I believe this is safe to post since the 12 hour window was over long ago, although solutions have yet to go up.
Taking the test this year was interesting since my printer’s color ink has gone haywire and I went entirely without grays or reds this year. On some puzzles that meant I had no gridlines whatsoever.
For the test as a whole, 344 is my expected score. I haven’t talked with enough people to know what rank that will get. Here’s what happened on individual puzzles.
Update: An email suggests I have 1st, with a score of 338. The decrease is due to a coordinate messup on the differences puzzle. I blame the gridlines not showing up on that one.
This one fell in the usual way. You get about a quarter of the ship cells, then stare for a bit, then plunk a guess down and eventually tweak to the answer.
I’m still kicking myself for missing that 2 in row 1 as long as I did. Once I got that the pencil almost never left the page. It’s funny to think that Corral has gone from being the last puzzle on the test (on some early USPCs) to being the second, although to be fair the Corrals on earlier tests were far harder than this one.
It seems like most of the recent USPC Masyus like to do clusters of white circles which result in a zigzag pattern. Other than that, there was also the matter of untangling the loose ends of the black circles in the bottom left to avoid making a loop, which is yet another pattern I know I’ve seen in a nikoli.com Masyu before. Though, I imagine for people not experienced with Masyus that this was awfully hard for the third puzzle on the test.
4. Lab Rats (Take 1)
I found the difference on C2 almost immediately. Nothing else came similarly quickly, so I put this on the todo pile.
I enjoy a lot of pentomino types, this one included, and this was a great construction too. I particularly enjoyed the finish of the right half where you realize there are two pentominos and they aren’t going to squeeze in very easily.
6. Digi2rd Search
Oof. I got it okay without dropping a tremendous amount of time, but that nasty diagonal “QUI” in the top left had me put QUI9 there early. Eventually my total inability to get BREAK EVENS and KEYSTONE led me to reconsider my current words. That cost me several minutes, but I was able to correct it.
I think this is the smoothest Sudoku solve I have ever had on a USPC. Especially compared to last year where I broke it three times and failed to realize the third try had broken at all. This was a nice puzzle with a lot of box-line reductions, which seem to be used a lot in USPC Sudokus.
This was a well-made puzzle. Start with the 2-3 pattern on the bottom, then run clockwise around the center from there, getting pairs of shaded cells next to each pair of clues. It’s a nice use of the 2-2 clues. Then after noticing things can’t be connected through the bottom, you can finish by navigating through the top.
9. Persistence of Memory
About the level of those OAPC puzzles (if a bit longer), and much easier than the one on the Snake Variations contest. You can start with the horizontal domino, since the top right tells you it passes through the left edge. That gives you information on the P that transfers to the vertical domino, and the unit square follows shortly after. Once you have the small shapes, a ton of the snake is determined already, and the larger letters can be worked out roughly in the order SPCU. I remember not enjoying the Snake Variations Contest puzzle much since so much of my time was spent looking for similar shapes, and I decided I didn’t like these puzzles in this size. But there were no such issues here. Great work from Serkan.
10. Number Tower (Take 1)
Ugh. I’ve already made my displeasure with this apparent to the organizers, so I’ll keep the rants short here. In any case, the trick didn’t come to me on first look, so I moved on.
Props to the Zotmeister for pulling off a puzzle with no given numbers. A nice solve even if I don’t usually like Sudoku variations. Also as usual for him, there’s an extra minor rule that isn’t actually needed. It’s only after the test I see the bit about all squares that can be boxed are boxed, but I know I didn’t use that anywhere in my solve.
Those of you who looked at that math problem I posted were in luck; that was definitely the right kind of practice. As I predicted, the problem I posted was indeed far harder, since it allowed backtracking through the individual regions. This had a much cleaner solution. Once you figure out three path counts in three by three squares (center to edge has 7, edge to nearby edge has 8, edge to far corner has 10), the multiplication and addition is easy. 56+70+700+7000+70000 = 77826. The way I see it, there was no counting puzzle this year; just one more math puzzle.
Having my color ink not work made this interesting to work on though. All that showed up were the S, F, and 8 arrows. I basically had to solve it looking at the PDF on the computer.
This was even more interesting due to no color ink, since none of the grid lines showed up at all. I eventually brought up the PDF, got the best zoom level I could, and then PicPick-grabbed it and solved it there. I use that program for most online tests, but generally only to start the first puzzle while the printer is rolling the first couple pages out.
As for the actual puzzle, this had a tricky work-in where the 4s on the edge are much more constrained than they might initially appear to be. Once you got those, the massive zigzags in the empty spaces started. This almost always happens in simple loop types with empty space since you can’t get uniqueness otherwise. So despite the size, the rest of the puzzle falls fairly quickly.
14. Tight-Fit TomTom (Take 1)
A very tricky puzzle! I would have given this more than 20 points personally. I originally got tripped up by not finding any way to be able to put an 8 in R1C4 and have that cage work, so I made a deduction from that. Unfortunately, the puzzle only broke much later. I went to print another and did the next one while waiting for that.
15. Star Battle
There’s not a large space of things that can be done with vanilla Star Battle (one reason I am very impressed with the ones Thomas posts on his blog). So having done quite a lot of these, this didn’t bring anything new to the table. But I’ll give credit to the constructor for making a difficult puzzle, involving some subtle use of “N regions in N rows”, without forcing the use of guessing. A lot of difficult ones have no logical solution to speak of.
14. Tight-Fit TomTom (Take 2)
This try went much better. I confess I still don’t have the logical solution for this one, since once I realized my 8 deduction was wrong I basically exploited my earlier contradiction and put the 8 in R1C4 right away. It was still pretty difficult even after that. TomTom continues to be a much, much more brutal type than KenKen, not just because of the slight rule fix.
16. Kakuro Cards (Take 1)
Oops, printer can’t do reds, so I had to copy the diamonds and hearts down from the PDF. Serkan didn’t give us many sums to work with. I spotted the 1-3 pair with diamonds quickly, but nothing seemed to come of it. A couple minutes later, I decided I’d come back to this one.
17. Follow-Up (Take 1)
Oh crap, that’s a big grid. SKIP!
(Now that it’s over, I see the pair of Qs, Zs, Js, Ys, Vs, and Ws mean that there are many reasonable work-ins. But with a word bank that large, I’m not so sure I would have been able to get through to the end in a reasonable time. I think I made the right choice.)
I had a Python script brute force the positive integer solutions to rows and columns beforehand and had studied that list pretty thoroughly before starting. I started by applying it to the bottom row and… that put a 1 on the edge, which I knew was impossible for lengths less than 12. Starting to suspect my script had been bugged, I do the bottom row from first principles, and then the first column quickly broke. After more soul-searching, I eventually consider the possibility of 0. That simultaneously fixed both of my issues and also threw my brute-forcing out the window. But fortunately the puzzle wasn’t too tricky. The small amount of possibilities for each row and column meant intuition could be used to speed things along.
I am aware of at least a few people who did not catch this trick. Again, I wasn’t happy with this and the organizers already know that, so I’ll leave it at that.
19. Siamese Fences
Hello, Colorlink! Thank you for the 35 points.
In all fairness, it was a tough puzzle worth the 35, but having constructed as many of these as I have, it didn’t present me too much difficulty. I ended up doing the whole thing on one grid, copying the clues from the other grid lightly pencilled in and using a thick light loop for those clues. It worked out better than I thought it would.
20. Thermo Skyscrapers
Color ink not working meant I got a blank 7 by 7 grid for this one. This is probably where the printer lost me the most time, since I had to copy down the thermometers and I made sure to check I had done it correctly. The solve itself was hard and what you’d expect from the penultimate puzzle on the test. But after that Skyscrapers round at the last WPC and all the practice I had done for it, it wasn’t terrible. Oddly enough, I think my favorite part of the puzzle was getting the last 7s at the beginning of the solve, like the one on R3C7. Filling in the highest number is not usually such a tricky or interesting part of Skyscrapers solves.
For sure a much bigger grid than the Rehbers I had been solving yesterday (under the wrong rules no less). I should note that I was impressed with the USPC theming in the center; that probably wasn’t trivial to pull off. This was certainly worthy of being the last puzzle on the test, and it was a slow, careful solve. After getting the first few deductions in the lower right part of the center, I eventually started relying on intuition to try to deal with the wide open edges. I was very fortunate to end up correct for all of those guesses except for my last one, where I wasn’t getting TROLL and ACER to work out in the last two columns. Turns out I had them swapped, and a few careful tweaks later I managed to put it away. I felt very fortunate on this one.
16. Kakuro Cards (Take 2)
I think the break-in for this one was indeed quite hard. The key was that 4s couldn’t be used in the two 8 sums. R2C3, R4C2, and R8C6 all had to have 1-3, so that meant R5C4 could only have 5. With some numbers finally down, I was able to make steady progress toward the answer. I ran into a hangup at the end since my copying down of diamonds and hearts put a wrong suit in one of the spaces. I count myself very lucky that it only mattered at the very end of the solve and that it was very easy to fix as a result.
17. Follow-up (Take 2)
Now it was down to this, Number Tower, or the Spot the Differences. I spent maybe 10 seconds on this one, then decided against it and took another look at Number Tower.
10. Number Tower (Take 2)
I was convinced the row with five bars was probably the 3, so I was trying to figure out how to interpret the rules to make it work. The best I came up with in the small amount of time left was that some large rectangles could represent blank space (with no number), and the number sizes could be distorted. Now of course this makes a tower that is held together by mere wishful thinking, but I had nothing better, and it seemed to work out. Until those top two rows. I was pretty sure I had ways of making the second row either 1 or 7, and no real way to proportionally distort either to make the top row work. With time running low, I threw down one of the two as a guess.
I found later what the real trick was. I also found I was a swap of 1 and 7 away from the right answer… meaning I picked the wrong guess at the end. ^@#*&!
4. Lab Rats (Take 2)
About 10 minutes to try to mop up these differences. I think this one was pretty easy this year, without a lot of subtle “this thing moved” or “these lines aren’t as long” differences. Unfortunately I only had the time to find 9, and another minute might have been enough to see the flipped mouse on G3 before I did some final checking.
That final checking actually saw me correct a word search entry error, where I misread one of the 1s as an I. Whew.
So the end result was everything out except for Follow-up which I never started, Number Tower which @#$!@$%, and one missing difference on Lab Rats which I was happy with considering how bad I am at Spot the Differences puzzles.
Overall it felt like it went well. There was time lost on puzzles like the word search on TomTom where I broke both. There were also efficient solves on things I’m good at like the Tapa and Colorlink. So I believe it was a performance of my usual grade, albeit not quite the caliber of what I put up in the 2011 USPC. We’ll see where it lands me.