Game between famous defenders. Strong players -- Idea behind move/ Line Questio?

Below is the link to Ulf Anderrson vs Tigran Petrosian played in Yugoslovia in 1983.

http://www.chess.com/games/view.html?id=341225#

 I have double sided question revolving around the innocent exchange on move, 21 Bxf5.

1) Firstly what were the reasons behind move 21 Bxf5? White has given up the bishop pair.. allowed castle and assisted in building a more solid pawn structure for blacks king side. What are the benefits for white despite ceding these things via Bxf5?

2) Secondly, exploring the alternate of 21. Bb5+ which springs to mind . Castle is lost and 21..Kf7 seems the reply. Are there lines where white is punished for doing this? It seems not from what I can tell, but does not necessarily provide a huge advantage either. Here is the only solid line for both sides that I can see (without engine assistance).

21.. Kf7 22. Qf4 Rf8  23. Ng4 Kg7 Seems to defend fairly effectively yet white maintains good pressure and piece activity

Comments? Other variations?

Bxf5 was most likely played to relieve the pressure on d4. Although losing castling is not the end of the world, I would not agree that making the guy lose castling with Bb5+ is useless (It has it's awkward points, like if a rook could get to c7, breakthroughs on the e file could be in the air), just that it's not nearly worth giving black so much pressure on d4. White certainly has nothing dynamic after Bb5+ so he would only do this if his game was solid. Here, after Bb5+ white would be rushed to do something drastic which isn't really possible, meanwhile his center is collapsing.

The light squared bishop on b5 or anywhere here seems much inferior to the f5 knight, because its only strategic purpose was to prevent castling but it's an otherwise useless piece, and even against a weakened king, it's not easy to attack with useless pieces!! If white's pieces were more potent, then black may want to avoid kf7 like the plague, but here he figures other things are much more important. Bb5+? would turn the game around in black's favor I think.

 I have reread this a couple of times. Great post

Interesting question! My take on the position was this (as everyone has pointed out) Bxf5 relieves the pressure on d4 and exchanges the so-so Bishop for the good Black N... it's a straightforward variation, pretty clear, and I think Andersson liked the general direction of the game Bxf5 was heading toward. 

Buuut... I had a hunch that White had better -- Bxf5 lets a lot of tension out of the position... but I couldn't see any better alternative than Bb5+ which upon further scrutiny doesn't accomplish much -- so, I plugged the position into Fritz running Stockfish 1.7 and lo and behold Stockfish has a clear preference for another move (Bxf5 is its second choice move and Bb5+ doesn't show up in the top three -- stockfish agress -- depriving castling rights in this position just isn't a big deal).

See if you can find Stockfish's choice -- IMO it's a classic strong computer move: tactical vision leads to a strong quiet move. Stockfish sees the tangled tactical balance between the Black N pinned on f6 and the weak white pawn on d4 and sees a repositioning that will allow it to add to it's edge. It took me a bit of chin scratching to see why the engine made this move and what it threatened.

Roll over between the brackets for Stockfish's move for WHITE instead of Bxf5: [Qf3]

My chin is about bloody by now, so I have no clue but I'll venture a shot with 21. Bc2.. 

I think you are too afraid to give up your bishop.

nice post. Allow me, however, to correct your paraphrasing of what is a great quote;

" Modern chess is too concerned with things like pawn structure. Forget it, checkmate ends the game" Nigel Short(C)

It was in response to JGpyth27's post about which alternative besides the move played or discussed here was suggested as best by the computer. What would you suggest?

I included the computer's answer in my post... it's in between the brackets at the very end of my post (#7) in white on white. if you highlight-select it you can see the chess-engine's first choice. Stockfish gave Bxf5 about a +0.60 eval... the first choice move got about a  +1.20

Oh i see.. I would be interested in seeing the variation as given your engine computed.. can you post it or edit it into the brackets?

Well... Bc2 actually does seem reasonable looking at it with a more open mind (still don't like Bb5+ though, puts the bishop too much out of place), with maybe the more ambitious idea of moving the queen, playing g4 to kick the knight, and be aggressive on the kingside/e file against the e6 pawn. I have no idea how it would actually work out, but it does seem worth looking at and I can see why the computers might like other moves than Bxf5. Bxf5 I think leads to a safe edge for white, but that doesn't mean a bigger advantage couldn't have been had with the more aggressive Bc2. The move seems to be quite consistent with either of these players' styles.

Black's game is very loose, and if he doesn't generate enough dynamic pressure on d4 or free his game with ...e5 (not only hard to achieve, but opening the position with a poor king position is always risky), he just has lots of weaknesses. Again, I'm looking at an eventual g4 as part of white's plans. The critical response is Bc2 Rc8, after which d4 can be defended with Nf3 in response to rc4 once the bishop is defended or moves (indeed it could go to b3, preventing rc4 even though it's otherwise out of play it may be worth it). Black can try to double on the c file, but what's more significant, the c file or black's loose, underdeveloped position? I'm guessing the latter.

But the point is giving up the bishop pair with a move like Bxf5 doesn't automatically give black anything special.

Haha, was a little confused for a moment what JG's cpu suggested: Qf3 (you had to highlight the brackets). Lol yeah, that's a very computer like move, just going right in for the kill if it's tactically justified, immediately looking at g4 maybe, not only kicking out the knight but exposing the f file right away.

I mean it does make sense: trying to tactically take advantage of black's loose king and stuff is the most dangerous idea, IF IT WORKS THAT IS!

Anyway, in response black can play ...0-0, trying to expose the queen on the f file, but white can probably be happy with getting in g4 (or maybe some other attacking idea I'm probably missing), with much more active pieces and quite a promising looking attacking position.

Petrosian really misplayed this game in general I think, starting with the opening.

Yeah, the engine has no problem going down the most complicated unclear line if it thinks it has an edge. I think most engines, stockfish included,  don't simplify unless it is clearly best to do so. 

Here's a Qf3 line from Stockfish... it doesn't do anything too dramatic, but it leaves Black a pawn down and terrible kingside pawns. 

Hmmm

*Edit* I actually looked at Qf3 for a while before making this thread.. but I didn't include it in the original post because of Nxd4 , 0-0 plan for black. Is there something wrong with Nxd4 that I'm missing?

*EditII* Ah I see it was just one of the lines. We'll did it give one where black plays Nxd4? That seemed to be the most sensible from what I could tell if 21 Qf3 

Wow somehow I missed the strength of a potential Ng4.

My human mind was absolutely shocked to see Re1. I'm sure the computer has it all calculated out, but I'd certainly like to see some variations myself!

I felt the exact same way about Re1! And I did follow the computer analysis down that line a while... it's ridiculously complex. White doesn't regain the material for a LONG time. But the attack is very strong... it's definitely worth a look. Engines make some beautiful moves, really they do.