In short, I did not have any real expectations for Prague and the WSC/WPC.

If I had goals, they matched my 10:6:3:1 statement from 2017. I was hoping for 10th at the WSC (6th place team) and 3rd on the WPC (1st place team). The better shot at the WPC tied to there being more of an experience factor across the dozens of puzzle genres at a WPC where competition rust means I will probably not be competitive at just sudoku at a WSC again. In Bangalore, I finished 10th in both events, so behind that goal for the WPC. In Prague, at the WSC, which precedes the WPC, I again finished 10th. With a day off for tourism, and having worked through the instruction booklet and many of Palmer Mebane’s practice puzzles for the US team, I was ready for anything as the WPC started, planning mostly to do my best. The organizers had structured rounds somewhat in the same way I structure genres on GMPuzzles, and I felt more comfortable than usual that I could find flow in most of these rounds.

Note: One change that I’ve done over the last two years is to focus on what I am doing in each round, but to completely ignore the scoreboard and not get caught up in the emotion and horserace of the championship. That was true this year as well, so in the summary that follows, until Day 3 morning, I don’t actually know any of the other competitors’ scores or how well I am doing except from gut intuition. There were many great competitors at the event, but I will mostly focus on the eventual top 4.

Day 1 morning

One challenge of not “competing at puzzles” for awhile is that you lose some of the skills that are required in competition at the highest levels. These skills aren’t about how well you solve different puzzle styles logically — I’m still testing those skills regularly as I edit for GMPuzzles — but how you solve different puzzle styles fastest. Pencil and eraser and trail-and-error is the right approach sometimes, or after getting some things for sure by logic.

While I know this from experience, each year — not having learned the puzzle designers’ style — picking the balance of logic and “intuition” let alone the right way to view high point puzzles challenge-wise takes some adjustment. I stumbled out of the gate. In Round 1 (Shading), a genre that often rewards intuition, I quickly got through the Shakashaka (the first in a set of first puzzles in rounds on a grid shaped like the Czech Republic) but then ran into harder logic than expected on the Heyawake. I grinded to a solution, but needed lots of tweaks (and later when the round was graded, was shown I had not actually gotten it right). I jumped forward to the hard Cave which was easy by logic and then got to the hardest puzzle of the round, the 100-point Coral. Here I did a little on the right by logic and then lots of intuition in the middle. I had the wrong logic at the start and my guessing never worked. Long after the round, I still have not found a fully logical path to solve this puzzle, but regardless the time spent on 100 points not earned in Coral and 80 thrown back in Heyawake was a high cost, considering 5 other puzzles I never looked at. 275 after 45 minutes.

Round 2 (Objects) should have been a strength. Indeed, with a lot of easy value puzzles it was really just solid logic through all of the 40 minutes. I finished the round, but I somehow managed to not write one last crossing word into the only high value puzzle the Scrabble, throwing away 85 points I should have had. It seems this was a common error as other top solvers including Ken Endo were at 345 like me and not 430+.

After two bad rounds, Round 3 (Skyscrapers) was my lowest of the championship. Palmer Mebane had written a full practice round in advance, meant to introduce all the new styles, but there were a few too many new rules and/or I was a little impatient on small 5×5 grids to use intuition over logic. This was a bad strategy and I had two answers that were marked wrong, and a lot of unlooked at puzzles. Getting 265 on a round where some were saying finished (and would be around 600+ points) was far from how I should be solving.

In short, the morning of Day 1 was one of the worst possible starts to the tournament I could have felt. I was wondering at lunch why I was here, why I was still trying to compete and not do more important things. Why don’t I wind down GMPuzzles or hand it over to other people to run. I’m wasting time trying to keep up with puzzles when I have other things I should be doing in my life. I took a long walk and played some Hollow Knight after lunch to clear my mind. Again, I wasn’t looking at scores but I knew I was mostly playing to do my best for the rest of the day and take it from there. With much longer rounds to come (60, 90, and 90 minutes) in the afternoon, there was still a lot of puzzling to do.

After Day 1 morning: Score (rank)
Ken Endo 1560 (1st)
Ulrich Voigt 1360 (3rd)
Palmer Mebane 1245 (4th)
Thomas Snyder 885 (24th)

Day 1 afternoon
Clearing the mind helped, or at least it seemed to. I felt I did fairly well (closer to 10 points per minute) in the Polygons (read: irregular geometry) round back from the break. I skipped the Slovak Sums and a few other puzzles, but had a good logical flow through the combos and slitherlinks and six winds. Got very lucky on the Boomerangs with a right guess in the first 15 seconds (I was not intending to spend too much time on it). Closed out by actually solving some of the (Hexa) skyscrapers which made me feel better after the bad round 3. Only “Coral” would still be on my mind as a possible kryptonite from the Czech constructors. While a better round, I continued my errors in round streak with a repeated digit in the Kropki part of the pyramid combo.

Round 5 at 90 minutes focused on Variations and Round 6 at another 90 minutes focused on Combinations. The subtle difference is that a “Variation” is a new riff on an existing style that usually introduces just one or a few rules while a “Combination” is blending at least two existing styles’ rule sets as is. Due to constructing a lot of puzzles, I expected a lot more comfort in Combinations and shifting between two styles of thinking but even the Variations weren’t too unusual. Nicely, many of the styles had 2 or 3 puzzles so you could get into them. My strategy for Round 5 was to avoid the high value multiplication puzzle (math-heavy puzzles are not my strength) and also some of the dominoes but get the rest. And this is exactly what I did, although I still left points on the table with errors with 25 + 70 points lost on the Fillomino with given numbers and the reduced domino with sums. Overall nice puzzles, just wish I was solving them more cleanly without mistakes but the 705 would be an ok score.

Round 6 was the closest to the point average I would be since round 2 as I also finished all the grids. While I started the round and ended the round (and apparently never got a correct answer) on the Nurikabe Tapa, which lacked a clear deducible path to the solution, I completed everything else. The Snail End View Untouchable had an incredibly high point value but determination and solid logic got me through it. Indeed, for most of the afternoon I was trying to be more patient and just think through answers, except towards the very end of some puzzles, and the points played out this way. I wouldn’t know it until later, but my afternoon was second only to Ken Endo’s as a session and I had recovered from the 24th place start to 5th.

Day 1 PM -> Day 1 total (D1PM Rank; Day 1 Rank)
Ken Endo +2490 -> 4050 (1st; 1st)
Ulrich Voigt +1985 -> 3330 (5th; 2nd)
Palmer Mebane +2030 -> 3275 (3rd; 3rd)
Thomas Snyder +2110 -> 2995 (2nd; 5th)

Day 2 morning
Day 2 brought another set of comfortable puzzle genres, and more rounds with the first puzzle on a grid of the Czech Republic. Round 7 (Paths) had more low point puzzles than high point puzzles so after getting through the two largest ones I distributed my time amongst all the rest except for the Country Road which I never had a chance to look at. The 480 points in 45 minutes was my highest ratio to that point, although for 7 straight rounds I had a broken puzzle with the smaller every third turn loop.

While I had talked about sticking with logic as much as possible, Round 8 (Dissections) was going to force me straight into some intuitive solves again which I welcomed as a challenge. Indeed some things can be done by logic, but in many of these styles like Pentominous a little bit of trial and error helps to visualize the contradictions and get to the answers. I solved from the back to the front, starting by being surprised by the ease of the Burokku at 100 points, but skipping the “codded dissection”. After Slash Pack I jumped to the big pentominoes start, assumed the solution needed to be greedy, and basically wrote it down except for one small tweak of one of the two Y pentominoes to fulfill the unusual 2×2 rule. Going forward from there I had 12 minutes left with just the codded dissection, and got it out with a little under 6 minutes on the clock. I checked that all grids were complete, but then turned in as soon as possible for 5 minutes of bonus. I could have checked — all my earlier rounds had had small mistakes — but nothing was obvious to check and find across all the puzzles. It would not be until later that evening that I learned I was clean and earned some bonus. Finally!

Last of the morning was the Numbers round, and I am generally weaker at math-heavy puzzles. But a few of these were more number placement than heavy arithmetic. Doubleblock would join Coral as a style that I did not want to see from these constructors, but there were some phenomenal TomTom (Jiří Hrdina’s puzzles being some of my favorites from the day). One challenge in general at WPCs is that by creating certain puzzles and writing certain rules, teams have a chance to guess secret constraints about the grids. In this case, a few teams had figured out, given the very unusual rules for Star, that there was exactly one possible answer given a likely assumption that all given numbers formed a consecutive set. Besides that one negative (from a competition fairness standpoint), and my own time fiddling with the Antimagic Square by guessing but not hitting the solution, I did adequately in the round.

Again, not looking at scores until Day 3 morning, I felt I had done ok but had no confidence what that meant (8th? 5th?). Turns out I had another stretch as the 2nd best solver for the half-day session. Ken Endo, who would be the best in each of these sessions, was again first.

Day 2 AM -> Thru Rd 9 total (D2AM Rank; Rd9 Rank)
Ken Endo +1785 -> 5835 (1st; 1st)
Palmer Mebane +1605 -> 4880 (3rd; 2nd)
Ulrich Voigt +1380 -> 4710 (7th; 3rd)
Thomas Snyder +1640 -> 4635 (2nd; 4th)

Day 2 afternoon
A very long competition was continuing with a few more novelty rounds and then the first team rounds. Round 10 (Double Trouble) focused on standard styles with a doubling of the rules. One thing I have critiqued at WPCs is that there are so many rules. This year’s instruction booklet had 83 pages. And the rules are so often written by authors with new words as opposed to common, consistent definitions with just the new parts called out like “Standard [genre] rules; also, [the one new thing]”. So it takes a lot of time to just read things I should already know and sometimes within that I’ll miss the few words which are different. Long story short, at some point in every WPC I will be solving something without knowing the rules and therefore miss the puzzle. This year it was the Double Shikaku with a constraint of clues in dominoes being split (in other puzzles it was often any numbers can be in any of the grids). So I got to an “answer” for 90 points that wouldn’t prove right as I wasn’t solving their puzzle. I also struggled with another of the Doubleblocks but got everything else.

Round 11 (Regional) played with sub-area shape variations on top of standard styles. I mostly got the points I expected except for a couple where I encountered errors while solving and didn’t think I could tweak. Round 12 (Innovative) was similar, although I hit the first non-unique puzzle in the competition which gave me some concern. These were shorter rounds and I only felt I did ok, but I was more consistently around 9-10 points per minute. I still had some errors, including another repeated digit written right at the end of Mirror Labyrinth. But if my morning of Day 1 was a low point, I felt I’d recovered so that the playoffs and a strong US team result were possible.

Two team rounds followed, both hit by having a task the whole team was working on at the same time and trying not to bump elbows. Both had fairly high challenge but we got through both, one time a little ahead of Japan, our main expected rival, and one time several more minutes behind. Notation was a huge challenge for us on that round, and we had to erase and restart a few hexes of that Coral/Easy as ABC variation.

Day 2 PM -> Thru Rd 12 total (D2PM Rank; Rd12 Rank)
Ken Endo +1775 -> 7610 (1st; 1st)
Palmer Mebane +1160 -> 6040 (12th; 2nd)
Ulrich Voigt +1305 -> 6015 (4th; 3rd)
Thomas Snyder +1180 -> 5815 (10th; 4th)

Day 3 morning
Having not read the scores throughout the tournament — maybe even thinking I was out of the podium running after the Day 1 morning — I was filled with energy to see the scores on the morning of the 3rd day. I was in 4th, and had had many rounds where I was outperforming Ulrich and Palmer so was capable of having a good result.

But here is where the competition structure gave me a small challenge. There were three rounds to the playoffs in this event: Players 7-10 would compete with puzzle selection and time bonus favoring the better ranked players. One person would advance to join Players 4-6 in a semifinal. That playoff would set up the finals with 1-3 and the winner of the semifinal. I was starting the day just ahead of others for the top semifinal spot; Nikola Zivanovic was 5 points behind with 5810 and Kota Morinishi at 5790 before Bram de Laat at 5510. I felt comfortable I wouldn’t fall out of the semifinals, so I started to think about how to skip them entirely. At 7 AM in the morning I’m starting to visualize how I want the day to go, to get to what I thought was a 5% chance of winning the playoffs.

The very first thing, in the last 90 minute round of the individual competition, was that I needed to solve my brain out and also hope for an error or poor round by one of Palmer and Ulrich to make up 200-225 points. Specifically, I figured this meant finishing at least 12 but more likely 15 minutes early, something I wasn’t that close to all tournament except in Round 8. If I could do that I could maybe get a straight path to the finals, I’d probably be close in time to everyone else there except for Ken Endo. Then, solving my brain out one more time and hoping for some Ken Endo mistakes was the path. Only two of these four things were in my control, and after breakfast and a good morning playlist to get mentally ready, I got started on Round 13.

Moreso than in past tournaments, I did a lot of gameplanning for how to go through rounds. Here I decided on a lot of preprogrammed jumps. I started in the “Wrong” section where all digits were +/- 1, but solving backwards so I could end on the Doubleblock. I then jumped into the coded section starting with the other Doubleblock, then solving forwards through the Coded and Liars sections before solving the easier missing clue starting puzzles. I had time targets for each of these runs, another thing I’m not used to doing. With 180 points for the “Wrong” section I was hoping to see at least 72 minutes on the clock. I actually saw 76! The Doubleblock was the only thing that slowed me some but it was the hardest of the 4. I didn’t make its first move logically — I used a forcing deduction that proved false — but that gave me everything I needed to solve it. Getting to the coded doubleblock I then wasted 2 minutes solving it like is was a 1-4 puzzle, not 1-5, at least in interpreting some of the sums. Probably lost some time there, but saw the Aha from constructing very constrained skyscrapers before in the next puzzle to make back some time. The coded Coral went fine, and then I hit what was a huge potential concern: The Coded Arrow.

One thing about competing at the WPC for so long is that I have a very good sense of puzzles I am good at and others I am not as good at. ~10 years ago some of my negative posts might have been complaining about F*ing arrow puzzles. I’ve taken time to get better at the regular form here, but when you use coded clues or +/-1 clues, it can get much harder. One lucky break I caught was that this year the new Toketa puzzle volume 6 had a section called the “Truth About Arrows” which included some new construction rules including an Octagon rule. I hadn’t bought these before, but decided to buy all 6 volumes in Prague and had read, based on a pointer from Walker, the Arrows section the night before. That Octagon rule proved quite helpful. Having found the 0 and 1 letter clues by logic, I could then identify that 2C + D = A + B + C + F from the octagon rule. With only 2,3,4,5,6 leftover, I could identify C and D as being 4 and 6 in some order and the grid made F = 2 very easy. The 95 point puzzle probably took 40 points worth of time. My time check for here, with 380 points more solved, was 34 minutes left on the clock. I had 44! The Liar puzzles went a little more at normal not accelerated pace, at least the Liar diagonal slitherlink where I was concerned that I might violate the diagonal cells lying constraint which was not too clear in the written rules. But I got through those and then cleaned up the 85 points of puzzles at the very start and had ~12:30 left on the clock. I turned as quickly as I could through the booklet to confirm that I had solved all of the 17 puzzles and turned in without checking any individual puzzle any more. My visualization of the morning needed a lot of time bonus and I was going to get every point because it could count.

Afterwards, Palmer mentioned maybe getting lucky towards the end and finishing one of the Doubleblocks which probably kept him in 2nd, but Ulrich had not done the Coded Arrows. 12 minutes + 95 points would be 215 gap (I needed to make up 200). I just needed to be clean. We had a team round in here then while waiting for the scores, where we split up the task and solved fairly well as a team but had a non-unique loop in the grid we never fully understood.

For the first time that day I had an anxious wait to see if my crazy fast and focused solving had worked. At lunch, not expecting the results up yet, Ken Anderson told me something about 7 minutes and being in the finals. I rushed upstairs to get the context. I had achieved magical goal 1 in my visualized path to a title. And the results were indeed super close. While Ken Endo continued to be well above the rest, 2-4 were within 35 points and I had snuck into 3rd. Ulrich had a time advantage and first puzzle choice in the semifinals but he would be the one to have to do extra solving.

Rd 13 –> Final (Rd 13 Rank; Final Rank)
Ken Endo +1030 -> 8640 (1st; 1st)
Palmer Mebane +725 -> 6765 (6th; 2nd)
Thomas Snyder +930 -> 6745 (2nd; 3rd)
Ulrich Voigt +715 -> 6730 (7th; 4th)

Continues here: Playoffs