WPC 2015 retrospective: day 2, mostly

Below, continuing the story of day 1, I’ve collected some thoughts on the rest of the WPC in Bulgaria. And as before, failed to cut down the length.

Round 8: Instructionless

Day two started with an instructionless round. I was a bit scared of this, expecting very obscure rule sets. But the puzzles turned out to be quite fair in my opinion. I found the correct instructions for all but a latin square puzzle, where the clues indicated a difference of diagonal two-digit numbers. I particularly enjoyed the loop puzzle type where clues indicated the distance between two visits. There’ll be one on the croco WPC series, so I had to think of a name: Umweg (“detour”).

I did run into quite a bit of trouble with an object placement puzzle where you had to place pins of different lengths, with equal lengths not touching, not even diagonally. Unfortunately, the NED-part of the rules was not clear from the example though I did assume it. But then I repeatedly arrived at a contradiction right at the start of the first puzzle of this type, so I dropped that rule, and put quite a bit of time into the first puzzle that now turned out non-unique. That’s probably about 60 points lost. In the end, I missed the pin and latin square puzzles, as well as the hardest arrow puzzle. (260/305)

Round 9: Different Areas

Next up was round where all puzzles had an extra “different areas” condition. For some number filling puzzles (Sudoku, Ripple Effect, Magic Summer, Kakuro) this was just an extra-region type constraint. This was probably the individual round where we put most time into preparation (we had a Loop, a Kakuro, two Magic Summer, a Tapa, a Snake and a Star Battle). On the other hand, the most useful thing here is to know the individual types well. This one felt quite good: The Loop was just a bit of quick trial and error, I skipped Sudoku (could only lose time here) and Ripple Effect, got through the Star Battle and Tapa well enough but not with precise logic. The Snake probably took a bit too long to get right, and in the end I was quite lucky to correct the Magic Summer in the last seconds after I’d introduced a small error mid-way. Rather optimal over-all, so I’m surprised this was one of my comparatively bad rounds. The beer, maybe? (350/425)

Round 10: Something Different

Last before lunchtime (over lunchtime, rather) was a long mixed bag of not-so-common puzzles. There was enough material here that I was able to skip the ones I didn’t feel comfortable with. Particularly the Scissors and Mars and Venus types (two unfamiliar object placement puzzles). In my first run through (starting with the Pentargets if I remember correctly) I did the all the fun stuff. There were a bunch of tetromino types here, those might have been intended for a tetromino-themed round. Nothing too memorable here. Numbered Rooms is a strangely fun type. Towards the end, the first XO sets too long, so I skipped the other one. Unfortunately I tried the Cards puzzles too late, though I luckily managed to get the smaller one out within the last minute or too. I’m not sure whether others were affected similarly, but this is a great example of why you shouldn’t put really hard examples in the instruction booklet: The Cards puzzle in the instructions was so hard I didn’t even consider the type for a while. The competition puzzles were actually both quite fair and straightforward, possibly even overvalued. Like the previous round, no errors and a puzzle correctly solved in the last seconds. This one was actually good in comparison. (710/690)

Round 11: Sprint

After lunch, a sprint round with a bunch of relatively easy puzzles. I heard some criticism that the puzzles weren’t sprint-level easy, but I think they were quite appropriately easy, certainly compared to puzzles in other rounds. Too many puzzles to finish for sure, so if a sprint involves finishing, it did miss the target there: The best score was 250 out of a possible 365 points. On the other hand I prefer this over the “sprint to the brick wall” of London 2014, where the time to sprint through 90% of the puzzles didn’t matter because most everyone got stuck on the far harder No Four in a Row puzzles at the end of the round.

In terms of puzzle selection, this one went decently well. I skipped the Different Neighbors which I’m sure would have taken me too long. Broke the Product Skyscrapers but wasted no extra time here, wasted a bit of time on the first but harder Tetroscope (this was another type with a terribly hard example). The Two Circles involved a bit of guessing but fell fast enough, the Nurikabe was not at all trivial so not sure it was good value for time. The Yajilin also required a bit of trial and error to resolve, and then I put way too much time into the Every Second Turn, which I broke and erased two or three times, no idea what went wrong there. It seemed a bit large for the points anyway, and solving it three times for 0 points is a bit suboptimal. I missed the Encrypted Skyscrapers during the competition, but that has quite an elegant solving path! (185/225).

Round 12: Tough Puzzles

Then came the last of the individual rounds, intimidatingly labelled “Tough Puzzles”. After the previous day’s Hybrid round, I expected these to be a series of impenetrable trial-and-error puzzles. But, over all, not at all. It did start quite badly with the Yajilin: That one has a nice solving path about half-way through, then required a two-level case distinction (for me, I did hear others managed to cut one level, but I haven’t so far). I got stuck there repeatedly, going off to solve a couple of the other puzzles then giving it another try. Thankfully, I did manage to finish it in the end, though I’m pretty sure I could have scored better points if I hadn’t tried it. The other puzzles were quite approachable, I missed the Domino Castle, Coriolis and Hula Hoop. And those were a joy to solve afterwards. (615/580)

Round 13: Domino Yin-Yang (Team)

The final round of the day was a team round of Domino Yin-Yang puzzles. Due to short-term difficulty adjustments, the sheets had to be reprinted, so the round was delayed bit by bit until after dinner. (The delays are one thing that could have been handled better: Delays are fine, but the frequent indefinite delays meant there was barely any down-time, which made the whole thing more stressful than necessary.)

I really liked the concept of this round: The puzzle type involved placing two sets of dominos in a grid satisfying Yin-Yang rules. There were four “small” puzzles to be solved individually (though helping each other was allowed); these were 10×11, and left over one domino from two full sets of 0-6 dominos. The individual part went quite smoothly from my perspective: Our Yin-Yang expert Philipp had made some useful deductions regarding the border of a Domino Yin-Yang, which turned out useful in getting started. I’m not sure how critical it was to figure this out. I did have a bit of trouble not losing track of things on the big grid, somehow my notation could have been better. I made two smaller mistakes while solving, but fixed one myself and the other with Florian’s help. By this time, he had finished his puzzle, and Philipp had been working on the final puzzle for a while.

That’s when the fun started: Using the border rule, Philipp found a rather convincing contradiction in the final puzzle. There was a bit of a worry that something had actually gone wrong in reprinting the puzzles; that sort of thinking tends to creep in after a series of printing errors. But we started double checking the individual puzzles, and I think it was Ulrich who asked: Why is there gray domino lying on this red clue? It turns out that Florian’s brand of colour blindness allowed confusing the red and gray clues. This had resulted in an extra red domino for the final puzzle, which made it unsolvable. Here, knowing about the border really paid off, it’s entirely unclear the contradiction would have been that convincing and early in the solve without knowing that.

So Ulrich spent the next 5 minutes or so carefully inverting that individual puzzle, while the rest of us worked on the final puzzle. We got through that one quite well, Ulrich joining us part-way through. We were probably not making full use of all four solvers towards the end of this, but I’d say it was a pretty good solve. We ended up finishing first with 10 minutes left, 2 minutes before the German B-team. No other team finished the final puzzle (at least not without error), with four teams finishing the four individual ones (both Japanese teams, USA and UK).

Round 14: Encrypted (Team)

The last round before the play-offs was a team encrypted round, with 8 individual puzzles linked by shared encrypted clues. By the time we started, individual results (and those for the first team round) were out, so we had a pretty good idea of where we stood: A solid lead over Japan, which would allow us a minor mistake on one of the puzzles, possibly even a proper broken puzzle. That’s not really that reassuring, considering the most likely way to fail at a round like this is to make a faulty deduction early on, which would spread all through the puzzles.

Over all, the round ended up as a careful and steady solve. We had roughly split up the puzzles equally, with each solver taking two. Mine were the Minesweeper and Slither Link, and I started out by crossing off lots of 4-9 ranges, then the Minesweeper clues, and also borrowed the Arrows for some simple clue eliminations. I remember some initial surprise regarding the Kakuro, which had crypted sum clues instead of crypted solution digits, contrary to the example.

Then it was a steady back and forth. I made a bit of progress on my puzzles, though I missed the fact that the total number of mines was given for the Minesweeper entirely, which would apparently have allowed earlier progress there. There were a couple of uniqueness deductions in the Minesweeper which we had decided to skip and only use for verification.

As I remember, one of the break-through points was Philipp noticing that the length of the Snake allowed making a sum-of-clues deduction for the columns. We double-checked that one carefully — it seemed like an estimate you definitely didn’t want to get wrong. I also remember brief trouble on the other end of the table, related to some mistake on the Kakuro, but that was resolved quickly enough.

Soon enough, we had a couple of puzzles where all clues had been resolved. (For me first the Slither Link, then the Minesweeper; I think Philipp had put the Tents to the side quite early.) I borrowed the Arrows which was still pretty empty at this point and found a few missing clue transfers, then set to wrapping up some of the puzzles. I remember having to make a case distinction for the Different Neighbors, so we checked that one particularly carefully. We ended with a thorough checking round, verifying correct clue transfers and solutions for each puzzle, and handed in, two minutes after Germany B and with 21 minutes left. Serbia was done 2 minutes after that, with USA and Hungary back another two minutes. A total of 13 teams finished. It seems we were not that fast relatively, compared to the Domino round, say. I do wonder how much missing that total mines argument threw us back, not that it mattered in the end.

Individual Play-offs

That was it for me in terms of competition, but we still had the individual play-offs to watch. This must have been summarized better elsewhere, but I’ll try to provide an overview. Details are mostly based on the “live ticker” at the German LM forum.

The mode was the same as in London 2014, one of the innovations there that I hope will stick: The first 10 play off in three stages, starting with a staggered solve of three puzzles by players 7-10, then four puzzles by players 4-6 and the winner of the first stage (whose disadvantage is that of player 7), and finally five puzzles to be solved by player 1-3 and the winner of the second stage. That means player 10 has a realistic chance of medalling, but the first three have a rather safe advantage due to their performance in the main rounds. Besides a time advantage, players also get to select the puzzle types that are used, with top players getting first pick.

Within a round, you have to solve the puzzles in order, and after handing in you have to wait for one minute, after which you get to start the next puzzle provided the previous one was correct.

The first round had Zoltán Horváth (7), Yanzhe Qiu (8, 0:41), Nikola Živanović (9, 0:59) and Florian Kirch (10, 1:30), competing to solve a Polygraph, a Kropki and a Magic Summer puzzle. Here, Florian took the lead on the first puzzle, and went on defend that to advance to the next round.

The second round had Kota Morinishi (4), Bram De Laat (5, 0:02), Ko Okamoto (6, 0:04) and Florian Kirch (7, 0:08) solving Rollercoaster, Hexaislands, Pent it Black and Fuzuli. This round was dominated by the Hexaislands; that’s a puzzle type which typically doesn’t allow for logical progress, you have to mostly just search a large solution space efficiently. It’s typically not feasible to reliably work your way through, in the sense that you need to make strong assumptions to come to convincing contradictions. So, mostly just try things and tweak things until you have the correct solution. Or ideally you just draw it right away. Anyhow, the solving time for this type, and this specific puzzle in particular, has a very high variance. You might argue that top solvers should have strategies to work through this reliably, but at least at the difficulty level here, it can take an almost arbitrary amount of time to solve. Certainly the solving time is random enough that the one puzzle decided this round, with Bram advancing to the final round. It’s a good thing the puzzle wasn’t chosen for the last round, though I think that would have been a reasonable strategy for someone with a large time disadvantage.

The final round was between Ulrich Voigt (1), Ken Endo (2, 2:51), Palmer Mebane (3, 5:35) and Bram De Laat (4, 8:22), the puzzles were 2D Magic, Chaos, Different Neighbors, Kakuro and Coral Maximum. I’m not entirely sure of the choices, but the ones I remember are Ulrich with the Kakuro, Palmer with the 2D Magic and Ken with the Coral.

I should remark here that I got a pretty good idea of how the play-offs unfolded by sitting in the first row, but really, this sort of thing could be transported to the audience in much better ways. As it was, the main thing we saw via projector was what puzzle who was working on. It wasn’t even clear who had made up time on another player without tracking the distances yourself. And of course no web-cams or commentary. Certainly it took being very much a puzzle expert to get anything at all out of watching the play-offs.

But on to the final. There were big gaps which meant few upsets. As far as I remember, the main events were:

  • Palmer lost time on the 2D Magic. I was really confused when I heard he chose the type, since that really doesn’t seem to play to his strengths. I wonder if he thought he had an advantage here having discovered those extra regions?
  • Ken overtook Ulrich on the Kakuro. Another unlucky choice, though this one made sense to me.
  • Ken defended his lead on the Coral, which both players seemed to take a non-rigorous approach to.
  • Uvo finished the Coral for a safe second place, and Palmer defended third.

Here’s a video of the end of the competition.  (I do hope it’s viewable without a Facebook account. Unfortunately, far too much of the online presence of the WPC takes place inside Facebook’s walled garden.)

I’ve solved the Kakuro afterwards and struggled with it, though it does have an entirely fair solving path I didn’t find. I haven’t seen the Coral or the other puzzles yet, unfortunately.


That’s it.

Results

From a competitive point of view, I’m very happy to have brought a team gold medal back. Thank you Florian, Philipp and Ulrich for making this possible! (And everyone else who supported us on the way there.) I was quite worried after that first day of having ruined that for everyone, but thankfully we recovered. After a solid second day, I even managed to place third among the German A team, for a total sixth place among the Germans. (Yes, the B-team did really well.)

Individually, it’s a bit less positive. The good part is that despite doing really badly on two rounds, I still placed better than in previous years (17th including B-team members, while I placed 18th both years before), so there’s clearly some improvement. I was hoping to try for the top 10, and while it would have been a stretch this year even if I’d done less terribly on rounds 5 and 8 (355 points down on place 10), that seems within reach. Particularly if I go in a bit more relaxed, having my first successful A-team WPC behind me. That will require qualifying again, though.

Puzzles

Having looked at the previous Bulgarian WPC and some Bulgarian national championships in preparation, I must say that I was very happy with the puzzle selection in general. Having gone through the whole set in the past couple of weeks, I don’t think there’s a single puzzle in there which I find completely intractable, and there were a couple of those in the older competitions. (Can anyone tell me how to solve Vertical Sums, say from this forsmarts book?)

Among the individual rounds, my favourite was clearly the instructionless round, which I thought was really well done. The Möbius round was also great fun. Then the different areas was a nice idea, but that mostly stood out from the rest because it did have a real theme. The poker round was well done, it’s just not a type that thrills me. Other than that, the individual rounds were a bit too much of a huge mixed bag of more and less interesting puzzle types. Somehow, I thought that worked better my previous two WPCs (Beijing and London, which are really hard to measure up to).

Regarding the team rounds, I thought the colour round was a great idea with not so great execution (compare the review of day 1), the domino yin-yang was excellent, and the crypto round was a solid execution of a proven team puzzle type that did a great job of really testing team work.

Organization

I won’t say a lot here. There were some problems as mentioned above, and I feel things could have been handled better, but particularly communicated better. I want to thank everyone involved in making things work out in the end, can’t say I thought it would go this smoothly: One day before the competition start we had no instruction booklet, it could well have turned out that there were no puzzles either.

Advertisements

One thought on “WPC 2015 retrospective: day 2, mostly

  1. Pingback: WPC 2015 retrospective: day 1 mostly | Maybe Puzzles

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s