A question for the constructors out there...

[rhu]

Has anyone figured out what percentage of trigrams can actually be clued? I've been seeing more and more cases lately where I've been surprised by what I consider a contrived clue for a particular three-letter entry. I understand that sometimes a great theme requires a sub-optimal word in the fill, but it seems to me from my armchair that we're seeing more cases where an ordinary theme gets more than its share of contrived fill entries.

Which led me to wonder: What percentage of trigrams are cluable in some way? What if we break it down into trigrams containing at least one vowel and triconsonant clusters?

Clearly, at one end of the spectrum are entries such as CAT, nearby are common abbreviations such as PDQ and the ever-popular SST, obscure abbreviations such as LGA and DLL, further along are arbitrary roman numerals and consecutive-letter sequences such as DLV and FGH.... but if you absolutely had to clue ZFV could you do it?

Comments

[cazique.livejournal.com]

I'm reminded of the (perhaps apocryphal, perhaps mis-remembered by me) story of Henry Hook telling Eric Albert that he could clue any trigram in a crossword. Eric tested him with something resembling ZQF, and Henry replied "Cat, in some cryptograms."

(anyone know if this story is true, by the way?)

[530nm330hz.livejournal.com]

"Cat, in some cryptograms" might be acceptable to HH but it does hint at a class of clues that would fall into the "very contrived" category but that probably eliminate most of the otherwise uncluables:

"Caesar-shift cyphertext for cat" (or for any other legit three-letter word) would be solvable once you had one letter. Definitely a clue of last resort, but if I were a solver and encountered it I would grimace and accept it (if the rest of the grid were worth it).

[cazique.livejournal.com]

I figured Henry was joking.

[530nm330hz.livejournal.com]

With Henry, how can you tell? :-)

[jangler-npl.livejournal.com]

My wordlist has 941 three letter entries, but I haven't put Roman numerals, letter-sequences, or partials (ON A, AS I, etc.) in there. I'd estimate about 1,500 of the 17,576 possible 3-letter sequences are "legitimate," in the sense that a puzzle with an otherwise flawless construction would not suffer in quality from their inclusion. But I'm picky.

[mello-cat.livejournal.com]

Greetings from a random blog surfer who happened across this question and got intrigued (probably because I do construct crosswords). In my database* of approx. 6000 puzzles I see 2,448 unique 3-letter entries (excluding 20 that contain numerals or '&'). So that's about 14% of the (pure alphabetic) total possibilities that have appeared in puzzles I've done over the past five years or so. Of those 2,448 entries, 232 appear only once and 132 appear only twice, which may (but doesn't necessarily) imply they are pretty marginal. I'd guess a fairly large percentage of them are ones I'd balk at putting in a puzzle.

The number of 3-letter entries is growing, but pretty slowly. I've got 28 single-appearance 3-letter entries that were introduced so far in 2006.

*The database contains NY Times and LA Times since early 2001, NY Sun since its inception (April 2002), Wall Street Journal with random gaps since mid-2001 or so, CrosSynergy since August 2005, United Press Syndicate and USA Today since November 2005, and a smattering of other random puzzles I happened to run across in the past five years.