I found this interesting because I keep an ongoing record of how long it takes me to solve the daily crossword puzzle (I use the free one in the Los Angeles Times).
Although the graph is entitled "Average word length for NYTimes Crossword answers, 1994-2017," it should more properly be described as "entry length," since quite a few of the answers on the late-in-the-week puzzles are multiple-word entries.
I found this graph at the Data Is Beautiful subreddit, where the discussion thread has some interesting observations about the software puzzlemakers use in their craft.
FWIW, my worst time on a Thursday puzzle this past year occurred in one this past December which required entering the following "words" -
sentildeor- which came from the following clues:
Terragona title, in detail?Fiendish.
Arles animal, in detail?
Toulouse trace, in detail?
Augsburg above, in detail?