You sunk my
March 31, 2018 2:51 PM   Subscribe

 
Boy howdy I lost a couple hours to this the other week. It's a very nice little thing, satisfying in the same scratching-the-procedural-itch way as things like Picross. My biggest complaint about it is just that it feels like it's got a higher proportion of busy-work clicking to actual solving compared to my favorite puzzles, but that works really well in its favor when you need something more on the soothing side of zone-out puzzling vs. an eyebrow knitter.
posted by cortex at 3:03 PM on March 31, 2018 [2 favorites]


but where is the win animation?
posted by parmanparman at 3:33 PM on March 31, 2018 [2 favorites]


If it helps you can just manually throw a deck of cards across the room to celebrate.
posted by cortex at 3:37 PM on March 31, 2018 [32 favorites]


These are great. GAMES magazine has had these in every issue for many many years ... I’ve always liked them better than sudoku/nanpure.
posted by freecellwizard at 3:46 PM on March 31, 2018 [3 favorites]


This is good.

If you like these, you may (will definitely) like nonograms. Which you can play here and here
posted by Horkus at 4:05 PM on March 31, 2018 [3 favorites]


This gives me the same thrill as crossing off boxes in logic puzzles.
posted by Ruki at 4:12 PM on March 31, 2018 [3 favorites]


I sunk my battleship.
posted by ckape at 4:40 PM on March 31, 2018 [3 favorites]


Aw, that's fun. Thanks.
posted by seanmpuckett at 4:50 PM on March 31, 2018


This combines the familiar gentle ease of kakuro puzzles with the childlike wanton destructiveness of Battleship. Perfecto.
posted by duffell at 5:29 PM on March 31, 2018


Excellent; however, puzzles may have multiple solves and the rules will count you as incorrect even with a correct solution. The worst solve I have had 4 permutations... and that's a problem...
posted by Nanukthedog at 6:02 PM on March 31, 2018 [1 favorite]


Nanukthedog, I would be super interested to see a screenshot of such a specimen; I had thought I'd found the same a few weeks back once or twice before twigging to a relatively subtle bit of ruleset that I hadn't appreciated before. But I'm not convinced that ambiguity isn't still possible. I definitely saw in practice the problem of ambiguous solutions in randomly generated nonogram puzzles a few years back when I was working on a puzzle generator/solver program, and I wouldn't be surprised if (a) this puzzle type is likewise susceptible to those ambiguity traps and (b) this generator doesn't successfully avoid such cases entirely.
posted by cortex at 6:06 PM on March 31, 2018 [1 favorite]


Yeah, I thought my first one was ambiguous and was a little ,"grr", about it. But it turned out I just hadn't scrolled down far enough to see the rule about no touching diagonally.
posted by Horkus at 6:20 PM on March 31, 2018 [1 favorite]


This is a reasonable interface, which is nice. Logic puzzles with annoying interfaces are the wrong kind of frustrating. Case in point: numberlink at nikoli.com. I generally like numberlinks, and I generally like nikoli's puzzles, but I really really don't like their numberlink interface.

Good points to this battleship interface: click on the number for a completed row and it will get marked as completed all at once.
Automatically greys out ships you've completed.
Greys out numbers for rows you've completed.

Boos: won't let you check a solution just from entering the ships, you have to make all the other squares blue by clicking something (a global button for "I got all the ships" would be nice).
posted by nat at 6:35 PM on March 31, 2018 [4 favorites]


Oh wow. Had no idea you could click the row/column numbers, thanks very much.

My favorite logic puzzle of this sort is Nurikabe, which has pretty elegant rules and a really satisfying puzzle progression. It's another Nikoli one, and there's several apps -- the Conceptis app is good.
posted by rifflesby at 7:14 PM on March 31, 2018


Needs more Rhianna...
posted by Naberius at 7:49 PM on March 31, 2018 [1 favorite]


This was not obvious to me at all: double-click to guess where boats are.
posted by the Real Dan at 8:22 PM on March 31, 2018


BTW, if you click on a column/row header where your displayed ship spaces match the count, it'll autofill the rest with water. Click a zero, the whole row clears, etc. Also, the seeded ship spaces that display are accurate to the 'segment' of boat showing, so if you see one end of a boat that's not a complete circle, you know there'll be another segment directly adjacent. Add in the fact that boats can't touch and that's a host more spaces that are cleared right out the gate... gets to the point that the max size field can be crushed in about a minute, zero inaccuracies.
posted by FatherDagon at 11:52 PM on March 31, 2018 [1 favorite]


You can also click (or tap; it works on my phone too) and drag to fill in multiple squares at once.

The best hint (that the author doesn't mention) when doing the 15x15 puzzles: the shape of the ship pieces given at start are correct -- so where there's an end piece you know there must be at least one more piece you can add on to the "boat" side. If there's end pieces facing each other, you can often fill in a boat for free.
posted by seanmpuckett at 5:35 AM on April 1, 2018 [1 favorite]


Thanks for all the UI hints: that made life a lot easier, especially the dragging.

I did find my first apparently ambiguous grid after some, err, extensive research, and it was a 4*4 square in the top left of a 10*10 puzzle where I could place a couple of 1s, and a 2 interchangeably*. But I didn't save it for puzzling out later, so I could be wrong. The way I did it worked fine, though.

*This is an incomplete description, because clearly that can only be placed one way.
posted by ambrosen at 9:36 AM on April 1, 2018


Awesome, thanks! An AskMe recently turned me on to Simon Tatham's Portable Puzzle Collection (also available on iOS), and this tickles the same GAMES magazine/Web 1.0 pleasure centers.
posted by ejs at 12:42 PM on April 1, 2018 [1 favorite]


The source code in unminified JS is easily viewed. [ spoilers ] The ships are placed randomly but legally. To determine which positions to reveal as hints, a built-in solver is called repetitively until there's enough visible information on the board for it to be solved without guessing. So, every puzzle is logically solvable, which both satisfies and disappoints me. I mean yes it's nice that they're all solvable, but the machine has already done it so why... Well, it's good news though, because now I can have my free time back.
posted by seanmpuckett at 2:31 PM on April 1, 2018


If you use the Stylish browser plugin and would like a visual aid or cheat (depending on how you feel about it), add the following CSS rule:
.battleships table.puzzle th.clickable {
	background-color: rgb(179, 212, 224);
}
posted by ardgedee at 2:48 AM on April 2, 2018


This was very fun at first but took little time before it became too easy, including the largest board. And I'm not especially bright.
posted by waving at 6:40 AM on April 2, 2018


I think this implementation errs far on the side of straightforward solvability for my tastes, yeah. Whether that's because the puzzle generator was designed for relatively easy puzzles or because designing more complex ones just requires more scheming, I don't know. It works fine as what it's meant to be, an introduction and mild past time, but there's rarely any crunch to the puzzles.

Which I feel is the default tendency with random puzzle generators of all sorts, from past experience: it's much easier to work up a ruleset for generating a logically valid puzzle (which is of course your top priority when building such a machine) than it is to work up a ruleset for generating interesting puzzles (which is a secondary and more subjective goal).

Which I think is a genuinely interesting topic in its own right: how do you characterize "interesting" in a puzzle, how do you bake in difficulty automatically in a way that creates a sense of flow and a rewarding pace of "a ha!" moments for human solver rather than just creating e.g. exceptionally annoying tile-counting tasks to slow them down.

I was starting to flirt with that a little on the design side myself a few years back, working on nonogram generation and solving routines, and it's something I really ought to get back to at some point because I was pretty pleased with the progress I'd made breaking my own solving process down into a couple dozen discrete subtasks which my solver would work through in rotation, while recognizing some subtasks were mentally easier to execute than others. Working through that, and thinking about which nonogram/picross games I liked vs. didn't, the main thing that occurred to me is that the best puzzle sets (a) did a good job of characterizing the rising difficulty of the subsets of puzzles they were offering and (b) ratcheted up that difficulty mostly by incrementally introducing a larger proportion of the high-effort subtasks while making sure that they existed in a mix with simpler, quicker "a ha" moments to let me as a player celebrate solving a tough bit of logic by having a couple quick bites of simpler stuff.

I never got as far as implementing a logically sound generator—I spent basically all my available programming energy at the time on breaking down the solving process into me-like hueristics, and got burnt out at that point—and so I don't know where I would start with trying to write a generator to create interesting puzzles from scratch. But I'm fairly certain I could use that mix of heuristics I came up with to create an adjudicator to rate an existing puzzle for rough interestingness and difficulty, based on the variety and effort of the subtasks required to solve it. At which point a mediocre generator could suffice, if it's fast: have it chuck out a thousand random puzzles, let the adjudicator rate them all on difficulty and variety, and then just keep the best one or best handful. Repeat as much as needed.

I doubt that scheme would yield puzzles that are as good as human-generated ones, but it might be a path to creating satisfyingly crunchy ones. (With nonograms of course there's also the tradition that the solved puzzles will look like something, which is another challenge entirely that I haven't even considered in my work yet.)

But all of that is to say: I think it requires either a competent human designer or a more elaborately designed algorithmic generator (or maybe a thoughtful mix of both: generator to come up with candidates, human agent to adjudicate and flag the best ones for public consumption) to get consistently good, crunchy puzzles for something like this.

All of that to say, coming back to this got me to check and see if Conceptis had a version and, yep, "Battleships" is on iOS at least and plays just fine (pretty similar to this in UI and interface, in fact). And the puzzles are immediately meatier, in a way that has made clear just how chill these puzzles are. I'm finding myself having to slow down and really think through small, "easy" puzzles because every one is pushing the logical pieces in front of each other in a way that feels designed in just the way that these ones don't particularly.

Conceptis generally does a good job with their puzzle clients, I've found; it varies a bit (I don't like their nonogram implementation very much, but I'm fussy about those and almost no one gets 'em right, but their nurikabe app is fine) but while they're never brilliant they're also never awful, and they all ship with enough free starter content to keep you plenty busy for a while and decide if you like the puzzle and their app implementation of it enough to toss a couple bucks at additional puzzle packs. At this point if you're digging the battleships puzzle but want more crunch, I'd totally recommend grabbing their version on your phone.
posted by cortex at 7:21 AM on April 2, 2018 [3 favorites]


Cortex, thanks for the feedback and suggestion. I'm going there now.
posted by waving at 7:44 AM on April 2, 2018


On a revisit, the 12x12 puzzles seem to be more fun than the 15x15s, especially if you click "new puzzle" a few times to get one with fewer hints. Those seem to be more rewarding in terms of logic puzzle "ok can you figure this out, sassyboi" rather than just ticking off rote click-drag things.

If I had more motivation I'd do a bunch of analytics based on generating millions of puzzles and sorting the solutions by minimum number of required hints. I mean can you imagine the rush of solving a 15x15 based on maybe 4-6 hints (rather than 16) that you /knew/ was solvable logically? I don't even know what the minimum is.
posted by seanmpuckett at 11:40 AM on April 2, 2018


Minimum viable hint count appears to depend on which hints are provided and what process you use for solving it.
posted by ardgedee at 12:34 PM on April 2, 2018


I dunno, I like that you can just logic your way through the puzzles fairly algorithmically. It's soothing. You get a board with just enough information to solve the puzzle, and if you just walk through it step by step, you can always solve it with zero mistakes, no backtracking, and no guessing. It's definitely easy and I don't feel a desire to sink lots of time into it, but it's a pleasant little game that reliably yields a series of minor victories. I think it's nice.
posted by Anticipation Of A New Lover's Arrival, The at 5:23 PM on April 2, 2018


« Older The 2018 Hugo Finalists   |   “I give to you the sound of corn flakes” Newer »


This thread has been archived and is closed to new comments