Welcome back, to the final episode packed with puzzles and drama. For the morning, we had the last two preliminary rounds on the schedule. By this time I knew where I was at: Around 14th place, which is a decent result for me relative to past performance. I managed to decide to be ok with it, particularly with not having come in to the competition in best shape.
Round 13: Twisted (17 puzzles, 90 minutes, 810 points)
As a wake-up round, we were served a round made up of “twisted” variants, which translates to puzzles where the clues were replaced by question marks, coded by letters, lying or “wrong” (off by one, a.k.a. knapp daneben; I still hope that we can establish knapp daneben as the standard international term for this).
My approach to this round was once again mostly front to back, skipping some low-pointers for the last minutes, and the puzzles I didn’t like. First was an ABC-box (20), not entirely easy but a smooth enough start with all the croco-puzzle training.
Next a somewhat twisted Coral with Questionmarks (15): clues were in order, and all 2s were given. I was surprised right now about the “all 2s” rule thus am unsure whether I remembered it on the contest. It also turns out to be entirely redundant for the competition puzzle, so maybe I got lucky. I slightly question the choice of including the rule if it’s not relevant (though it does affect the solve a bit), but well done to the authors for ensuring that the example does require it. To end the question mark block, a straightforward Tapa with Questionmarks (40). That one felt like cheap points to me.
I skipped the following Tents With No Hints (10), then sunk quite a bit of time into the Coded Doubleblock (60). That one had a pretty 2018/MMXVIII theme, and required reasonably limited casework. My mistake was to forget that 0 was a valid clue/replacement. (Another question I had explicitly clarified in the forum Q&A beforehand…) I gave up for a bit after proving it broken, but thankfully thought of this later and got it out when revisiting.
Next a Coded Skyscrapers (60), which was pretty straightforward if you noticed a certain regularity about skyscraper grids. I’ve seen/constructed enough of these to find that approach quickly. After that it still needed a bit of limited case-work, but still, a pretty puzzle.
The following Coded Coral (85) had a fair non-trivial break-in via the coding of numbers 1/2 and 3/4. I made one case-distinction to resolve 3/4, but that seemed to just point me at the correct logical deduction.
After this a 4×4 Coded Arrows (95) which I was happy to skip. I wonder if this would have been similarly manageable as the Skyscrapers if I was as familiar with the base type? I solved it after the fact with Martin’s help (or rather, he solved, I made mistakes), but it seems a very tough case-worky solve even after noticing certain helpful restrictions.
The pretty Coded Laser (80) seemed limited enough (if you remembered that 0 was not allowed as a substitution). It took me a bit of erasing, but I got it out well enough. Presumably the difference with the Arrows is mostly that I’m more comfortable with the base type, particularly with quickly sketching and verifying a solution. Or maybe it was just that much easier?
That was it for the coded batch. Next up, liars. To start, a Liar Diagonal Slitherlink (55). With the full grid, I don’t like this type quite as much as plain Liar Slitherlink, but still they’re loop puzzles, and the diagonals add a nice twist to the standard Slitherlink deductions. I kept finding the breaking restrictions quickly enough, though more than once I broke the puzzle locally due to missing an obscure way to resolve things differently with diagonals. Still, a good solve of a good puzzle.
The Liar Masyu (55) went similarly well. This one was a “knapp daneben” variant; I’m comfortable enough with the concept from countless instances on the Rätselportal, e.g. this Slitherlink. I had one or two contradiction-scares, but those were imagined.
I left the Pinochio Fillomino (20) for later, solved the Polygraph (35) quickly (with inverted clues relative to the “standard”), then went on to sink time into the next Doppelblock: Wrong Doubleblock (85), a plain 6×6 knapp daneben. The solve here went a bit strangely; Based on my Doppelblock experience in this championship, I quickly started guessing, and got pretty far pretty well after placing a block in the top left corner. There was enough logic in there that I felt I was on the right path, but eventually I reached a contradiction. After erasing and not getting anywhere with other tries, I tried reconstructing my first almost solution, and found a way to tweak it to resolve the contradiction. Phew.
The next two puzzles, a Wrong Snake (25) and Wrong Shikaku (30) were both easy, and the final Wrong Products (40) was quite a bit of fun to solve, with a mix of certain placements and placements that felt right.
In the remaining few minutes, I rushed through the left-over fillomino and tents, thus finishing with only the arrows missing. I came out clean (though the polygraph had one square left I’m not sure I really resolved consciously), with another good result at 100% of 10th best.
Favorite puzzles: Coded Coral, Wrong Products
Score: 715 of 715/810/1030; no errors
Team Round 3: Ariadne’s Thread (90 minutes)
This was followed by a large team round, a combination of loop puzzles. It consisted of 9 individual puzzles of various loop types that were to be assembled into a 3×3 square with one long loop connecting them all, with some restrictions on how often the borders between the loops could be crossed.
The type makes for some really fun twists in solving the base puzzles, since certain standard deductions grow less certain when you’re allowed to leave the puzzle through an edge, while noticing where such exits are forced is helpful/necessary for assembling the pieces.
We struggled a bit getting into this one; particularly, the Country Road had a non-obvious logical step that made it resolve without external input that I missed, but Ulrich found it when having a look later. With that, I believe we placed the first two or three puzzles, and found at least a rough idea of the global course of the loop. I’m sure our preparation helped here somewhat: Philipp had prepared a very tough practice puzzle based on some deep global arguments related to the loop orientation (Password Path and Bahnhöfe had to be oriented consistently, but crossings…). That didn’t apply as such, but ensured we didn’t miss any deductions on that level.
At this point we got stuck again for a bit looking for certain deductions, eventually opting to try some things that “looked good”: The Tents (Ulrich) solved nicely in one corner and fit in with the rest, and similarly the Bosnian Snake (me) only seemed to fit well in another corner and resolve uniquely given the one entry we had. Unfortunately I was wrong with the “uniquely” here: There was a second solution for a second exit. At that point, we had most puzzles resolved and placed correctly, except that the Bosnian Snake contradicted adjacency to the Yajilin due to the wrong exit. Given the combined uncertainty of how we’d arrived at the current state with the seeming certainty of that wrong exit, we ended up backtracking a lot of things, re-evaluating a bunch of the placements and individual solutions that were already correct, and seeing what could fit elsewhere. I remember trying once more to place the Bosnian Snake in most of the other unsure spots, and even solving the whole Country Road anew. We finally broke through by a combination of hard casework on the Yajilin by Ulrich, and Sebastian eventually reworking the Bosnian Snake for that exit. Thanks, guys!
So with about 5 minutes on the clock, we handed in. At this point, every puzzle must have had two people double-checking it, I’d spent the last 5 minutes checking every puzzle again myself. Unfortunately, while we apparently caught a few local letter twists in the password path earlier, one such twist still escaped us (I didn’t even realize this was a likely error), so we didn’t come out clean after all in the end.
Ultimately, we lost a serious amount of ground here, losing between 1000 and 2000 points to the teams around us. Thus after being head to head with team USA for second place before, we dropped to a clear third, with the top four evenly spaced. Well done on USA in particular for clean and fast solve, finishing first with 36 minutes on the clock.
After lunch, it was time for the individual playoffs. There was a clear cut after 10th place, so not a lot of worry about protests. I only have an outside perspective on these; perhaps some of the participants have more to share.
The playoff structure was quite similar to recent years; I think the three-stage playoff has proved itself well. In contrast to previous years, the organizers set up a system with video feeds and excellent live commentary by Jan Novotný. There’s certainly room for improvement as the commentator didn’t seem to have much control over the choice of camera, nor could he see what was going on outside view. I felt it was a huge improvement over recent years even for the live audience, despite the downside of being in a different room from the players. But more importantly, this is a great way to increase visibility outside the core community. I highly recommend reviewing the live commentary on YouTube, particularly to relive the drama in the second and final stages. Spoilers below!
The organizers adapted the time bonus calculations relative to last year, to be more realistic in view of Ken Endo’s domination of the preliminary rounds. This meant that there were rather low time differences between all the other contestants. The first round had Nikola Živanović, Philipp Weiß, Bram de Laat and Zoltán Horváth fighting for a place in the second round, with a difference of just 2 minutes between first and last through three puzzles. All players got through the initial easy Overlapping Squares (25) quickly enough. This puzzle was Nikola’s choice, and it makes sense for him to choose an easy puzzle to make the most of his time bonus. The came a Galaxies (45), Philipp’s choice, and most players but Philipp seemed to struggle with this one. It seemed that Philipp caught up with Nikola here, and both handed in a correct solution around the same time, but they got it back as incorrect. The playoff was interrupted soon after. It’s still unclear to me whether it was just due to a wrong correctors’ solution, or whether the puzzle had multiple solutions.
Apparently, the players were offered the choice to jointly move on to the second round, but opted to continue with a substitute puzzle after the results of the first puzzle instead. This seems like a hard call, particularly for Philipp who had made up time on his puzzle of choice, but apparently he was ok with it.
Finally, Nikola prevailed to grab the fourth spot in the second round, next to Ulrich Voigt, Kota Morinishi and Walker Anderson. Again, the field was separated by 2 minutes.
Here, the first two puzzles (Singleblock and Futoshiki) were rather uneventful. The third puzzle, a large Loop Around Pentominoes shook things up however. The audience downstairs was suffering seeing Ulrich’s touching X and L pentominoes, but most players had some trouble with this one. I remember a lot of erasing by Kota. Nikola did eventually get it out with a good lead, but went on to struggle with his choice of Pinocchio Fillomino. I believe Ulrich was still fighting the loop when Nikola submitted the first of several wrong solutions. Finally, Ulrich submitted the loop and blazed through the fillomino to win the round. (If someone would take the time to extract the players’ times on the individual puzzles from the video, that would be much appreciated.)
The final had Ken Endo starting with a solid and well deserved 7 minute time advantage over Palmer Meban, Thomas Snyder and Ulrich Voigt. One important note here is that there were only 9 puzzles in the pool to be chosen, presumably because the organizers found another buggy puzzle while double checking after the quarterfinal incident. This meant that instead of giving Ken a second choice between the two remaining puzzles, Ulrich got to choose two out of the last three puzzles.
Starting on his choice of Nurikabe, it was quite clear that Ken was struggling with his nerves. The puzzle was not particularly different and should play to his strengths (well, any puzzle would), but he seemed to lose roughly half of his bonus. Over the next puzzles, the field and Thomas Snyder in particular got closer, and finally all players were working on the giant final Wrong Products (numbers 1-20). Ken’s lead over Thomas was down to 1:30, with the other two another 3:30 minutes down. I remember thinking I was glad to not be sitting up there, because I really would not enjoy solving that kind of puzzle. It was tense for a while, but Ken ended up breaking the puzzle and Thomas finished first with 6 minutes to spare. Ulrich placed second, having caught up all but one minute on Thomas. The other two did not manage to finish the puzzle, thus placing third (Ken) and fourth (Palmer) according to their starting positions.
Congratulations to Thomas for finally succeeding at a WPC, a well deserved victory. To Ulrich, for an amazing semi-final and the next podium in how many years? And of course to Ken, who is clearly the best non-playoff puzzler by a huge margin.
Due to the roughly equal steps in points, teams started with about 2, 4 and 6 minutes delay after Japan. The format was a welcome take on the “weakest link” concept – pure weakest link can be very cruel, putting a lot more focus on the individual than the team, with a tendency to place the blame for team failure on a single player. Senec 2016 comes to mind. Instead, players had the choice to help another player with their individual puzzles once they were done with their own puzzles. This innovation may have been informed by the very unequal distribution in difficulty of the individual puzzles (Galaxies and Coral were valued around 50 points, while Japanese Sums and Easy as ABC were valued around 100 points), but I think it’s well worth using even with more balanced individual puzzles.
One nice touch is the effect that due to the helping player having to wait out the correction window of one minute together with the designated player, there’s a real time cost involved with helping. Then, there’s a somewhat interesting decision to be made in terms of whether you want to optimize for being complete at the team table early on, or whether it’s maybe better to have one player at the team table early to start on the solve, since typically teams don’t parallelize well to four players. However, that decision relies a lot on the concrete team puzzle; had we known that there was good progress to be made even without all the transferred components from the individual parts, we might have made a different choice.
Regarding the puzzles of the round: Each player was given two grids of their puzzle type, and three grids of Xs which had to be matched to the two puzzles. The left-over grid was then taken to the team table, where it had to be matched to one quadrant of a battleclouds puzzle.
As such, Philipp and I got through our Coral and Galaxies smoothly and went on to help Sebastian on the Easy as ABC and Ulrich on the Japanese Sums, respectively. Watch me here working on the Galaxies. Ulrich had both grids half solved at this point, so I got one, and soon started guessing myself to most of a solution in two levels, since there didn’t seem to be a lot of progress to made with strict logic. You can watch me not yet making much progress in the video. Ulrich finished his in the mean-time, and we wrapped mine up. Unfortunately, we’d missed one doubled digit in a column (presumable because it was still in guessy candidate notation at that point), which lost us one minute (times 2), but this was fixed soon enough after. The other pair handed in very close to our first submission, but they stayed at their table a bit longer: Due to some miscommunication it appears that they did not transfer the forced blanks to one of the puzzles. I’d guess they submitted four times over all, but I’m not absolutely sure.
At the team table, Ulrich and I found a few concrete deductions, then went with a horizontal distribution of the large ships that seemed to work very well. The solve is a bit of a blur, I remember doing things here and there, particularly with placing the X grids, but it really was very much a back and forth, particularly once Philipp and Sebastian had joined. I also had no idea how the other teams were doing at this point. I think the US team had two players at the table already when Ulrich and I came there, but that’s about it. Anyway, the puzzle resolved nicely with a few hiccups, we spent a small amount of time checking then handed in, aware that the other teams were still working. Soon after (20 seconds?), team USA was done, so there was a tense wait knowing we’d probably not have the opportunity to fix any sloppy mistake we left in the grid. Thankfully, we were marked clean. World Champion!
USA was also clean, while the other two were still clearly in the middle of it. Hungary still had players at the individual table even at this point. We waited outside, and at some point the Japanese team noticed the cause of their endless contradictions: They had apparently taken a wrong sheet from one of the individual tables. Unfortunately, by the time they noticed (and recovered the missing sheet which had mysteriously disappeared – I’m very happy with how the organizing team dealt with that bit of unfair play and will leave it at that), Hungary was close enough to the finish that they didn’t manage to catch up. The Hungarian team had done a great job preparing for the final players to join their table and wrapped it up in no time, well worth watching the video to see some of this work.
I’m very happy with and proud of our result, coming back after the poor last team round to overtake USA. Another unfortunate playoff for Japan, particularly since this wasn’t just a case of us making less mistakes:
Clearly it was a mistake to take the wrong sheet to the final table. Personally I worried about this and was not safe from making the same error since I didn’t mark my sheets well. But, there’s also an aspect of unclear rules and inconsistent marking behavior here: There was apparently no clear agreed upon line on the invigilators’ role in the process. I personally only handed in the puzzle grids and dealt with the X grids myself. But what if I’d handed in all five of the sheets, would only the left-over grid be returned? All three? Apparently, some invigilators handled this differently from others.
I find the argument quite convincing that the rules should prevent errors like this: Why do we even have invigilators for the individual puzzles, if not to ensure that players make it to the team table with the correct ingredient for solving the final puzzle. My recommendation for future contests would thus be to clearly make this the responsibility of the invigilator, but at least to clearly specify how this will work.
That’s it for now. Some kind of stats/summary post probably to come, but I hope you’ve enjoyed the write-up so far!