Closed jonadsimon closed 2 years ago
Determined a sharp correlation between mean word length & number of words. But his relies heavily on the distribution of words, could in principal have 1000 words with a small mean length
Can't use the naive mean length --> num_words cutoff since mean length is a function of the underlying distribution. Should see if there's some obvious long-word-truncation going
Input "flamboyant" is a clear counterexample: even when adding only 222 letters (< 15^2) among 38 words with a mean length of 5.84, the script is unable to do it.
Indicates that while length is informative, additional retry logic is still needed.
Have a better understanding of the situation now: word length & letter overlaps are both important
Using the analysis performed in https://docs.google.com/spreadsheets/d/1csESQNB7O_sZgsCGOGhZJpLpZTEvhHjILlNVjmFI_mQ choose the number of words to place on the board as a function of the mean word length + num letters