Suck My Kakuro

[tablesaw]

I just finished this month's puzzle test at Logic Masters India. This month's Puzzle Marathon was a different format. Instead of setting aside a few hours and trying to solve as many puzzles in the time limit, there were ten large puzzles that could be solved at any time over the course of nine days. Solving any puzzle under an hour net some bonus points, but any puzzle correctly completed is worth a base amount of points, no matter when it's finished.

I did okay. My best puzzle was the Loop the Loops, which is a combination of two of my best puzzle types. But I had some catastrophic failures on the Star Battle and the "Small Neighbors." Still, because of the way the way the puzzle was structured, I followed through and eventually got them entered correctly. In fact, because of thes coring system, I was able to take breaks to recompose myself. After trying to solve the Small Neighbors puzzle after work, I ended up getting on the Metro and finishing it on the ride home.

I knew that the worst for me would be the last two puzzles, a hard Samurai Sudoku and a hard Kakuro. The sudoku gave me a lot of trouble, but eventually I worked through it after a few hours (with a sleep break in between). At that point, I'd earned all the points I could from the test—the scoring discarded the worst score, and since I knew that I wouldn't be able to solve a large, hard sudoku in an hour that it was just going to be the base points, like several others. Still, there was a special notice appearing on the site for people who solved all ten, so I figured I'd push through, for completeness' sake.

That was one of the most grueling puzzling experiences in my life.

Kakuro is my least favorite puzzle type. I hate the number of combinations that have to be constantly checked and rechecked, especially at harder levels. And this puzzle inevitably magnified everything about Kakuro that I hate. And that's in addition to the portions I had to do over a few times because of fatigue errors. Total time for one Kakuro: 22 hours and 50 minutes of nothing but pain.

Okay, admittedly, a significant portion was food, sleep, talking on the phone, playing Batman: Arkham City and working. But still—PAIN.

I guess that's a point in favor of the normal contest type: I can ignore the Kakuro and know I'm not losing any points.

Comments

[devjoe.livejournal.com]

The Kakuro didn't take me that long, but at just under 2 hours, it was my third-longest solve, so know that you are not alone here. I also spent 2.5 hours on the different neighbors. It was the Samurai Sudoku, though, that kicked my a**. Doing it during the work week - and solving a whole other contest while I was supposedly doing the samurai - added to my time, but it still took me many hours of actually working on it out of a solving time of 3 days and change.

My fastest solve was actually the braille word search, perhaps due to having spent time trying to solve Blackout from this year's mystery hunt before working on it, and hence being somewhat more familiar with braille than I normally would be. Solved in 16 minutes. With this kind of event, it's theoretically possible to train yourself up for each puzzle type during off-clock time and then do each one while the strategies are freshest in your head. But I wasn't really worried about that, it was just a fun thing to do, and I was happy to solve all the puzzles and get bonus time points on 3 of them.

[tablesaw]

It was odd to see so many people complaining about the sudoku, because I solved it in about exactly the time I expected to solve it. Maybe I'm doing better at sudoku strategy than I thought, though I still have a hard time actually seeing the places I can use the information. On medium-level sudoku, I fairly regularly go through complicated logic to prove the contents of one cell, then notice that I could have filled it in using much simpler logic if I'd been paying attention.

As for training, I'm very glad that I'm not at the level that I need to worry too much about that kind of optimization, but I still need to log some practice to get more adept at various types. I especially need to stop making the rookie mistake of not working on practice examples before a test, which probably would have saved me some time on different neighbors.

[(Anonymous)]

I was passed on this link, and as author of aforementioned kakuro I think I deserve the right of reply to this little tirade.

In my opinion the puzzle was completely fair - and I'm sure Deb would say the same thing about the samurai sudoku. The crucial thing with large puzzles is knowing where the right place to look is. As it happens, the samurai could be solved puzzle-by-puzzle going clockwise from the top left - but I don't think many people cottoned onto that fact, which is why it took so long for everyone, myself included. My kakuro was designed so that once you had a break-in, everything else followed - any attempts to flit about the grid scanning for other break-ins if you missed something would probably have left you very frustrated. There was certainly no need to know many obscure combinations, and those that there were always reduced to something more familiar and unique once you had a digit or two.

My big tip for the future is if you are stuck with a puzzle, look to generate further progress in a puzzle around the areas you have already made a bit of progress. I think the thing you need to remember, especially with these LMI tests where people are trying to show off the best of their own puzzles, is that there is always a "nice" solving path to these things.

Sorry to hear I have been responsible for so much "pain",

Tom.C
http://blogs.warwick.ac.uk/tcollyer/

[tablesaw]

Oh, I'm sorry if my rant gave you trouble. The things I hate about kakuro are . . . kakuro. So I assume that everything you magnified were things that you were supposed to magnify. Keep in mind, also that I haven't done anything above a medium-level kakuro in a long time. In short, not every painful (for me) puzzle is badly designed, and I had no reason to think that this one was.

But since you're here, I'll give some more specifics.

The lack of multiple break-ins made this a much more daunting puzzle, initially, since it took me a long time to find the starting point. I actually started the kakuro after finding in error an my sudoku, and the search for break-in was enough to drive me right back to the sudoku.

(I don't know if my memory is accurate, but I seem to recall forcing a way into the section below the W before I found the true beginning.)

After finding the break-in, things moved steadily (though still gruelingly, as I was fighting for each cell). This continued up through solving the M where the sudoku logic worked nicely. But starting at the W, things took a turn for the worse. For whatever reason, the pairs I was seeing in the W were not lining up correctly, and they required a large cascade to disqualify guesses. I ended up reworking this section three different times (at least one due to simply writing down the wrong number).

The lower right was the worst, though. This was the main place I was dealing with having to keep track of all of the possibilities of numbers to move forward (even when my feeder numbers were correct). If there was a clean path of solving down her I was far, far, away from finding it. I don't think it's too far off to say that I spent as much time in the W and upside-down L as I did in the rest of the puzzle.

Actually, let's take a moment to look at solving path and difficulty. If I had more experience with kakuro, I might have known that there was something fishy going on when I needed more guesswork and combination hacking. I know that when I get seriously stuck on other puzzles, it's a good time for me to check for errors. But since I don't know kakuro as well, I had no indication that I'd gone off track. And for whatever reason, once I'd reached the W, it took a very long time to come to a contradiction. (Earlier errors were contained enough that I could wipe and redo without losing too much work.)


Might as well put a note here about my experience with the sudoku. I did solve the odd and the extra region in order, but I was having a lot of difficulty with the irregular. I assumed that I would need the X next (and I already had managed to wring out a few digits that had helped with the top left). I put in a guess that led to a complete grid and used that to break into the irregular.

I screwed up the irregular, but in the process, I learned how the geometry worked, so when I printed out a new grid, I was able to work the irregular as it was supposed to be done, since I still wasn't sure if my solution for the X was unique. I got everything right the second time, and it turns out that I did have the X correct, so everything worked out pretty well. I'm now back solving some of the Shapes and Sizes Test to get used to irregularly shaped sudoku again.