Write down the branches that failed. They are the only map you get of the dictionary's edges.
Most players keep track of wins. A dead-end log tracks the opposite, and it is a better teacher. Every time a branch collapses, write one line: the rung you were on, the change you tried, and what it taught you about the dictionary. Three fields, one line, no judgement.
The format matters less than the habit. A page of scribbles is fine. What you are building is a personal map of the places the four-letter dictionary refuses to go, and that map is more valuable than a list of successful routes because it stops you repeating the same failure.
Dead ends cluster into families. A whole corridor can be closed because the only exits are plurals, or names, or words the puzzle does not accept. Once you have hit the same family twice, you stop paying tuition for the same lesson. Three lines in a notebook can save a dozen wrong turns over a season.
There is a quieter benefit. A log makes failure feel like data instead of a verdict. You finish a bad round, write two lines, and the round becomes an entry. That change in framing is worth more than any single insight, because it keeps you playing on days when the score is discouraging.
You do not need a special board to do this. free practice gives you as many attempts as you want, and a failed branch there costs nothing but a line in the notebook. If a dead end keeps appearing, read the hints for that day after your own attempt; the nudge will point at the corridor you were missing, and the entry writes itself.
A log of failures sounds gloomy and is not. It is the cheapest way to acquire local knowledge, and local knowledge is what a four-letter dictionary rewards.