them's crossing words

Constructor notes: Adventure Awaits

7/2/2024

Adventure Awaits is a midi (11x11) D&D-themed crossword that I just finished putting together. Before I start diving into the details, go ahead and give it a solve. It'll help to have some context as I elaborate on the components of its construction. Go ahead, I'll wait.


(this area intentionally left blank)


Okay, crossword solved. Let's talk puzzles.

The idea for this crossword started with one of the spanners: BALDURSGATE. I had recently finished Baldur's Gate 3 (incredible, BTW) and felt that it became a big enough name in the gaming community that it was worthy of a puzzle. That said, I wasn't really sure where to turn for the other themers. The most obvious approach is other D&D cities, so I started thinking of some names that could fit the bill. Waterdeep is a big one, a classic setting for many campaigns. Sword Coast is the general region, maybe not quite what I'm looking for. Neverwinter is less known, but then again there's a game named after it called Neverwinter Nights. That's it: games featuring D&D cities! And so the idea was born.

Now, most acceptable crossword themes have a secondary element that adds a little bonus intrigue to the solve. Unfortunately I couldn't think of anything to tie the two cities together. Instead, I opted to sprinkle in lots of D&D references into the clues and work extra hard on the fill to compensate. Maybe not as clean as some other puzzles, but I think it works.

But before diving into the fill, the next component of a crossword is placing the black squares. I started by inserting the themers equidistant in the grid, picking a grid size that lets them span 11-squares across. I then introduced black squares with the goal of segmenting the grid into corners and cutting down on the overall length of words. The goal at this stage is solvability, reaching a point where my crossword software can fill the grid with words that offer lots of possibilities.

There's no real rhyme or reason to the introduction of black squares other than a few key guidelines:

Pinwheel shapes are really common in crosswords thanks to the convenience of segmenting a grid into discrete corners, so I try to deviate from that pattern as much as possible. I also like to fit as many long words as I can, particularly in midis because they often lead to interesting words. That technique really worked out for me because it eventually led to TRAVELPLANS and MARTIALARTS, two spanners that do a great job supporting my original themers. I stumbled upon these spanners by complete accident.

With solvability achieved (and by that, I mean a healthy amount of possible fill that my crossword program can quickly provide), the next step is to start experimenting with fill. This step always takes the longest for me, I often find myself going down word selection paths that end on some bad crosswordese or otherwise impossible words, causing me to delete the initial seeds and start fresh. There's also some wiggle room here to add more black squares if the fill is being difficult. For example, I added two more while filling because I couldn't otherwise arrive at words that I actually liked in the corners.

With fill, I start by reading through the suggested list from my program and select a word that's within theme. Something like ARROW, where I see relevancy towards D&D (arrows being a core component of many D&D classes) and tolerable conditions for a solver (an arrow can be clued in many different ways and is very common in regular speech). From there I experiment with different word sequences, selecting my favorite for the next fill and repeating for the rest of the corner. As is often the case, the first few selections end up with really bad results, like EMU, EDDA, or the other dregs of your wordlist. When this happens, it's best to go back a few words, try some different combinations, and continue to experiment until the entire corner has a consistent quality.

Once you repeat this process a few times you start to notice contentious spots in the grid, areas where a word will often be awful due to the position in the corner and the intersection with your themer. A good approach is to start the corner with that word. Try to maximize the quality of your hardest word, and often you'll arrive at nicer fill at the easier edges.

In the case of Adventure Awaits, I really struggled with the upper-left corner. Nothing was feeling quite right, I was constantly arriving at bad fill no matter the technique. What finally worked for me was actually swapping my two themers. Originally I had BALDURSGATE first, NEVERWINTER second, because I thought ending on a W would make for easier fill. After swapping their positions, I had no problem finding something I liked for the upper-left corner. Sometimes you have to drastically change the plan to compensate for the reality of your wordlist.

I'm not sure whether it was due to my themers or my black squares, but every corner in this crossword was pretty difficult to fill. It's not normally this bad; I had to revisit every corner probably 10 times before I found something that I could qualify as acceptable. Even then I had to make some concessions, EDDA, SGT, EMTS, and AGESA being some egregious examples. That's just how the cookie crumbles.

Let's talk about some of the happy accidents that occurred during the fill. My favorite surpises were TRAVELPLANS and MARTIALARTS, two full-length spanners that appeared spontaneously from my wordlist. They both fit the theme beautifully while maintaining a solid final grid score. I was really fortunate to just stumble into these and their positions, I wish I could say that it was premeditated but it was all luck. Once I saw them in the wordlist as potential fill, I doubled down on them, making sure that I worked the corners to fit.

EMUS and EMOS kept popping up in the upper-right corner and I was not about to concede to either of those words. It took so much tweaking to arrive on something else, and that something is EMTS which is honestly not that much better. But hey, at least it's not EMUS.

With the fill completed, the final step of construction is cluing. There's no secret formula for this step. By the time you're filling the grid, you probably already have a rough idea for potential clues. Maybe you have a few words that are reserved for wordplay (e.g. ARROW), or some that are simply proper nouns (e.g. EDDA). Either way, it's worth looking up some of the trickier words to make sure that you're cluing them appropriately, that is, matching tense or case. Other than that, the process is going from word-to-word and thinking of something clever, cute, funny, or easy.

Speaking of easy, one thing I try to do in my first cluing pass is to seed easy clues. The goal is to maintain a balanced difficulty, I want at least one or two gimmes in every corner (at least for an average difficulty crossword, like this one). If the corner is particularly esoteric (looking at you, EVAL vs. AGESA) I'll lean easier, avoiding wordplay and cleverness. An initial easy-clue pass is a great starting point for a grid because it lets your imagination run wild for the rest.

It's hard to know whether or not a crossword is solvable until you have someone unfamiliar with the grid playtest it. Until then you're really just guessing, who knows if "Earth, according to sci-fi" is as obvious as you think it is when initially writing it. Everything that you clue is done with the prerequisite knowledge of the answer, there's no way to breach the gap in solvability that accompanies a blank canvas. So grab a playtester or two and watch them run through the grid.

In my case, playtesting revealed some serious painpoints around EVAL and AGESA, which I had originally clued as two programming references. That's just too much programming for one crossing. I laxed EVAL by rewording it as "Assessment, for short" from "The E in REPL", making it more generally approachable. I also noticed that NEVERWINTER is much less well known than BALDURSGATE, so I tried to make the upper-left corner easier to compensate, making PAIN an absolute gimme so that the wordplay in PATHS and ARROW is easier to guess.

And that's about it! Thanks for following me in this construction journey. If you're wondering what the title is referencing, check out this hilarious Joel Haver video: Playing an RPG for the first time. Till next time.

In defense of trivia in crosswords

6/3/2024

Trivia has long been a bugbear of mine. A grid full of missed references is like a desert full of empty oases. You scour about desperately from pool to pool hoping you'll find a sip of water but instead you're hit with a clue like "The only bird with calf muscles."

Recently I've been thinking more about the subject of trivia in crossword puzzles. Why does so much of it exist in good crosswords, and why does it bother me so much?[1] What I've come away with is a newfound respect for trivia and its rightful place in the crossword grid. It doesn't serve to belabor a solve with esoteric references, but to emphasize the difficulty of the grid and the puzzle mechanisms at the crosses.

Ideally the words in a crossword grid are recognizable, guessable, or interesting to a solver. That’s the baseline. But if the every word in the grid falls into one of these three categories, what's left for a constructor trying to introduce difficulty? That is mostly the responsibility of the clues, which may range from rote, “Opposite of NNE” (SSW), to esoteric “The only bird with calf muscles” (EMU). Clues must define their respective words and they're given a huge amount of leeway to do so, be it through trivia or other means.

The problem is that trivia exists on a binary spectrum: you either know it or you don’t. That’s why it’s so grating to see a piece of esoteric trivia blocking a solve. How are you expected to know the answer?

The answer is you're not. At least, not directly. Crosswords employ obscure trivia as a means of instilling difficulty. The constructor knows (probably) that the general public does not have some innate knowledge of bird calves, but trusts that their grid is tight enough for solvers to piece it together from contextual clues.

The best you can do with trivia clues is to appreciate the small details that serve as the real hints. "Ethiopian gem" (OPAL) makes it apparent that you're searching for a kind of fancy rock; "The only bird with calf muscles" implies that you're looking for a pretty big bird (although you'd be forgiven for not knowing what an EMU is in any context).

On top of hinting at the parameters for the answer, trivia serves up some post-solve flavor. You found the solution, nice work. Here's a little factoid for your future self. Puzzles would be dry if these types of details were totally omitted.

Maybe I'll lighten up the next time I'm stumped in a grid full of Taylor Swift references, thinking of the greater puzzle elements at play in the broad game of crossword calculus. Or maybe the constructors could just choose some better trivia.


  1. Good crossword is subjective, of course, but generally my definition is a puzzle that is constructed for the purposes of pleasant solving and instilling "Aha!" moments. ↩︎

Puzzmo games, ranked

5/27/2024

Puzzmo has a whole bunch of games. Which ones are worth playing?

All of them, of course, but I have favorites. Here's my tier list of games, with a few extra Zach Gage games thrown in for good measure.

S-tier

Cross|word: look, this is a crossword blog. I'm not sure what else you could've expected here.

A-tier

Flipart: admittedly, this game doesn't have a whole lot of depth (pun intended). But as a palate cleanser between games or as a quick diversion while waiting in line, Flipart is unrivaled.

Knotwords[1] and Good Sudoku[2]: see the footnotes below, but these games together get more playtime than any other Puzzmo game, save for the Crossword.

B-tier

Typeshift: this game rules for its tactile feel alone. I'm all about scrolling those columns up and down, even though I am not the biggest fan of the "make a word from scrambled letters" style of puzzle.

Wordbind: very similar to Typeshift, but without the scrolling goodness. I think the best way to describe it is as a horizontal word search.

Pile-up Poker: the puzzle in Pile-up Poker is a fun one and offers some randomness that isn't present in other Puzzmo games. But five sets of four deals is just too much poker. This is one of the few times that I think a Puzzmo game overstays its welcome.

C-tier

Cube Clear: I'm a little bit of a SpellTower hater and Cube Clear is by all looks just a smaller version of SpellTower. However, the smaller size and rule changes actually help solve my analysis paralysis with SpellTower. I think they're great changes, although I don't play it often.

Really Bad Chess: when I was playing more chess in general, I was playing more RBC. It's not a problem with the game, just my attention span! RBC is a simple chess variant, but it's effective.

D-tier

SpellTower: my biggest gripe with SpellTower is the analysis paralysis of looking at the huge grid and deciding on a path towards the solution. The game is kinda built against quick experimentation, as evidenced by the high score walk-through where a high-skill player copies the grid into Photoshop to map out a complete solution before entering it into the game. Too much forethought for my taste.


  1. Not on Puzzmo, but should be! A fantastic word game, if not hindered a tiny bit by random puzzle generation. ↩︎

  2. Not on Puzzmo, but an obvious newspaper classic. The guys over at Cracking the Cryptic have demonstrated to me that sudoku can be just as deep and varied as crosswords when a great constructor is involved. ↩︎