Ask Dr. Sudoku #17: Thoughts on digital solving options and puzz.link

0 Flares Filament.io 0 Flares ×

[Update: As of 2021, we are now routinely using penpa-edit and more info is here.]

While I am mainly a pencil-and-paper puzzle solver, I always thought GMPuzzles would eventually find some digital outlets. Not necessarily one outlet — our different styles have different needs and a good app for Sudoku/TomTom is probably quite different from a good app for Tapa/Nurikabe — but at least some outlets where we would be content providers. While I will soon have some of my TomTom puzzles as part of one app-based release, this is the exception and not the rule after 7.5 years.

A few times over the years, including again two weeks ago, solvers asked us to consider linking out to puzz.link, an open source project that gives constructors an option to share online versions of several puzzle genres. We did try this last week for Star Battle, one of the easier styles to think of digital solutions for and where the puzz.link implementation is really good including the ability to add star notes on lines and corners of cells. We received positive comments from several solvers, which suggests at a minimum that this year we have to get serious about digital options.

While I’ve been pointed at this project before, I hadn’t properly explored it until this trial week. I’m still mainly a pencil solver but I didn’t mind the puzz.link interface for object placement/shading puzzles in my experimentation. I could see a role for maybe half of the puzzles we put out with a tool like this. I worry this coverage is too low, and misses important genres for us like Sudoku (+ variants) and TomTom that have better implementations elsewhere.

I also have my own quirky dislikes of the puzzle presentation. The following concerns may not be too important to solvers, but they are important to me as a puzzle publisher. We care about our authors and about puzzle themes. The default puzz.link implementation allows sharing a grid with other users as a web link, but it gives no apparent option to connect a name or a title to the grid when building a link. While the puzz.link database picked up our posts (listed as from “gmpuzzles” with a link back to the web post), the actual solving experience is completely removed from recognizing the author/origin of the puzzle. Related to the first item is that the current implementation is to click out from this site to go elsewhere, to a page with a different visual style. With some work (contributing directly to the puzz.link project, or funding development to extend it) we may be able to embed our own version of the code here on this site.

My vision for digital solving of our puzzles is larger than this: at minimum an interface through which we can reliably and securely sell puzzle packets, that works well on computer and mobile device screen sizes for solving without any printing required. But at maximum with many benefits that aren’t possible on paper. A few are in tools like puzz.link or what Nikoli.com had like showing the solvers what rules may still be incomplete/broken. We get so many solvers asking us about duplicate solutions where we always just respond by highlighting the rule they missed. So that kind of visual highlighting, and interactive tutorials, would be great to expand the puzzle community. Another mode I’ve not seen routinely could be a “walkthrough” mode for a puzzle like a TomTom where we can code 3-4 natural next steps that should be found by a solver so when they ask for a “hint”, they can get a dynamic one for the current state of their grid. Leaderboards of times, watching the solving paths of others, and other competition friendly but community growing aspects would be great.

But maybe I’ve let my big vision halt us from starting somewhere with a basic web interface for our puzzles, understanding we cannot cover all. Puzz.link may be the closest of the things we have seen to be able to start using, and it has grown up naturally over the last few years to be a very good resource. But I have to evaluate how much work is needed (and who could do it) to extend the code to get to where I am comfortable with it for many of our genres at GMPuzzles. We’re going to give this more thought, and also recheck one or two other options that might complete a patchwork solution to get 80+% of our puzzle genres covered.

While we are building out a plan for eventual coverage this year, we are interested in hearing your comments on how you solve our puzzles and what makes a good/great digital option in your mind.

  • Michael says:

    While I don’t have a problem with solving puzzles on paper, I much prefer to use a computer for it, and in cases where the puzzle is already in a digital form, it would take exceptionally good reasons to move to paper. Admittedly, I don’t have a printer, so I don’t really have much of a choice anyway – but even if I did, I would not be using it unless I *had* to. It is much easier to fix any mistakes I might make (especially if I don’t see the mistake straight away – I can undo my way back to the point I messed up), and I don’t have to worry about being able to read my handwriting. 🙂

    My preference for solving digitally is strong enough that I’ve written two tools of my own specifically for this purpose:

    First, I wrote a program specifically for Sudoku which doesn’t do anything fancy in terms of solving and doesn’t know about variants beyond simple grid size, but it does automatically handle basic counting, while still letting me override possibilities in specific cells. I like to use this because simply counting which numbers have been used is trivial – the challenge is spotting the relevant patterns, and the program doesn’t do any of that (although it can fill in singles if I really want it to).

    For other puzzle types, or Sudoku with non-standard regions, I wrote a second program which is very flexible and can handle most puzzle types – it can do different grid types, different cell styling, different controls for each puzzle type, etc. This is what I use for pretty much every non-Sudoku puzzle, extending it as necessary. There’s very little in the way of assistance here of course (although I did code in automatic inside/outside coloring for Slitherlink), but it does the job.

    In the rare case where I have a puzzle that doesn’t really fit with either of those programs (e.g. a grid type I don’t support and don’t want to write support for at the moment), I *still* keep it on the computer by screenshotting the puzzle and solving it in Paint.net.

    My tools are not perfect, but they’re *good enough*: they let me do what I need to do without too much hassle (and they were fun to work on). It’s the same thing with Cracking the Cryptic: I’ve been doing their puzzles for the past month or two, and their online Sudoku app is good enough for the puzzles they tend to do – e.g. their check function doesn’t know how to check more than basic constraints – but it’s *good enough* for most of the puzzles they do (even non-Sudoku puzzles occasionally!), and if they really do need something more than their tool offers, they use something else (like the site you mentioned, or Penpa).

    The grand vision sound cool, and I’d love to see it happen, but I think it’s better to have at least *a* tool than no tool. Whether that should be a custom one or a pre-existing one, I don’t know.

    If I had to mention the one feature I’m missing the most at the moment, it’s really just keeping track of whether I’ve solved a given puzzle or not – we had that before the re-launch with the answer string, but I’d be completely fine with just basing it on the honor system and not verifying it. If I’m not confident in my answer, I can always come back when the solutions are posted.

  • SS says:

    Just to add my data point – I am a paper and pencil solver. I like how much enjoyment I can get from a simple sheet of paper.

    Online solving such as the old Nikoli was sometimes nice for the “Try” option – but that probably led to some bad brute-force attempts instead of trying to solve more logically.

    To echo the comment above, I do miss the answer checker and tracker.

    • drsudoku says:

      What do you miss specifically about the tracker? Having a column marked solve with a check mark as you scan through the site? This used to be a feature on the tracker, even without an answer entry requirement, but the applet has changed code recently and I can’t get the “SOLVE” option to be up on its own. I can ask the developer about that if that is the most useful piece to just mark for yourself when something is done.

      There are a lot of tracker features we wanted to have that aren’t there, like an all up view of the GMPuzzles site or views by puzzle type. It seems to be the bare minimum of tracking right now and is very difficult for us to develop/maintain so we are considering taking it down (for an eventual replacement).

  • SS says:

    Yes – the main feature I miss is the indication to myself that I have solved the puzzle. With the absence of “Solve”, I have been tempted to click “Fave” – but I don’t want to skew the voting! (Though if I Fave everything, I suppose I wouldn’t!)

    It would be nice to have a more full-featured and less finicky answer checker/tracker. It would also be nice to not lose the existing tracker for past solving history, if at all possible. I am still slowly working my way through the old puzzles. 🙂

    • drsudoku says:

      OK — I have an open request to learn how to get a SOLVE option back up when an answer is not input (we had been able to do it before but I have not been able to replicate since the latest version went up). Hopefully that gets basic functionality back for the ~20 puzzles we’ve posted this year.

  • Michael says:

    My opinion about the tracker is the same as SS. I can’t always get to puzzles the same day they’re released, and it gets very difficult for me to keep track if I miss more than one or two. It’s also very necessary for the backlog of old puzzles (although I at least have a pretty good idea of where I am there, since I go through them in order and I saved a decent number of them to files using my tool).

    I also agree it’d be awesome to have a more full-fledged tracker (especially for the backlog of old puzzles…), but I hope it’d be possible to either transfer the existing data somehow, possibly just in the form of getting a list of puzzles that were marked in the old tracker (so I can copy the data and re-mark every puzzle later).

    If it’s not possible to export or transfer the data, I’d appreciate it if we could at least get some form of heads-up a week or so before the old tracker is removed (so I can at least attempt to make a list manually).

  • Carl says:

    I prefer computer solving as long as tools are good (some inefficiencies acceptable but not too many).

    The four sudoku apps by Cracking the Cryptic is the closest I have seen to that. The Nikoli website used to be quite good too, but sadly never worked on mobile. The old Nikoli apps weren’t bad.

    I think clean presentation is a part of it, but I think the tools are more important (e.g. in the case of Sudoku-apps, selection of multiple cells, or in Star Battle, showing the squares that are blocked out as you). Some of the things that are part of your vision (e.g. highlighting errors) are probably too hard to do for a v1 – I would be more interested in support for a broader class of puzzles.

  • IHNN says:

    “I worry this coverage is too low, and misses important genres for us like Sudoku (+ variants) and TomTom that have better implementations elsewhere.” Penpa is another open source online interface (english translation here: https://rjrudman.github.io/penpa-edit/ ) that trades genre-specific error messages and checking for the ability to use virtually any common puzzle symbol in one grid. It even supports hexagonal grids and a constructor-set answer (a small browser alert displays on an answer that matches that). I’m unsure if the answer checker can support Tight Fit Sudoku, but such a grid can be set within. To the best of my knowledge, the only puzzles on this site that Penpa can’t support are Quintessence (maximum grid dimension of 30 in either direction, though this could be corrected) and possibly some old Patron puzzles, if they’re in the 31×45 range. Though puzz.link could likely support those if they’re standard puzzles.

  • Matej says:

    I don’t have a printer. Last week was the first time I could easily engage with the puzzles on this website without finding some tool like puzz.link and inputting the puzzle myself.

    Even if the tools only work for say 50% of the puzzles you publish, that is still a significant number of puzzles. Please, consider using even the non-ideal tools – you can always continue looking for an alternative.

  • nkm says:

    I usually solve on a tablet with a stylus, by exporting the pdfs into Microsoft OneNote. This works pretty well, but I would definitely shift to an online system when it gets good enough.

    I found puzz.link pretty good, but not quite good enough to get me to switch. Biggest complaints: No way to pause the timer, and the limited tools for making notes. For star battles, I really want to be able to mark “a star is in this group of cells” in some way.

    The best online puzzling tool I have seen is the Cracking the Cryptic web app for sudoku. I assume that’s not open source, but something like it would be great. I like that there are a bunch of different kinds of notes you can make, and that everything can be controlled from the keyboard, including moving around the grid and changing between different kinds of notes.

    On a final note, I want to echo others who miss having a tracker of some kind. It’s nice to be able to see at a glance whether I have or have not done a puzzle in the archive, and also there’s just a satisfaction of completing a checklist.

    Your vision for the future sounds awesome! I love the site, and I can’t wait to see how it develops.

  • drsudoku says:

    I know the Cracking the Cryptic team is trying to get their tool more open for others to use eventually, and I may try to ask for an upcoming sudoku week if that is something we can explore as another beta-test for us and for them.

    Also, the solver applet should now show “Mark solved” for all of our recent puzzles that did not have a key. Hopefully this restores all the functionality people were using (just don’t misclick!) as it is on the honor system.

  • InvalidD says:

    I use paper and pencil for number placement puzzles, but mainly Paint for other ones. Because of that, I find the puzz.link a bit monotone (lacking visual differentiation), but that’s probably just me.

    Also, thank you for adding “Marked solved”!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.