In Andrew Soltis's book The 100 Best Chess Games of the 20th Century, Ranked, his number 1 game is a correspondence game between Yakov Estrin and Hans Berliner, from the 1965–1968 World Correspondence Championship. This is of course a provocative choice, and Soltis, when making his selection, evidently believed that Berliner's play was sound. Soltis quotes Berliner as offering to play anyone from the position after Black's 9th move (as Black) for stakes of $1,000, provided he was given draw odds.
I have a rather old engine, but I decided to see what it thought of the game. It says that Estrin could have significantly improved at move 12. This critical position is diagrammed below.
[FEN "r3k2r/p1p2ppp/3b4/1p1n4/3Pp1bq/5PN1/PP1P2PP/RNBQKB1R w KQkq - 2 12"]
Here, Estrin played 12.Bxb5+ but my engine prefers 12.Qe2, and gives the main line as 12...O-O 13.fxg4 Bxg3+ 14.Kd1 with an advantage to White.
Is my engine missing something or does this line refute Berliner's play?