that last point).

Everyone could keep trying forever if there was no arbiter counting, yes.

[Ray Sayers]

Here is the very famous game where GM Epishin failed to do it:

http://www.chessgames.com/perl/chessgame?gid=1533865

[David Blower]

I practice the W once in a while against Rybka ....

One of the things about KBN v K is that although it's impossible to find if you don't know what you're doing, it's really not that hard if you do. The most difficult thing about learning it is doing the 'revision' every now and then to keep the method in mind.

The trouble with practicing against a computer, in my experience, is that they just shunt the king to the wrong corner straight away. No doubt this is 'correct' in that it delays mate the longest, it's not at all how a real person would play.

I would guess that a real person would put the king in the centre of the board, and also use their king to try and attack pieces.

[Ray Sayers]

I just had my first practice of the triangles against Rybka and it worked a treat.

Also, Rybka didn't just run straight to the wrong corner; it did try to play against the pieces. But it's quite easy to push the king back and indeed if you try harassing the pieces too much you end up getting pushed back into the right corner faster.

I did it in 39 moves this way. I think I lost a couple of moves somewhere though!

[Arshad Ali]

The trouble with practicing against a computer, in my experience, is that they just shunt the king to the wrong corner straight away. No doubt this is 'correct' in that it delays mate the longest, it's not at all how a real person would play.

If your opponent moves to the right corner, he just makes life easier for you -- cuts down the number of moves and amount of time you need to administer the coup de grace. I'd have thought most club players know that when defending, they should move their king to the wrong corner. Perhaps your objection is a point Nunn made in one of his endgame books: that you need to practice against suboptimal strategies as well, because these do occur and an opponent who wants to trip you up may resort to this sort of thing.

[Clive Blackburn]

The trouble with practicing against a computer, in my experience, is that they just shunt the king to the wrong corner straight away. No doubt this is 'correct' in that it delays mate the longest, it's not at all how a real person would play.

If your opponent moves to the right corner, he just makes life easier for you -- cuts down the number of moves and amount of time you need to administer the coup de grace. I'd have thought most club players know that when defending, they should move their king to the wrong corner. Perhaps your objection is a point Nunn made in one of his endgame books: that you need to practice against suboptimal strategies as well, because these do occur and an opponent who wants to trip you up may resort to this sort of thing.

Obviously the defending side will try to make the game last as long as possible because that puts the pressure on the attacking side not to slip up and step over the 50 move limit. The best defence is to try to keep the king close to the centre of the board for as long as possible, while making sure that the king can always run to the "wrong" corner when it is forced to.

I have no idea why a computer would run straight to the wrong corner without being forced to.

[Arshad Ali]

I have no idea why a computer would run straight to the wrong corner without being forced to.

I think the point being made is that when the engine is forced to move its king back, it makes sure it moves it to the wrong corner. But playing against an engine can be tricky -- e.g., with K+Q versus K+R, a human with K+R keeps his rook close to his king. An engine might occasionally move it away from the king because that lengthens the winning process. Again pointed out by Nunn.

[Andrew Bak]

I once played against a friend practicing this endgame where we tried both sides multiple times with 3 mins for both sides. It's amazing how sub-optimal defence can throw off the attacker and force them to flounder around.

The only problem with practicing against tablebases or computers is that you only get optimal defence and are limited to the patterns that you have to recognise.

[Jonathan Bryant]

I think the point being made is that when the engine is forced to move its king back, it makes sure it moves it to the wrong corner.

That's exactly what I was saying, yes.

'Sub optimal' defence isn't necessarily easier to play against I think. Quicker in terms of moves, yes, but from the king in the corner I can make all the moves (19 to mate I think it is) with barely a thought. Against a less exact defence I have to use my brain.

[David Blower]

Surely the point Jonathan is making is that a human opponent will try and make things awkward for another human opponent.

Although it may take longer I can easily beat the computer with its strategy of placing its king in the wrong corner. A human opponent knows this is the standard training position, so might decide to do something different.

[Kevin Thurlow]

I managed it in a Surrey KO match a couple of years ago with 3 minutes against 2 - actually it started as BN v N, but his N was trapped on the edge and I had to win it first. I know nothing of triangles and Ws, I just did (roughly) what it said in an old endings book, probably Averbakh, maybe Fine. I did mess it up once but luckily had time to rescue it. It was rather pleasing when our GM pronounced himself impressed and thought I must have been practising it!

Mnay years ago John Sargent organised some Civil Service "Selected Endings" tournaments where you played your opponent twice in the same ending, but with opposite colours. KBN v K was included. I won the toss (important that, you don't want to show how it's done)so my opponent had the BN first and was only able to draw. When I had the BN, he ran straight to the corner I wanted...

[Christopher Kreuzer]

Many years ago John Sargent organised some Civil Service "Selected Endings" tournaments where you played your opponent twice in the same ending, but with opposite colours. KBN v K was included.

Out of interest, can you remember the other endings? Were QvR and R+NvR and R+BvR included? What would people include in such a tournament today if they had to pick ten (or more if you want) such endings? Would you include standard rook and king and pawn endings? And what about KBB v K? It is not that difficult, but it is surprising how even endings like that can get 'difficult' under time pressure!

Not sure if it is plausible in OTB play, but if you have five minor pieces (e.g. NNNBB or BBBNN), can you checkmate against KQ? I suspect in the latter the extra (third) bishop is useless, but with three knights and two bishops, not sure whether the queen is capable of a perpetual or disrupting mating nets long enough for the king to escape?

[Matt Mackenzie]

"Just" four minor pieces should usually beat a Queen. Fine demonstrated this long ago and others have backed him up.

[Kevin Thurlow]

"Out of interest, can you remember the other endings? Were QvR and R+NvR and R+BvR included?"

Well, we are going back 30 years! I think those were and things like R+P vs R, and simple K+P pawn endings (not that everyone found them simple). I seem to recall a R+6 vs R+5 against a weak player, where I won with both colours. Also, B + Ps vs N + Ps.

John had some great ideas for special events. (He also had selected openings and the feared all-play-all simultaneous, 5 players, thus 10 boards, 30 minutes each, being fit enough to run to each board helped.)

[Alistair Campbell]

Is there anyone here who can't checkmate with K+B+N versus K?

This is something I’ve practiced occasionally –(but only the W method – not the 3 triangle method, which I’d never seen before).

I’ve only once had the opportunity to do it in real life, and failed. I found myself in an ending of B+N+1 v R+3 with 5 minutes left on my clock. I won the 3 pawns, but as I tried to nurse my pawn up the board, my opponent sacrificed his rook for the pawn, and I ran out of time floundering around. I seem to recall that my opponent stubbornly refused to do what he was meant to do, like head for the wrong corner, and I used up my little remaining time trying to figure out where I was in the “pattern”.

I note Wiki says this ending occurs about once in 5000 games. Does that seem right? I’ve only witnessed it 4 or 5 times.
Do people also practice, say, Q v R or R + B v R or some of the R+P endings?

And given there is chat of adjournments elsewhere, would you adjourn K+B+N v K and would you expect it to be resumed?