An uncomfortable truth first. At almost every level below master, you lose more rating to your own one-move blunders than you ever gain from clever plans. The brilliancy you found last Tuesday won you a game. The rook you hung on move 23 lost you three.
Fix the leaks before you buy new taps. It is the highest-return work in amateur chess and nobody wants to do it, because it is not glamorous and it does not involve learning an opening.
Why we blunder
Blunders are almost never a knowledge problem. You know a queen takes an undefended rook. It is an attention problem, and it comes in three flavours.
Autopilot is the big one: you play the move you decided on three moves ago without re-checking what has changed on the board since. Then there is tunnel vision, where you are so pleased with your own attack that it does not occur to you the opponent has one too. And time trouble, where the checking habit is the first thing to go, because checking is the part of your thinking that feels optional.
The blunder-check that takes ten seconds
Before every move, run the same short scan. Boring, identical, automatic. Checks, captures, threats, for both sides.
Does my move allow a check that hurts me, and is there a check of my own that wins something? What can my opponent capture now that they could not a moment ago, and what can I? Is anything of mine hanging, and is there a fork or a mate being set up behind the move they just played?
That is the whole routine. Most blunders die right there, before they reach the board. The hard part is not learning it, it is doing it on move 23 of a game you are enjoying.
Watch your own king
A large share of one-move disasters are mating-pattern blindness rather than piece-counting errors.
If you have castled and never touched the pawns in front of your king, you are one back-rank mate away from losing on the spot, and you will not see it coming because nothing looked wrong. Learn the feel of a thin shelter. Spending a quiet move on luft when the position is calm is one of the cheapest insurance policies in chess.
Train the patterns that keep biting you
You cannot concentrate your way out of blunders forever. Effort is a limited resource and it runs out around move 30.
What actually works is making the patterns automatic, so the danger registers before you have decided to look for it. Drill forks, pins and back-rank ideas until the shape sets off an alarm without conscious effort. Recognition is the real defence. Vigilance is what you fall back on when recognition has not been built yet.
Slow down at the right moments
You do not need to think hard on every move, and trying to is why people end up in time trouble, which is where the worst blunders live.
You need to think hard on the critical ones. Just after a capture, when the position opens, when your opponent does something you did not expect. Those three moments contain most of the blunders you will ever play. Spend your clock there and move quickly everywhere else.
Do this for a few weeks and the rating graph stops bleeding. It is not a dramatic improvement and there is no moment where it feels like progress. You just stop losing games you were winning, which turns out to be the same thing.