In the opening and early middlegame Tarrasch outplayed his opponent and with simple 22...gxf4 he could have achieved material advantage with lasting positional dominance. But he was not completely satisfied with that as he had a notion that white can have some chances for counter-attack and so he tried to find anything better here. By this he lost a lot of time for thinking and the continuation with which he came out appeared to be no good. After several suboptimal moves of black Schallopp launched very powerful and resourceful attack culminated by sacrifice of Rook in move 34 which Tarrasch highly appreciated in his annotations. At that point white had the upper hand in the game and black was in deep troubles. For example, in post mortem analysis James Mason suggested a "subtle" defence 36...Rfg8 against threatening 37.Qxh6+ with intention to reply it by 37...Rh7 but Schallopp immediately refuted it by 37.Bf7! with win of white. [Maybe an improvement of this Mason's idea can be inserting of 36...Bc5 with intention 37.Nf2 (37.Qxh6+ leads probably to perpetual check only.) 37...Rfg8 as here 38.Bf7 is no option for 38...Rxf7 39.gxf7 Rxg1+ with check. But the position is very complex and it is quite easy to miss any tactical blow that refutes this attempt.] Unfortunately for him, Schallopp in great zeitnot erred with 39.Bxa6 while 39.Bd5!? and especially 39.Nc3! (recommended later by Steinitz as an easy win due to Knight's quick accession to the attack via e4 or d5) would have kept white's edge. 41.Rg5 was another mistake but it only accelerated the inevitable end.