Minority Attack

Spanton (168) - Steve Dean (154)
Royal Beacon Seniors 50+,  Exmouth, Round 4
Veresov
1.d4 d5 2.Nc3 Nf6 3.Bg5 c6 4.e3 Bf5 5.Bd3 Bxd3 6.cxd3!?
6.Qxd3 is much more popular in ChessBase's 2019 Mega database, but Stockfish10 and Komodo10 narrowly prefer the text.
6...e6 7.Nf3 Be7 8.0-0 Nbd7 9.Qe2 a6!?
Black castled in the two games to reach this position in Mega19, but Komodo10, at least for a while, prefers the text.
10.Rfc1 0-0 11.a3 Ne8 12.Bf4 Bd6 13.Bg3 Qe7 14.b4 Bxg3 15.hxg3 e5 16.dxe5 Nxe5 17.Nxe5 Qxe5 18.d4 Qe7 19.Rcb1 Nd6 20.a4 b5
One of the standard responses to a Minority Attack when White does not have a light-square bishop. The idea is Black accepts a weakness on c6, but masks it by dropping a knight into c4.
21.Qe1 Nc4 22.Ne2 h5!?
Black seeks counterplay by weakening the white king's defences.
23.Nf4 h4 24.gxh4 Qxh4 25.Nd3?
It was probably time to try to bail out, for example with 25.axb5. The engines, especially Stockfish10, strongly dislike moving the knight from the kingside.
25...g6?!
A better way to double on the kingside was almost certainly 25...Rad8, intending ...Rd6 etc, not least because this also gives the option of ...Rg6 or even …Rf6 in some circumstances.
26.g3 Qh3 27.Nf4
Realising the error of my 25th move.
27...Qf5 28.axb5 axb5 29.Qe2 Kg7 30.Qd3

In this position Black came up with an unusual way to proceed

*****
*****
*****
*****
*****
30...Ra4!?
An imaginative way to switch the attack to the queenside, but the engines prefer 30...Qf6 or 30...Qg5.
31.Qxf5 gxf5 32.Nd3 Rfa8 33.Rxa4 bxa4 34.Ra1 a3 35.Ra2 Kf6 36.Nc5 Ke7
Not 36...Rb8?? 37.Nd7+.
37.Kf1 Rb8 38.Nd3 Rh8 39.Ke2 Rh1 40.Ne1 Rh8
The engines' 40...Nxe3!? 41.Kxe3 Rxe1+ 42.Kd3 Rb1 43.Rxa3 Rxb4 44.Ra7+ Kf6 45.Rc7 is also approximately equal.
41.Nc2 Ra8 42.Kd3 Nb2+ 43.Kc3?!
Trying for more than the position justifies.
43...Nd1+ 44.Kb3 Nxf2 45.Nxa3 Ne4 46.Rc2 Kd7 47.b5 cxb5 48.Nxb5 Rg8 49.Rc7+ Ke6 50.Rc6+ Kd7 51.Rc7+ ½–½