Analyzing missed h4/h5 pawn pushes in own OTB play.

Methodology:

Here is a categorization of the circumstances where h4/h5 thrusts can be played:

1. Fixing a weakness in an endgame (or avoiding a weakness being fixed by opponent)
2. Kingside attack with opposite side castling
3. Kingside attack with same side castling or uncastled king
4. Disrupting opponent's pawn storm

I could not really find any examples for either 1 or 2. This is perhaps unsurprising, as 3 and 4 are usually associated with much higher risk (details in their respective chapters).

Part 1: Pawn storm disruption

This is a typical idea in various Sicilian lines, hence I should be well-acquainted with it. Unfortunately, I usually only go for it when it's part of my opening preparation (i.e. when I know it works).
Usually the justification I give myself to not play it is that it's easy to make a mistake when playing on the same side of the board as the opponent. This is a good general rule, I think the point in the following cases is that I was not, in fact, the weaker side on the kingside, thus h4/h5 was justified.

Part 2: Kingside attack with same side castling or uncastled king

Similarly to part 1, my reluctance to play h4/h5 here stems from the fact that they are commital moves and I prefer to attack without weakening my king. Obviously this is a limitation that I would like to get rid of.

Summary and conclusions

In total, I didn't play h4/h5 as a disruption to enemy pawn storm 3 times, and 5 times as a part of kingside attack with my king uncastled or short castled.

Zero misses for endgame and opposite castling circumstances suggests it's mostly about king safety for me - I have been reluctant to move h pawn when I felt I needed it to protect my king.

In terms of reasons, there were 5 cases where I didn't see the move at all and 4 cases where I saw the possibility but couldn't convince myself to follow through.

I think in the following tournaments I will try to go for the h4/h5 even when I will be scared just to get a feel for when it works and when it doesn't. Hopefully this will allow me to get rid of the bias.