Find the computerish idea

Technical questions regarding Openings, Middlegames, Endings etc.
MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Find the computerish idea

Post by MartinCarpenter » Tue Jan 14, 2014 4:25 pm

Something gently mind boggling that turned up when going through the ashes of a rather unmemorable evening league game.



Black to play and..... get searched for electronic equipment :)

It doesn't win outright by any means but does seem to retain rather more workable chances than anything else. Bonus points if you can somehow explain the 'logic' behind it and why it is seemingly the only move that does it.

In practice I'd been losing for ages and was a bit shocked to have escaped with the rather obvious sac that had just landed so just chopped on e5 and offerred a draw. Odd season. I'm playing quite awfully a lot of the time but having outrageous luck to compensate. Can't remember anything like it.

Ian Stephens
Posts: 117
Joined: Wed Aug 17, 2011 10:59 am
Location: Merseyside

Re: Find the computerish idea

Post by Ian Stephens » Tue Jan 14, 2014 6:11 pm

Hello Martin, at first glance I would try ...Rf5 blunting the Queens scope and threatening ...Rg5+ or is that too human?
Ex-President of Liverpool Chess Club, now mere Tournament Controller and Chief bottle washer.

User avatar
Christopher Kreuzer
Posts: 8806
Joined: Fri Aug 06, 2010 2:34 am
Location: London

Re: Find the computerish idea

Post by Christopher Kreuzer » Tue Jan 14, 2014 6:15 pm

Er, something like 1...Kf7, trying to bring the rook to h8? I don't think it works, but it is the only 'crazy' move I could find that might work, but 2.Qe2 seems a strong reply, so I'm not sure what the idea is if it is not that.
Last edited by Christopher Kreuzer on Tue Jan 14, 2014 6:16 pm, edited 1 time in total.

User avatar
Matt Mackenzie
Posts: 5206
Joined: Tue Mar 31, 2009 11:51 pm
Location: Millom, Cumbria

Re: Find the computerish idea

Post by Matt Mackenzie » Tue Jan 14, 2014 6:16 pm

1....Rf5 was my first thought too - but a bit *too* obvious, perhaps? :?:
"Set up your attacks so that when the fire is out, it isn't out!" (H N Pillsbury)

David Gilbert
Posts: 962
Joined: Sat Apr 04, 2009 10:03 am

Re: Find the computerish idea

Post by David Gilbert » Tue Jan 14, 2014 7:10 pm

What about Rf6 and if White allows follow up with g5, threatening Rh6. Just a thought?

Brian Valentine
Posts: 574
Joined: Fri Apr 03, 2009 1:30 pm

Re: Find the computerish idea

Post by Brian Valentine » Tue Jan 14, 2014 8:17 pm

It looks as though f4 slows down the Rf5 idea, so I'm wondering about ...Kh7 and if Re3 then Rf4 ( ->h4 etc)

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: Find the computerish idea

Post by MartinCarpenter » Wed Jan 15, 2014 10:28 am

Yes, Rf5 is far too human. You'd probably even consider it a normal candidate move ;)

Rf6 is actually the move so well done(?!) to David. The main idea of it isn't to attack though. Its to avoid swapping queens......

After Rf6 white seemingly has to go Qe2 - else the attack does get dangerous - and then black can play Bxe5 Qxe5 Qxf3 and retreat to f5 after Qe2. This is, I believe, somewhat non obvious :) If you take on e5/f3 directly then Qe2 squaps queens.

Why you have to start with Rf6 and not something else, like g5 (which white can't take of course) or Kh7 etc is not very obvious even after briefly looking at its main lines. Probably there isn't really any logic, just a bunch of concrete lines.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Find the computerish idea

Post by Roger de Coverly » Wed Jan 15, 2014 11:02 am

MartinCarpenter wrote: Rf6 is actually the move so well done(?!) to David. The main idea of it isn't to attack though. Its to avoid swapping queens......
It's going to depend on your engine, computer and settings as Bxe5, followed by Qxf3 is the obvious continuation even to engines. Having "forced" Rf6, there's next to nothing between the evaluations.

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: Find the computerish idea

Post by MartinCarpenter » Wed Jan 15, 2014 12:19 pm

Ah, Stockfish was very firmly about half a pawn better for Rf6 than anything else ~-0.9 vs -0.4.

So I was presuming it should be something real. Could be an evaluation engine quirk though, yes :) Fast computer so it might be some deep tactic that justifies it vs some of the other moves.

Roger de Coverly
Posts: 21301
Joined: Tue Apr 15, 2008 2:51 pm

Re: Find the computerish idea

Post by Roger de Coverly » Wed Jan 15, 2014 12:45 pm

MartinCarpenter wrote:Fast computer so it might be some deep tactic that justifies it vs some of the other moves.
Horizon effect or different weightings then. The hidden idea presumably is the threat of a transfer of the Rook to the h file in support of a mating attack.

Anand, I think it was, cautioned against over-reliance on engine assessments. He reckoned five years later you got a different answer as the evaluations got more sophisticated and the machines faster and deeper.

MartinCarpenter
Posts: 3044
Joined: Tue May 24, 2011 10:58 am

Re: Find the computerish idea

Post by MartinCarpenter » Wed Jan 15, 2014 10:03 pm

'Static' computer evaluations tend not to mean terribly much. Picking on a move like this means there's a reason. Not always a good one but fun to unpick the (twisted) computer logic whatever :)

I'm not totally sure about this being strictly right as its based on a bunch of rook endings. Computer aren't great at evaluating those and I'm no doubt quite a bit worse....

Of the moves: Even thinking of 1.. Rf5 clearly shows carbon based thinking as white can just play Rxf5 ;) 1 .. g5 probably should be in range of humans too as white can go 2 Re1e3 Bxe5 3 Rxe5 Qxf3 4 Rxg5.

1... Kh7 is more forgivable. The problem there is that white can go 2 Re1e3 and meet 2... Bxe5 with 3 f4 Qg4+ 4 Rg3 Qxf4 5 Qxg6+ Kg8 6 Qe6+ Qf7 7 QxQ+ RxQ 8 PxB etc. 'Obvious'....

So maybe 1.. Rf6 is a waiting move to prepare for 2 Re1e3.... After 2 Re1e3 Bxe5 3 Rxe5 Qxf3 4 Qe2 QxQ 5 RxQ you can go Kf7 and the rook is very well placed for the ending with f5/b6 to pick from. Much better than on f8.

The most likely horizon effect candidate would be that 1.. Rf6 2 Qd2 g5 wins. Takes my comp a few moments to work through that. No wonder really - after 3 Qxg5 you have to go 3.. Bxe5 4 Rxe5 Qxf3 or 4 Pxe5 Rh6. Not clear cut wins.

I spent a while rather puzzled why it felt that 1 .. Rf6 2 Qe2 Bxe5 3 Qxe5 Qxf3 4 Qb8+ Rf8 5 Qg3 QxQ+ 6 PxQ was a non trivial improvement on 1.. Bxe5 etc. That's actually the same position except for whites f2 pawn being on g3. However the clear second rank means that 6 .. Rf6 7 Re5 Rb6 8 Rxd5 Rxb2 9 Ra5 a6 10 Ra4 b5 is now winning due to the b4 breakthrough. So white has to defend a bit more passively instead. Its a seriously hard game sometimes :)