This story starts 4 years ago, when on a bit of a whim I took the 2019 USPC during its competition period. I had no real illusions of qualifying that year and wanted to see how I'd do - I was improving quite a bit over time and ended up placing 13th US. Far outside of the top spots, but a decent showing. The rest of the year and beginning of 2020 showed even more improvement with some excellent placements in the years' WPF Puzzle GP, and the US qualifying methods had made a mention of top 10 Puzzle GP guaranteeing a spot, which was something I was well on pace for!
Then, of course, we hit March 2020 and the covid pandemic shut in-person events down. I probably wouldn't have been able to make it to Shanghai anyway, in hindsight, but I absolutely would have done my best to try and was 2nd US in the puzzle GP. Of course, 2021 was also shut down. At the end of the GP that year, I ended up in 6th place despite a pretty rough answer key penalty in one round costing me a top 5 placement. This was a huge confidence boost that if there was a WPC this year, I would have made the cut. 2021 also saw the online WS+PC with 10 rounds - I really had to work to fit that into my schedule, and I wrote up my experience here. The short version is I placed 8th - within the top 10 previous WPC autoqualification - in spite of having no prep or practice time, which would really have helped in round 3. I was pretty close to 6th here and felt pretty good for the following year.
2022 was more of the same. I didn't do as well in the GP that year due to a large assortment of tech issues - ranging from a faulty keyboard to the GP server itself going down on me one round, ending in 18th. That there was a WPC came late and team selection came even later, and though I was offered a spot about a month before the event, I hadn't gotten a passport because I didn't think I'd done well enough to make a spot, going by GP results from those tech issues. This was really frustrating - arguably a third straight year I would have had a strong chance to attend the WPC officially, with no real opportunities to show for it. And with Thomas Luo performing incredibly, alongside Walker Anderson, Thomas Snyder and Palmer Mebane forming a strong core... it seemed like my future role would be first alternate, especially if WPC results would result in continual free passes for that core.
There was one glimmer of hope though, one opportunity I could be virtually assured of getting the chance to compete, and getting the chance to prove myself. WPC 2023 was to be held in Toronto, with the GMPuzzles team - including two of the US team core - writing. I even requested a passport in May so I'd definitely have one and be able to travel. I'd continued doing extremely well on LMI tests (rarely out of the top 5, winning 3 dailies) and PuzzleDuel (never out of the top 3 by rating, reaching 2700) throughout this time as well. By any measure - except one - I was virtually a lock.
The 2022 USPC - run in January 2023 - was a bit of a disaster for me. It's not worth dwelling on too much but I wasted over a half hour on the Score Cards and had two half finished puzzles at the end of my time, and hadn't even looked at what turned out to be one of the easiest puzzles that was also valued quite highly. This was stated as potentially having an impact on selections for 2023 so placing 4th was... not ideal. And as the registration deadline for Toronto - September 1st - loomed, there was still zero communication from the US organization. The longer this went, the more likely other historical results would be used, where I was the clear frontrunner (outside of the locks of tluo and Walker).
And then, a few weeks before, the USSC and USPC were announced. For September 2nd and September 9th. After the registration deadline. To say I was unhappy would be a bit of an understatement; in my mind I went from "I am going to be competing in Toronto" to
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH now there's a chance I won't make it because the results are entirely based on this one test and the USPC isn't usually representative of the sorts of puzzles in WPC instructions and USPC 2022 went really badly and it's a fixed start time of 1 PM and I've been sleeping from about 6 AM to 3 PM recently and I can't go backwards on my sleep schedule I just don't work like that so being awake and alert for 1-4 is basically the worst possible time for me and also the results riding on one test ignores an entire year of extremely strong results that really should make this decision easy and even if you were to go by last year only I'd still be fine AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHH"
Yeah. Lots of extra stress. Not fun.
And then I found out that USPC instructions only drop the day before the event. On a Friday. And obviously, people prepare and practice for this because of how much rides on it. And I wouldn't really have time to prepare. Awesome.
So I requested the day off from work so that I could rearrange my entire schedule to have some time instead of no time. I forced myself to stay up later each day to eventually cycle when I was waking up from 3 PM all the way forward to 6 AM. I thought ahead about what the prep day might look like. And I was still stressing constantly about how much was actually riding on this, knowing that due to the format one small mistake could cost me.
All of this to say, I felt like I had several years of additional pressure on me for this, that I felt I shouldn't even have to be dealing with. A big part of doing well on a puzzle contest is the mindset aspect and being in the right headspace, just like with any competitive discipline.
The instructions were already up when I woke up on the 8th, the day before the USPC. I looked through it immediately and was pretty happy with the mix on the whole - a lot of classic genres, some neat variants, and a couple offbeat rulesets. I also can't forget: Score Cards, the absolute abomination of a "genre" that killed my score the prior year. But here's the thing - I know it killed me last year when I wasn't expecting it, and so now I am expecting it. I grouped the puzzles into 4 categories: puzzles I didn't need to practice (stuff like Nurikabe and Yajilin), puzzles I just wanted to solve the example for (Network, Pentominous Skyscrapers, Minedoku Myopia, NonSudoku), puzzles I wanted to practice with material available (Canal View, Sudoku, Score Cards) and puzzles I wanted to toy with writing (Double Kakuro Invisible, Seven Star Fences). I also tallied up the points at 325 total (expected A-team cutoff of about 260-270 points) and figured out my plan of attack for the round - Nurikabe first to start strong, then into either more shading (Canal View) or the Pentominous pair before jumping around the variants. I was aiming to leave just Score Cards and NonSudoku to the end and decide what to attempt based on how much time I had remaining, which even if I got no Score Cards would still be 265.
29 hours remaining, time to make the most of it.
I started by printing out and solving all of the esoteric example puzzles, which didn't end up teaching much but the difference between solving a type for the first time and solving a type for the second time is notable. I then went through past USPC rounds and picked out Score Cards and 20+ point Nikoli Sudoku puzzles, finding 6 (including 2022, which I forgot the solutions for) and 4, respectively. I spent up to 5 minutes on each Score Card, looking up the solution if I didn't get it in that time, and noted down the trick to each one in turn for future reference. Most of the way through this process I realized that taking the result number and multiplying/dividing by .X for each given number was a really good way to quickly search the possibilities, as it turned the problem into just asking if a given value can be reached from 2 numbers instead of 3. I then did the 4 Sudoku puzzles in 45 minutes (with one broken puzzle) which was about 2 points per minute. Given that I am not a Sudoku specialist at all, getting finishing pace was really good, and more importantly I noticed that the 20 point puzzles never went past an X-wing while the 25 pointer had a hidden triple. Somewhere in this time I kept reviewing the answer key formats and eventually noticed that the exact answer rows were specified, leaking a lot of puzzle size information. Large Nurikabe and Yajilin, probably small Fences, almost definitely 12x12 Minedoku, for instance. And a 15x15 Pentominous Skyscrapers (oof).
Matthew - my 4 year old son - woke up around now, so I brought up the Canal View puzzles from the GMPuzzles Shading Variety Collection and the Double Kakuros from Art of Puzzles 2. I solved the entire Canal View section - prioritizing this since it's a 20 point classic on a less familiar genre - first, realizing a few connectivity tricks and potential uniqueness deductions. I also solved a dozen or so Double Kakuro puzzles, getting a feel for for how the larger cells influenced the common deductions and eliminations. It definitely made scanning harder on me! I finished that up around noon, having been preparing for about 5 hours at this point, and I decided to take a puzzle rest for a few hours and just try to relax and not stress, without much success.
Around 4 PM I went back to preparing by trying to write a Double Kakuro Invisible where every clue was a possible singles-only clue. I didn't get very far with this construction as I was struggling to get it to do what I wanted, but I learned a lot about what kinds of formations would force doubles. I then tried to write a Seven Star Fences with no Slitherlink clues, and while I did get most of the way through a puzzle with this I also didn't finish it. I found a few useful tricks that mostly weren't relevant to the actual puzzle, and figured out how to bookkeep the used loop orientations better. After a few hours of this I sat back, watched some videos, and went to sleep.
I woke up on the 9th around 7:30, plenty of time to wake up and be alert while also not being so early that I'm getting tired during the USPC. Perfect start, this was something I was genuinely concerned about. I did one more Kakuro, one Sudoku from The Art of Sudoku, a Minedoku Myopia (16x16!) that ft had made, and then spent the rest of the morning just trying to not get inside my own head. Noon, I prepared and ate lunch. 12:45, I went down to my room and prepared my solving area. Then, I just waited until 1 PM for the test to start.
Yes, I have in fact written over 2000 words before even getting into the test itself. Also, thanks to Nick Baxter for sending me my submission history so that I can have accurate stats for this part.
Nurikabe (10) first, exactly as planned. It was an 18x18 so took a while, but was very smooth. Submitted at 4:39 (2.15 ppm). I then elected to move straight to the Pentominous pair, getting the Borders (10) after another 6 minutes (1.61 ppm) and made a lot of quick progress on the PentoSky (35). I ran into a bit of trouble in the middle of the puzzle and carefully worked through the counting argument, steadily arriving at the final corner. Which ended up being a horrible bash that I was extremely careful about checking given the value of this puzzle, and I ended up solving it at 1:35 PM (24 minutes taken, 1.44 ppm). However, I made an answer key extraction error and did not catch it at this time, putting me at 20 points active with 20% of the time gone!
Given the slower variant solve, I then moved to the Canal View (20) and knocked it out in 5:54 (3.38 ppm). I was approaching 45 minutes in so it was time to look at the Score Cards (5+5+5, +15 for all three) and I got two of these in about 7 minutes. Didn't quite see how to get the first one, so I submitted my two answers and moved onto the Aqre (15). 3 minutes later I had it solved (4.66 ppm) with some nice connectivity logic through a region that restricted it much more than first appeared. Minedoku Myopia was next (25) where I took an embarrassingly long time to find the Star Battle-esque step that broke the 12x12 irregular open. I made a ton of progress without it, but ended up with a very slow feeling 11:54 solve (2.1 ppm) submitting just after 2 PM. I kept the variants rolling with the Kakuro Double Invisible (25) and knocked it out in 6:09 (4.06 ppm) - there really weren't any hard steps in this puzzle and it relied heavily on impossibly low values to force doubles, not sure why it was valued so highly. I then went for the Seven Star Fences (25) and ended up breaking it in the bottom left corner, after a restart I avoided making the same error and ended up submitting at 2:26 PM, about halfway through the time. I don't have a clean ppm for this puzzle as I had to go to the bathroom here, and took that detour to contemplate the missing Score Card. Exercise for the reader, how do you get to 26 from 5, 5, and 6?
At this point it was time to clean up the early easy puzzles rapid fire, which I did in order. Battleships (5), Tapa (5), Cave (10), Statue Park (10) in 10:40 for all 4 puzzles. Then the Sudoku (20) which didn't use any X-wings and instead had a tricky short cycle (???) and Masyu (15) in 17 minutes, and the Yajilin (10) in 3:18 (3.03 ppm) which was a really fast solve for the size. I then went for the Network (10) which was a slow, gradual solve of eliminating partial possibilities until the entire puzzle just collapsed after an 8 minute (1.24 ppm) struggle. Block Search (5) and Fillomino (10) both fell pretty quickly as well, though I miscounted the Fillomino and thought there was more slack than there actually was at first. And so, at 3:14 PM, with 46 minutes remaining, I found myself at my decision point, with 275 points and only the Score Cards (5+15) and NonSudoku (30) left. My plan was to go for the NonSudoku if I had 20+ minutes. Yeah. I think I had 20+ minutes. It only took 13 (2.34 ppm) leaving me with just 5, 5 and 6. I spent quite a while tinkering with different moves from 26 and got stuck on trying to get there from 65 (65*.4=26) and also stuck on trying to use an exponent as I hadn't used one yet. Eventually I realized 26+6=32, 2^5=32, if only I had a 2...
Wait.
I do have a 2
.5 = 1/2
1/2^-5=32
equals 32
That's it, that's the answer, did I actually just finish the USPC with 23 spare minutes? (370 points)
I submitted my final answer and then spent the entire remaining time checking over my answer keys and solutions, catching the Pentominous Skyscrapers error and resubmitting 5 minutes into this process (360 total points). I didn't catch anything else and was extremely confident in this result, I expected maybe one other potential qualifier had a shot at finishing this, and there were two open slots. Even if I had a solving error, I would still be well above my expected 260-270 threshold, which after seeing the actual puzzles I revised to 280. Good thing too, because I did forget (and miss multiple times!) to connect around the top right corner in the Tapa puzzle, costing me 5 points and my entire bonus, taking me down to a much closer to the pack with a score of 320.
But this was still enough to win the US side, ahead of Anderson Wang (missing Battleships and Statue Park, less than 10 minutes from finishing) and the already qualified Thomas Luo (similar pace as me, but didn't get the final Score Card at the end). The top scores overall were by Ken Endo (40 minutes bonus) and Cai Ji (30 minutes bonus), so had I taken the 5 seconds to finish the Tapa (requires noticing it of course) I wouldn't have been too far off from those marks and would have been third overall. It's a shame that my performance isn't fully reflected in the results, but this year I've been taking a lot of encouragement from how well I've been doing in general. The format of the WPC (no drops) is a format that favors consistency, and while I'm rarely the fastest, I also rarely have bad rounds. Here's hoping low variance carries me far there!
So yeah, to anyone reading this who'll be attending the WPC, hope to see you in Toronto!
It's about time I got to go to one of these.
No comments:
Post a Comment