Monday, April 08, 2013

Blue or Red Pill? XVII

White to play

Without Carlsen and Kramnik's mutual last-day implosion, Ivanchuk's win over the Norwegian kiddiwink in round 12 might have been one of the most significant games of the 21st century. As it was the tiebreaks went in Magnus' favour and the lost half-point didn't really matter.

As for the game itself, cutting off the king, a fundamental concept in these endings, turned out to be the key to Black's victory. Chucky, needless to say, wasn't unduly troubled although for the rest of us - well, me - things weren't so straightfoward.

Anyhoo, today we have a position based on a similar theme. White's got the pawn, but Black's rook and king are actively placed. It's your move and you're down to increments so you've got thirty seconds at most to choose. What are you going to do?

The choice is yours.

BORP Index
Rook and pawn Index

(position taken from de la Villa's 100 Endgames You Must Know)


Anonymous said...

The trouble with Rc1 is that White's progress seems to be slow so that the Black King has time to scuttle back to Kd7 and allow Rc8. For example, Rc1 Ra8+ Kb3 Kd5 b5 Kd6 Kb4 Kd7 b6 Rc8 Rxc8 Kxc8 Kb5 Kb7=

After Rh5 Black seems stuffed. The naive Re8 Ka4 Re5 Rxe5 Kxe5 Kb5 Kd6 Kb6 just seems a trivial win and I can't see any real improvements.

Dan Schmidt said...

My recollection from Rosen's book Chess Endgame Training is that Rc1 is insufficient with the pawn still in White's half of the board. White's king can't move forward to dislodge the rook without dropping the pawn that is too far behind him. I think Black's king just has stay on d5 and d6 so that White can't play Rc5 and then b5.

So, given two choices, I guess I'd play Rh5 by process of elimination, although if it is in fact correct I'd have to work out the rest from first principles.

John Cox said...

Rh5, bien sur. It's a general principle in rook endings (or so I once read somewhere) that cutting the defending king off on a rank is generally more effective than doing it on files. Besides, as someone said above, Rc1 is a fairly trivial draw - 1...Kd5, and what's White's next move supposed to be?

Jonathan B said...

Yes, you're all right - although it's interesting to see the different thought processes of getting there.

In essence, the defending cut off by one file is not enough for a knight's pawn in it's own half of the board. If the defender is well placed, that is (as the rook and king are here).

So 1 Rh5 can't have a down side. It wins, although I think 1 ... Ra8 is a tougher defence than 1 ... Re8.

Anonymous said...

Houdini couldn't see a win after Rh5 analysing to 30-ply. Also in my Rc1 Ra8+ Kb3 Kd5 b5 Kd6 Kb4 Kd7 b6 Rc8 line now Rc5! wins apparently. Back to the drawing board I think.

ejh said...

The difficult thing is working out how White makes progress after 1.Rh5 Kc4 2.Rc5+ Kd4. Obviously I can look it up if I want, but if you don't already know, it's not so easy to see. (Also, one is naturally likely to think - if 2.Rc5 is forced, why not 1.Rc1 putting the rook on the c-file straightaway?)

Anonymous said...

I have looked it up and I would try to find a rationalisation for Rh5 instead of Rc1. Distance to the queening square of b8 is always important and Rh5 keeps the King further away once White manages to get the b pawn to advance.

When you get these endings in practice, it's almost always against the clock even with a 30 second increment, so knowing what to play without much thought is often necessary.


Jonathan B said...

After Rc1 ...Ra8 loses. You should leave it on the b file to stop the pawn advancing.

Also, it's not necessary to go to c5 with the rook after 1 Rh5.

Sorry for brevity but its not easy to type comments on an iPhone

Jonathan B said...

Ps - engines are rubbish at rook endings so houdini not knowing 1 Rh5 wins doesn't surprise me in the least.

30 ply is no substitute for technique.

AngusF said...

engines are rubbish at rook endings so houdini not knowing 1 Rh5 wins doesn't surprise me in the least.

I suspect that engines aren't that good at endings at all.

My guess is that they have little or no algorithms to win/draw endings by technique.

Access to tablebases helps though only for positions with a limited number of pieces. (I see that ChessOK have recently announced access to 7-piece tablebases).

Anonymous said...

Of course AngusF is largely correct but a 30-ply analysis beast is still a behemoth of an assistant for checking variations. As humans we must try to establish underlying concepts and rules-of-thumb for playing out these positions. However as someone who's had access to chess computers for over 30 years I have to say Houdini is incredible. (And I'm only a cheapskate with the free version.) It was able to see mate-in-27 from a random KBNvK position for example.

ejh said...

Also, it's not necessary to go to c5 with the rook after 1 Rh5.

Unless I am missing something, Nalimov disagrees with you.

Jonathan B said...

Apologies, I didn't read your comment properly.

I meant 'in general' you don't have to place the rook on c5. After the specific position that we get after 1 Rh5 Kc4 I'd certainly trust Nalimov's opinion over mine.

Jonathan B said...

Although, as it happens, it's clearly the only move.

Dan Schmidt said...

When I tried Houdini 3.0 on this position it chose Rh5 within a few seconds and by the time it got to 30 ply its eval was +3.0.

I don't know what version you (Anonymous) was using, but there does appear to be an issue with Houdini 3.0 where if it thinks you don't have a legit copy it gives weird evaluations, especially in the endgame. (I am not accusing you of anything; after buying it I had to download it a second time before its evals made any sense.)