Log in

16 June 2007 @ 03:31 pm
USPC - full summary  
themed Fence Post

"reverse of evil"-themed multiplicity

An admittedly too easy fridge magnets

Two puzzles by thedan
"DK Fencepost" - [Edit: The first version of this I had posted had a missing digit in the fourth row, the 1 which you now see in this file. The error was mine in type-setting, not Dan's, and I apologize to those who stumbled with a non-unique puzzle. This goes to show that you should not try to do anything productive in the afternoon after the USPC when your brain is not yet recovered.]

"DK Multiplicity" - I greatly prefer his LED method, apparently from MS paint, to mine above

So when I saw the make-up of the test yesterday, I figured that finishing it would be fairly likely, maybe even in 2 hours. There were only 20 puzzles overall (last few years have been 22 and 23) but amazingly no bonuses except on the spot the differences. All the high point ones, except perhaps the magnets which is always a slog, were of types I tended to be able to complete without too much trouble. The double murder seemed very similar to some twin killers (minus box constraints) I've done before. I presumed the tiling would be the hidden hexagonal grid puzzle.

thedan and I shared some puzzles to test the newer types and I put my examples up above for those not too puzzled out to try in the future. I definitely felt I got comfortable with both the multiplicity and fence post from this exercise. The only puzzle I really worried about, as I've never gotten one right at any USPC or WPC, was Count Me In. Somehow I always overlook one or two of the rotations shapes, or forget I have 5, and not 4 or 6 fingers on each hand, or who knows. I can't count.

So, the test arrived and as I often do I solved the first page while my 2x copies of the test printed out. The battleships wasn't so bad and the diagonal sudoku took a little play with the diagonal (for so many givens in the corners, it was moderately hard - a 3 to 4 minute one certainly). Sum Figure was too easy to not do immediately and fell in a minute at most.
I then got back to my desk with both tests in hand and proceeded from the back to the front.

While the ORu Kakuro was a stretch of a title to get the palindrome, the puzzle itself was quite nice. After some searching for the break-in, it fell nicely outwards from that one spot.

The No Parking went smoothly with me trying to use On One's Honor as a break-in word even though it had several places it could go. One thing I like about these criss-crosses is that once you approach the answer, your grid starts to look more convincing. There is one particular set of crossings in the middle that give this feeling and once I found it, I knew I'd finish without problems.

The double murder was about what I expected it to be, although with only row and column cages, there weren't the innies or outies I was hoping for. It was a challenging 35 pointer - probably took about 10 minutes - but a good solve. Twice I wrote wrong numbers while copying the grid. One of those times I repeated a one in a column and thought I was at a broken point as I couldn't complete a sum. I was maybe a little frazzled by the effort it took to fall but it finally came.

Multiplicity was another fun puzzle then which, from our pregaming, I had a good intuition for. There is an early assignment you can make based on the 6/5 shapes and the 0 products, which led to an assignment of one of the products fast (from my puzzle written yesterday I further knew 6 had to multiply with 4 or 9 to give a 4 in one of them which helped a lot), and the other a bit slower, but still probably within ten minutes.

The Corral, while a type I've seen often, was a neat little surprise this time around. Not because it was a hexagonal grid, although that also was a neat surprise in '04. No, what was neat was a solving pattern that arose along the outside edge. Many people probably noticed this at some point in their solving, but it was a really neat embedded theme. Of all the puzzles, this is the one I'd recommend removing some points from, going to a 25.

Gull from Ipanema was a slightly less painful version of this regular type. I've done lots of the Nikoli Giants of these (well, 3 to 4) and so I have a good note-taking style to make thematic tiles, check them all, then move on. So it was an efficient 3 minute search I think to find the three tiles.

Magnets was my one worry, so I looked at it some, made a little headway, and skipped to the fun fence posts on the same page. As expected, that was a nice fresh type which followed the kind of solving strategy I'd looked into the day before. canadianpuzzler always provides fantastic puzzles for the USPC (my failures with Atomic Fusion last year not withstanding) and his new fence posts type and multiplicity were some gems.

Went back to Magnets, decided I still needed some comfort puzzles before solving the last biggie. So turned the page to my favorite easy Nikoli type Masyu. Only this wasn't an easy masyu with defined work-ins. The hardest work-in I've tended to need on a Nikoli is actually a black circle two squares from two touching white circles, negating a direction. One up from that in difficulty is what was here, namely white circles diagonal to black circles communicating the directions things can go. In one place, two diagonal white circles specifically gave a direction and in another both directions. Needless to say, I didn't see either of these at this point, so I made some guesses to get simple contradictions. Then I proceeded to blank when it come to the upper-right. The top was forced, and I had a loop right under it. I saw I needed to turn, but I only thought I could turn up and that was going to close the loop. After 2 minutes of erasing and thinking it was broken, I realized it can also turn down. Not sure why I wasn't seeing that, but its a typical hiccup after a stretch of hard puzzles so I can accept it.

Skipped over the tiling for the moment, did the very easy (except for answer submission) arrow ring. Carefully checking my internal, external, and loop square counts, I found the extra hidden "outside" square which didn't touch the outside edge. Round Trip was even easier

All About Rio, a type I like consistently reappearing if just for the couple minutes of relaxation, was next. Somehow this similar size word-search has varied from ~35 to 15 points in the last couple years. 15 feels about right. Double checking my letter-writing, I got that submission in fine.

Onto headache indu... I mean circuit maze. I'd gameplanned before that because of the black/white color problems, I'd just use a pink pencil and make the thin walls decidedly colored. Black and Pink don't cross streams in my brain, so I could see everything fine. I started in a likely point, worked along a path. When I found the dead-end problems, erased backwards, then continued this process. It felt like it could have been worth five more points, but I eventually got a loop. Took awhile to make sure I wrote all the letters in the circuit right. Turned out I initially had them counterclockwise instead of clockwise but I caught this on checking [although apparently it wouldn't matter]

Spot the differences went uberfast. My one worry was the two or so differences that had a dedicated home cell of difference, but another cell in which the difference honestly continued. The feather is a good example, as is the backwards paw, but I found 10 legit different differences and moved on.

Comic Strips was the fun puzzle I hoped it would be. I found me my scissors and tape and went to work. I initially made a false grid by trying to place TENUTA horizontally. Got the correct grid when I worked off another name and then saw how the TENUTA was meant to arise. A clean, enjoyable 15 pointer.

Looked up at the timer and I was at 40 minutes to go with the magnets to finish and the tiling the count me in. Went back to the magnets and used some forced guesses to get some contradictions and make progress. This is not to say that zotmeister's puzzle isn't doable. It definitely is. Just that my sharpness was not there at that moment.

Tiling could have been faster, as the wrong one stood out as very wrong pretty easily, but I checked all the other ones. Having several copies of this page was helpful which is why I preprint more than 1 copy of the test.

Finally, my nemesis, the counting puzzle. I figured a good way to not err was to meticulously count triangles that used a particular vertex, and work from the left to the right, or right to the left, or top to bottom, from each of the main vertices of the triangle. I got the nice count of 36 coming from the right. I got the same count, 36 coming from the left. Working from the top, I initially got just 34. I checked each of the vertices and didn't see missing ones. I checked again and found one of the problems to get to 35. I checked again and found another missing one to get to 36. Considering how hard it was to find those 36, I figured the triple count was good enough. I'd actually failed in three different ways to count one additional triangle. As I've said before, I really really cannot do counting puzzles. Dissections are the other kryptonite, if you really want to cut into my USPC score Nick, but counting puzzles I'm 0 for 6 in competition now.

So I submitted answers at 2:06 and change. I started checking everything (not rechecking the counting as I had three independently wrong verifications on that). Found I had the circuit maze entered counterclockwise, not clockwise, and from the "topmost" letter, not the "topmost" elevator. After two separate resubmissions to correct those problems, I was done in a little over 2 hours and 11 minutes. I finished checking everything else, saw no errors, saw the clock had run out, and my USPC was over. Until I saw the answer key and photoshopped the triangles, I thought I was perfect. The 36 for 37 will haunt for a little while, but I did as well as I could and again proved my belief that the last 30 minutes of a USPC are the worst for me. Year 1, I lost 5 points during that time, year 2 I gained 0 points, year 3 I gained just 10 points, this year, I lost 5 points again. So net result, 0 in each of the last thirty minutes. Maybe I'll petition for the 2 hour version of the USPC next year...

So many fun puzzles, at least 19 but my count might be off by one. Thanks to all the puzzle authors and especially Nick Baxter and Will Shortz for assembling another excellent test.
jrivet on June 17th, 2007 04:02 am (UTC)
The counting puzzle is actually fairly trivial (I didn't notice this until after the test, of course, though I did get it right):

Start with the big triangle. That's 1 triangle. Every time you draw a line from a vertex, perpendicular to the opposite side, that adds two triangles. Then you just have to redraw the puzzle, counting the lines.
motris motris on June 17th, 2007 04:07 am (UTC)
Excellent observation! One of many cool alternatives to counting that I would have loved to have found.

Even when I was submitting answers for the corral/arrow-ring I had a heck of a time counting and recounting squares in all ways to make sure I didn't have an error in my submission. Somehow, just as with basic multiplication at the WSC, counting is a skill I lose under pressure.
(Anonymous) on June 17th, 2007 09:55 am (UTC)
The Least Happy Place on Earth
Somehow, despite a goodly amount of (what seemed to me to be) legitimate preparation (old tests, foreign championships, etc.), having been immersed in the whole competitive puzzle culture for a handful of months, and an improvement in environment -- moving from a noisy hotel business center to the comfy confines of the Luxurious Zuffranieri Villa -- despite all these changes, by my count my score on the USPC went down. And it wasn't high, at all, to begin with.

There was at least 45min of "effort" directed towards puzzles that I didn't ever finish, which is only surpassed in disappointment by those minutes I spent on one (maybe more?) that I screwed up. I guess I can take comfort that the initial answer key I saw was indeed fouled up, else things would have stooped even lower.

Other than the dismal performance, my only real gripe was with the maze one. I think it's a groovy concept, but if the goal was to bewilder contestants with a bizarre Pollock-esque visual -- mission accomplished, I would think. I guess I was expecting to see separate level/floor representations. Not that I'd have gotten any points on it anyhow...not today at any rate.

Finally, after last year's USPC, I decided to take Fences puzzles more seriously, and have improved to the point that I'm not embarrassed with my times more than maybe 5% of the time. Well this year's attempt at Corral went nowhere...and even my pressure-free solve took far, far too long. Anyone got an idea where I could get my hands on a few more of that type? I think I'd seen one in a previous USPC, but are there more to be had? Whether one's on next year's test or not...I wouldn't mind attempting to exorcise those particular demons, especially since all the other demons from today are complaining about overcrowding... :)

Good job Thomas, and I hope Jonathan did similarly well (or better).

motris motris on June 17th, 2007 02:26 pm (UTC)
Re: The Least Happy Place on Earth
It seems everyones score went down a little from last year ....

The past USPC corral won't be tremendous help as it is hexagonal.

There are about 10-12 corral puzzles in the tuller/rios math and logic puzzles books that go under the Mensa labels. Specifically, I think they lead off the blue book (as opposed to orange). Those books have many good logic puzzles of different varieties and were my main practice before last year's test. They also mark a quaint time when Number Place and Cross Sums were how we referred to some puzzles.
jrivet on June 17th, 2007 03:26 pm (UTC)
Re: The Least Happy Place on Earth
There are also rectangular corral puzzles on the '02 and '03 tests.

Some of us still call them "Cross Sums". They only called it a Kakuro here because "Smusssor Cross Sums" doesn't make any sense.

phat_joe phatjoe on June 17th, 2007 08:21 pm (UTC)
Incredibly, I got the counting puzzle right, even though last night when I tried redoing it I got 5 different results.

Unfortunately there were at least 3 puzzles that I worked on all the way to the point of failing. I redid one of them (the diagonal sudoku which I made some kind of too-fast stupid mistake on) at the end.

I think my final score is 110, up from last year's score of 90. The difference was having a printer. Still, I should have had at least a 150 here. I was not "on".

nickbaxter nickbaxter on June 18th, 2007 12:11 am (UTC)
re: palindrome
Well, thank you for at least noticing the palidrome. I thought it was pretty good, and much better than the original (Cross Some).
lunchboy lunchboy on June 18th, 2007 06:08 am (UTC)
I didn't compete officially (had conflicting plans that day) but solved on the bus ride home from Boston. Scored 199, missing one of the difference in Brazilian Wild Life. Although I think there's one more difference -- the bar of the unicycle is vertical in the first picture, and noticeably off vertical in the second one. No idea what square one would assign that to, though. It's interesting that the magnets puzzle slowed you down -- it was one of the quickest points-to-solving-time ratio for me, whereas I didn't have the foggiest idea where to even start working on the Fence Posts puzzle, and even though I've solved Corral puzzles before, I got completely stymied on this one. Masyu I'm never good at, and this one was no exception.
motris motris on June 18th, 2007 06:41 am (UTC)
Looking into your comment more, as in past years, it seems one of the images in Brazilian Wild Life is 5% larger than the other. I tried to directly overlay even with the shrink but the bottom one is also about 4 degrees rotated from vertical. So while you saw it in the cycle's vertical piece, the difference would have arisen from "overall distortion" as in the instructions. I can sort of identify the rotation when I look at the snake/its trail.

Not sure why magnets was slow for me, although I definitely am best at number puzzles and then loop puzzles and this is in a slower category for me when I think WPC-wide. I might not be as comfortable making the logical assignments when a magnet is very very likely, but not yet in my mind solid.
lunchboy lunchboy on June 18th, 2007 06:51 am (UTC)
Yeah, I had one pair of magnets on the bottom row swapped from an earlyish incorrect assumption, but fortunately it was easy to fix.

Didn't occur to me that "overall distortion" meant anything other than "there might be print glitches, so don't worry about crap like that". In retrospect it makes sense that they would introduce a slight skew, so as to prevent someone from just taking a screenshot and flipping the image and thus being able to instantly stereogram the differences. Didn't notice it anywhere else, of course, because it's most noticeable when the original image contains a strong vertical.
(Anonymous) on June 18th, 2007 06:34 am (UTC)
DK Fencepost
I'm giving this one a go and...it looks to me like there are two areas of nonuniqueness (totaling 4 solutions). I'll be interested to see if anyone else gets to this conclusion...

motris motris on June 18th, 2007 06:42 am (UTC)
Re: DK Fencepost
Dan's original version had one point of non-uniqueness and I'm pretty sure I address this by adding in a digit here. Could you point to the places you see problems?
stigant on June 18th, 2007 04:48 pm (UTC)
Re: DK Fencepost
Yeah, I think I have 2 solutions to the upper right corner. I think there are two ways to deal with the 2 2's on the far right side. The path can go between them, or below the bottom one and to the left of both, then across the top of the top one. I'm not sure if the rest of my solution is unique, so there may be some other solutions. I can post the solutions, but I don't know how to spoiler them out.
motris motris on June 18th, 2007 05:00 pm (UTC)
Re: DK Fencepost
No problem - I got an email from Jason who started this thread describing a similar result to what you are pointing out so I will try to recapitulate it once I have more time this evening. You can always email my drsudoku at gmail dot com acount if you have an easy image to show your answer(s).
motris motris on June 18th, 2007 05:54 pm (UTC)
Re: DK Fencepost
My apologies to everyone. Indeed the error was one of transcription. The two sites you have found for ambiguities were all caused because I forgot to put in a 1 from Dan's text file when I went into Adobe Illustrator. The specific location is in the fourth row, the second to last column. There should be a one which touches the 3, and two 2's diagonally. Dan can write puzzles. I just can't copy-edit the graphic work right. I should know better to test solve the actual copy, not to hand-write a version and test that, then digitize and not retest.
lunchboy lunchboy on June 18th, 2007 05:19 pm (UTC)
Today I retried Fence Posts and quickly spotted the break-in point at the bottom right, after which the whole thing fell fairly quickly. Similarly, I retried the Corral, found the simple-to-correct wrong assumption I'd made at the beginning, and got through the puzzle relatively expeditiously (rediscovering my Corral solving methods in the process). Amazing what a difference in ability to focus the time pressure makes, even when you're not actually properly competing.
ivoryvampire on June 27th, 2007 01:14 pm (UTC)
Hi, Thomas!

I've just found your journal and dropped by to say that your performance on the USPC was impressive!
This is the first time I've ever entered a puzzle's competition. My score was not very good, only 100 points, but I had so much fun that I can't wait for next year's test! With a bit of practice I know I can do much better!
By the way... I'm the only Spanish on the list... XP

I've added you to my Friends list so I can get to know more about the "puzzling experience" XD Hope you don't mind.

Congratulations! And keep it up ;)