CRUCIVERB.COM

After I had two 54-word puzzles in the New York Times (one with 40 black squares on Oct 24, 1997 and one with 36 blacks on Dec 15, 2001), I didn’t seriously consider attempting a 52-worder. Merl Reagle had theorized for a long time that he thought a 52-worder was in fact possible, but would probably the limit for low word count in a 15x15. But, constantly reminding myself how difficult the 54s were to pull off, I never put my heart into trying to go lower. That’s not to say I hadn’t often toyed with 52-word grids. I had several different grid patterns that I was looking at as potentially fillable. But it wasn’t until I felt challenged that I really took to the task, and this finally happened in early 2005 when I heard that another constructor had sent Will several 54-word puzzles.

Now, when you’re talking about extremes in low word count, the rules for construction change a bit. In a “regular” puzzle, even a fairly wide-open themeless with a word count in the 60s, you build the puzzle based on the entries, often choosing them one by one for their instrinsic value and liveliness, stacking them, and usually placing the black squares around them as you proceed, as need dictates. Now, although this is all still true to some extent for extremely low word counts, when you get down to the mid-50s, the choices for grid patterns become a lot more limited. There are only a few basic patterns that allow for, say, a 52-word grid. So you need to choose the skeletal pattern first and add black -square chunks to bring the black count to the maximum allowable for an “official record-valid” 15x15, which is generally held to be 38 black squares. Thus the pattern is set in advance and you begin the arduous task of hoping to be able to fill it. In doing this, the #1 thing you need is patience, because you know you are going to be trying a zillion fill avenues that lead to dead ends. Also, with “regular” puzzles you have a lot more freedom to try to use really lively entries and high-point-value Scrabble letters (X, Q, J, Z, etc.). Although it would be nice to have this luxury with extreme low-count puzzles, unfortunately it just isn’t possible. For the most part you’re stuck with whatever finally works that makes the overall fill publishable. Still, I planned to get as many phrases in as possible, because I knew that Will would not accept a puzzle whose vocabulary was too bland, regardless of the word count.

I chose, as the skeleton, a grid with four 7x7 corners connected by six long entries:

Fig. 1

This has only 22 black squares, so I added chunks of black at the top and sides to bring the total to 38. I placed the blacks so that the four 15-letter entries going across would become two 13s and two 11s, and the two 15s going down would become 13s. This obviously is a lot more feasible to fill:

Fig. 2

Now, these huge wide-open corners are hard enough to fill in and of themselves, but the problem is compounded by the fact that they all extend into each other significantly. In particular, the 13-11 and 11-13 stacks in rows 6-7 and 9-10 had to work such that the two-letter endings of columns 3, 4, 5, 6, 10, 11, 12 13 were good endings for the seven-letter words in the upper half of the grid, and such that the two-letter beginnings in the same columns were good beginnings for the seven-letter words in the lower half of the grid. DINETTESETS looked good as a starting entry because of the ending letters it created (with the exception of the I):

Fig. 3

And on top of that, I wanted something that would both be a phrase and create good two-letter endings. REMOTE SENSORS seemed possible:

Fig. 4

Notice how all the endings to the seven-letter words in the top half have tons of possibilities (-ED, -ON, -TE, -NS, -OT, -RS). The only exception is -MI. Even the single-letter endings -R and -S are good. At this point I knew the upper left had to be done first because of the constraining -MI ending. Almost right away I put in STREET- for the beginning of the long down entry because I wanted to have many options for the bottom half. Lots of phrases start with “street” and –S, -T, -R and –E are all superb endings:

Fig. 5

It took quite a while to fill the upper left cleanly and in a fairly lively way. I have a word database that I can do letter-pattern searches on (using the GREP function, for those who care). This is equivalent to using the ever-popular electronic crossword-solving tool called a “Franklin,” except since I’m doing it on my own database, it doesn’t have a lot of the junk that Franklin has, and it has a lot more lively, interesting stuff. I was lucky to get the entries TOTASTE, SEAMIST and TAMIAMI in the fill:

Fig. 6

Now, on to the upper right. It looked good because of all those great enders. I noted that REMOTESENSORS could also be REMOTESERVERS if need be, and when filling I allowed for both possibilities. This area went through lots of trials. I had wanted something for the long down entry with lots of flexibility, but that didn’t work out. The fill I thought was cleanest constrained that entry to TRADEEDITIONS. But I felt those were common-enough letters to work with for the bottom half. I was pretty pleased with LAPCAT, TOPRATE and ADEPTAT:

Fig. 7

The bottom half was excruciating and took at least two weeks. Even though STREET- has lots of possibilities, the convergence of that with TRADEEDITIONS and the two remaining long stacked across entries was a real bear. The problem was finding entries to fill those slots that would result in feasible two-letter starting combos for the seven-letter entries going down. I always seemed to end up with one or two of these two-letter starters that yielded only a couple possibilities. Plus, the bottom right was already additionally constrained because of TRADEEDITIONS. Even though I hated to have an 11-letter entry that was a yawner, HERMETICISM above RAREINSTANCES seemed to yield decent starters (HA-, ER-, RE-, MI-, CA-, IN-, SC-, ME-). I also wasn’t thrilled that choosing this would force VED for the central three-letter entry, but Ved Mehta seemed fair enough after asking a few people whether or not they’d heard of him, and the fact that he got a lot of Internet verification (over 10,000 hits on the exact-name search I did). And obviously I was happy with STREETVENDORS:

Fig. 8

Filling these last two corners, with so many constraints, was painful, but in the end I was happy enough under the circumstances. NORANDA and EPHORS aren’t the greatest entries. But the former (Rouyn-Noranda) is actually a very important mining city in Quebec (one of the world's largest producers of nickel and zinc), and it got 260,000 Web hits. And I remembered “ephor” from middle-school history, so I didn’t think it could be all that horrible:

Fig. 9

One must bear in mind that this is an extreme, and for an extreme it turned out well. But ordinarily I would put more “zing” into a fill (there is no Q, Z, X, J, K, W, Y or B in this puzzle). Interestingly, after this had been published (January 21, 2005), at the most recent crossword tournament in Stamford, Merl did in fact show me a potential grid for a *50*-worder, saying it might be theoretically possible. That was a surprise considering how long he had assumed 52 was the low limit. Now, I love a good challenge as much as the next person, but sheesh! Well, maybe in another ten years!

- Frank Longo