Trouble with defending against pawn attacks

Hello all,

I played a game this morning where I played the opening and early middle game very well.  I was up a clear piece and engine said up as much at +8 if I'd seen a tactic involving bishop / his queen.  My issue is I can't defend.  All that I see beginning move 20 is that he's going to pawn storm me, and I only know that I need to trade pieces.  I systematically imploded in this game.  

Here's the position in the last move before I implode for good by capturing with the pawn.

I guess I don't know what the critical elements of their attack are and therefore I don't know what to precisely defend.  I would appreciate your thoughts and suggestions.  

https://www.chess.com/live/game/2896112748

Basic elements of an attack:
 - Open lines
 - Focus pieces on undermining or directly attacking a key line or square
 - Trade away important defenders

Basic elements of a defense
 - Keep lines closed, or if open already, interfere with key lines (even if pawn shelter is gone, you can move knights and bishops between attackers and your king for example)

 - Trade off key attackers

 - The most important element in defense is... stir up trouble for your opponent in a different area of the board (kingside, center, queenside). The usual advice is to meet a flank attack with action in the center.

This means while your opponent is preparing to open lines on the kingside, you should be trying to open some lines somewhere else. Direct defense is both technically and psychologically difficult, and in some cases it's objectively impossible to directly defend. However if you open lines and try to infiltrate in a different area, then you'll force at least some of the opponent's pieces to stay on defense, and even if the attack isn't stopped, it will make defense much easier.

13. 0-0
Objectively better than 0-0-0, but since you're a piece up, it was probably a good practical decision to avoid castling on opposite sides. You could have just 0-0-0 and probably won a relatively uneventful game.

I probably would have played 13.0-0-0
(The engine likes 13.0-0)

14.h3
A very bad reaction for at least 2 reasons.

First, if you believe your opponent is going to attack, then you shouldn't play on that side of the board. You should be moving pieces in the center or on the kingside.

Secondly, h3 makes it easier for black to open lines against your king. All it takes now is for black to play g4, and if your knight and pawn are still on f3 and h3, then for sure either the g or h file will open and this will make black's attack happen much faster. (You can even think of it like the pawn and knight are being forked)

I probably would have played 14.Nd4 or 14.b4
(The engine also likes those moves)

16.Qc3
I like it, pressuring a center pawn means you're stirring up some trouble in that sector.

17.Bf3
This feels a bit wimpy. I would have liked to continue with your idea and play 17.Nf3 or 17.c5

(The engine likes 17.c5, but not so much 17.Nf3)

19. e4
Very good. Keeping a line to your king closed (the b8 h2 diagonal). Also it freezes the e5 pawn in place, so it might be a juicy target.

20.Rad1
Feels random, the rook isn't conceivably threatening to infiltrate, on the d file, it's not pressuring a pawn, and it's not necessary for you to challenge black's d8 rook. Oh, I guess you wanted to capture on e5, and needed to defend the knight first. Hmm, ok.

I probably would have played 20.c5 or 20.Nf3

(The engine likes c5, and actually doesn't mind your move Rad1 either)

21.Rfe1
I guess you wanted to defend e4 but now Black has a simple follow up plan of f4, g4, g3. Why g3? Because it will fork your f2 pawn and h2 bishop, meaning guaranteed open lines against your king.

I would have played 21.f3 (to avoid the plan I just mentioned) or again 21.c5 (always do your best to counter whenever you feel like it's at all possible).

(The engine likes 21.c5 or 21.Qa5)

23.hxg
Opens the g file for black. I'd rather have played 23.h4
(The engine wants 23.Nb3)

24.Nf3
This cuts your queen off from defending along the 3rd rank. Black's key line is the g file, and he's ready to pile up on it, so Nf3 also seems slow. I'd rather play 24.Qh3. The engine doesn't like this and says black will trade queens... but that means this is a good practical decision because then the attack is over. And if 24.Qh6 Qg6, now you can play Nf3 if you really want, and your queen isn't cut off from defense.

(The engine likes 24.c5, and wants to meet Rg8 with Kf1, typical engine being cold blooded.)

25.g3
This looks terrible, but the engine points out you can answer fxg with 26.fxg3 even though that also looks bad, temporarily entombing the bishop. The point is in the game black could have attacked g3 again by playing 26...Bc5 pinning the pawn.

27.Nh2
Hmm. Usually this kind of move is made with the idea of hiding the king on h1. I'm not sure what your plan here was. I guess you were clearing the 3rd rank again for the queen. Notice this was basically wasting two tempi in your defense (you played Nd2-f3-h2)

27...Bc5
Black brings another attacker.
28.Qxe5
Misses the idea of Bc5. It was possible to keep defending along the 3rd rank with 28.Re3.

So to summarize, black opened the g file, piled up on it (Bc5 counts as piling up on it too). That was black's attack.

Your counter started off not bad, you were piling up on e5, but weren't consistent with it, making some slower defensive minded moves, and missing the chance to play c5. 

Then there were some bad moves that made it easier on black. Moves like 14.h3, 23.hxg, and 25.g3

Even though the engine says you can still survive after those moves (and the only alternative to 25.g3 was 25.Kf1), to me they stand out as showing some level of misunderstanding about the danger involved, and the usual ways of how to deal with this sort of pressure.