All the rest of the moves are the same except for a slight tranposition at the end, chesslab has black transposing moves 39/40. Oh and also whites 38th is c4 not c3, but still the position doesnt change.

This game is the only game in the chesslab database for Spassky/Polugaevsky in 1958 it was played at Riga so it must be the same game they are referring to.

[ Andy Soltis gives the following (somewhat superficial) analysis: 20...Nxa4; ('!') 21.Nf5!! Rxd1+!?;
This seems like a fairly natural move for Black, and it also follows general principles as well.

"When you are attacked, trade as many pieces as you can.")

<( One book gives the line of: 21...Rfe8?!; (but now) 22.Bxf7+!!, 22...Kxf7!?; 23.Rxd8! Rxd8; 24.Nd6+ Kg8; 25.Qf7+ Kh8; 26.Qxe7, ( ) but White is clearly winning here.

<One simple win is if Black plays ...Bc8; - to keep from dropping the Bishop on b7 - White simply plays QxR/d8+! and Nf7+.> )

22.Rxd1 Bc5!?; ('?!/?')
I guess Black is looking to activate this somewhat passive piece. *******

---> ( It seems much better to retreat the Bishop to the d8-square. >/= 22...Bd8!; An odd-looking move, but a play which is natural when you realize that if Black allows Qg5, he is simply lost.)

23.Bxa4!?,
This is the <correct> move according to A. Soltis ... he even praises this play for White. ('23.Bxa4!' - Andy Soltis.)

***

( Much better is the surprising move - in this position, of: >/= 23.e6!!, "--->" and according to IM S. Soloviov, White will have a great - and possibly winning - attack. )

***

23...Qxa4?!; (Really - '?')

I guess this is the main idea - but Soltis does not discuss any alternatives here for Black. (Black blindly falls in with White's plan. But the second player has a better move that I wrote into the U.S. Chess magazine about ... but never got a reply.)

***

[ MUCH better was the move: >/= 23...h6!; "~" (in-between) and two week's worth of analysis never revealed any forced wins here for White. ('23...h6!!' - IM Sergei Soloviov.) ]

***

24.Rxd8 Rxd8; 25.Qg5, " " and thanks to the mate threat on g7, Black will pick off a whole Rook here. ) <---

*******

Houdini_20_x64: 26/75 1:13:04 49,108,248,336
+0.32 13.Bxf6 Bxf6 14.Nf5 0-0 15.Nxd6 Rab8 16.e5
-0.29 13.Bd5 Bxd5 14.exd5 b4 15.Nb1 Nxd5 16.Nc6
-0.32 13.a3 0-0 14.Kb1 h6 15.Bxh6 gxh6 16.f5
-0.51 13.Qe2 0-0 14.Kb1 h6 15.Bh4 Qc7 16.e5
-0.61 13.e5 b4 14.Ba4+ Nxa4 15.Nxa4 Qxa4 16.exf6
-0.73 13.Qe3 0-0 14.a3 b4 15.axb4 Qxb4 16.Kb1 a5

Position with small advantage for white similar to 13.Bxf6 was arrived at anyway because of error 13...dxe5?! and 14...Bxf6 error, black could (should) have taken with pawn. For example, 14...gxf6 15.fxe5 0-0-0 16.exf6 Bxf6 17.Qf4 (or, maybe better yet 16...Bd6 17.Qf2) and black has comfortable game with slight edge, position is typical and should be familiar to Najdorf player. Interesting to see 14.Nf5 move (line above with 13.Bxf6) in light of recent kibitzing on Velimirovic vs Sofrevski, 1965 and related pages shown in links there.