I solved thirty ladders in a row with the same small rule: after each one, three lines in a notebook before I looked at anything else. Line one was my plan at the start. Line two was the rung where the plan died. Line three was the word I did not know. Thirty entries later, three things had changed.
The first was obvious. My average move count dropped by roughly one, and almost all of the gain came from the opening rather than the endgame. I was taking the same number of moves once I was close to the target; I was arriving at the mid-ladder in a better position. That was not what I expected, because the endgame feels harder.
The second was less obvious. The words in line three started repeating. Five words accounted for more than half of my listed gaps, and three of them were words I had seen before and rejected as unusable. My vocabulary was not the problem; my confidence in perfectly good words was the problem.
The third was the interesting one. My plans got shorter to describe. At the beginning of the month a line-one plan took a whole sentence with two clauses. At the end it was usually four or five words: name a pivot, name a fallback. Plans that are short enough to check are plans that survive contact with a puzzle.
If you only copy one part of this, copy the three lines. The tracking is cheap and it turns a two-minute daily game into a small piece of research with a stable subject: you. The daily ladder resets every day, which is exactly what a study needs. Keep it up for a month of archived puzzles and the pattern will show up on its own.
One practical note about the study design. I kept the puzzle constant and changed only the notebook, which means the gains are probably a measurement effect as much as a skill effect. That is fine. A measurement effect that makes you better is indistinguishable from a skill effect, and it is much cheaper to acquire.