**Rules** Fill the grid with digits 1-6, so that each row/column contains each digit exactly once. Furthermore, draw a single loop along the grid lines that doesn’t touch or cross itself.

The loop divides each row/column into groups of digits. Each such group of digits corresponds to a number; it’s the product of these digits if the group is inside the loop, and the sum otherwise. Clues outside the grid describe the groups in that row/column in the correct order; an asterisk (*) stands for an arbitrary amount of groups, including no group at all.

**Rules** Solve as a regular Geradeweg. In addition, any two straight segments which meet at a point must have different lengths.

]]>

**Rules** Solve as a regular Country Road. Additionally, any two straight segments that meet at a point must have different lengths.

**Rules** Solve as a regular Masyu puzzle. In addition, any two straight line segments that meet at a corner must have different lengths.

**Rules** Solve as a regular slitherlink. In addition, any two connected straight line segments must have different lengths.

I attempted to record myself solving again, video here. Unfortunately, I once again managed to hide half the puzzles, particularly my one mistake. Here’s my course through the contest:

I started with the clouds because they came out of the printer soon and are a type that I’m reasonable comfortable with. I made a bit of a mess of the first one, which is typical for me at the start of a competition for me. By the second one I was warmed up (or got a bit lucky). I pretty much just drew the solution in there.

Next, I didn’t want to do the arithmetic squares, and had decided prior to the contest that I didn’t want to risk the 151 point arithmetical city. The doppelblock choice was rather arbitrary, it was what came up on top after flipping forward a few pages.

(Usually as a top 10 puzzler I’d expect I should have to try a big pointer like that, but I feel like the GP has consistently missed the goal having a couple of finishers this year. Judging by the huge total point value, I was hopeful to be able to fill the 90 minutes with more appealing puzzles.)

With the exception of the overvalued second puzzle, these went quite poorly for me. (Watching my recording of the second puzzle, I even wasted a lot of time there missing an easy deduction in the top rows.) I don’t think anything went wrong in particular with the first puzzle.

On the third, with a rather inelegant rule set (intended to arrange the possible clue values cyclically), I could have been better prepared; it took me a while to realise that most large/small values still allowed deductions regardless of whether they were lying or not. I couldn’t break in without some casework, but eventually I seemed to be on the right track. There was a scary moment close to the end where I’d made a bad deduction and I was close to having to erase the whole puzzle; thankfully I was able to take a step back and correct my error.

It seems the GP test solvers are not very familiar with the type. Of those 3:34, I spent 41 seconds carefully entering the solution code – those NZs are quite confusing. It was a pleasant clean solve, though watching the recording is slightly maddening, what with me ignoring obvious deductions all over the place.

I really enjoyed this one. I’m a bit familiar with the type, but still charged right in with wrong deductions. Once I saw how the puzzle worked it came out very nicely.

After the Yagit I briefly looked at the Arithmetical City to see if it seemed approachable. However I was mostly irritated that the intruction booklet featured unused extra rules, and paged back to the type I was most looking forward to. The “unequal lengths” rule is a wonderfully simple variant rule that changes the type quite significantly.

My solves here were not particularly clean, but fast enough, particularly on the second puzzle.

Next I went for the skyscrapers. It’s not particularly strong for me as a latin square type, but as a German I’m well practiced… Neither solve was particularly clean, each time ending up with some quarter of the grid empty without obvious logical progress. I got the first one out well enough, but on the second I had already made a bad deduction before guessing, and ended up turning in an (obviously) broken solution.

Looking at the solve in retrospect, I wonder if I should go to more effort to catch errors like this. (Placing two 1s directly above each other seems like something that could be noticed without wasting too much time, but who knows?)

For the first of the snakes, I found the right place to start, but it took me quite a while to figure out that there were options for the bottom of the grid that looked different than what I started with. The second one was a pretty satisfying mix of experimentation and logic.

At this point I had the two Four Winds left, and still a decent amount of time on the clock. Neither of the two felt like a particularly fast solve; particularly on the second I felt that I was scanning the grid for spots to make progress quite a bit, which is usually not something I’m fast at. Judging by the points per minute, it seems I did ok though.

By the time I was done, I was down to about a minute. I gave the first puzzle a guessy try that didn’t work out. It might have been a better idea to double check, but chances would have been slim to fix anything in the time left.

Overall, a pretty good result for me, with 6th place after last month’s 5th. After a lousy first GP and missing the second GP as an author, it looks like I might end up in the top 10 again after all if I keep up this level.

Congratulations to the winner Hideaki Jo, and the whole Japanese team that dominated the round. You can see Yuki Kawabe’s writeup here, who solved the same puzzles 5 minutes faster and with one less mistake.

]]>The latest addition is the Nikoli puzzle type Scrin (スクリン); here’s a small sample puzzle:

https://puzz.link/p?scrin/6/6/g2h.zl.h4g

**Rules** Draw some rectangles, such that every clue is inside a rectangle (of the given size if specified). Different rectangles can only touch by corners, and all rectangles must form a single non-branching loop via these corner-connections. Rectangles without clues are possible.

- angleloop: Introduce new puzzle: Angle Loop
- doubleback: Introduce new puzzle: Double Back
- heyawacky: Introduce new puzzle: Heyawacky
- meander: Introduce new puzzle: Meandering Numbers
- nurimisaki: Introduce new puzzle: Nurimisaki
- satogaeri: Introduce new puzzle: Satogaeri (due to Lennard Sprong)
- scrin: Introduce new puzzle: Scrin

- Export: Support https URLs
- Language: Rework some English language messages
- bag: Rework as shading puzzle
- fillomino: Automatically draw borders between different numbers
- yajilin: Allow clues to be marked complete
- yajilin: Autocompletion for clues

- LineManager: Fix connection logic and colouring for pipelink, ringring etc.
- nurimaze: Fix broken cleared clue cells afer random key input
- tentaisho: Fix false error when placing star after drawing border in edit mode
- fivecells: Fix encoding/decoding of missing cells

**Rules** Fill the empty cells of the grid by drawing rectangles that consist of horizontal and vertical lines between cell centers. The sides of different rectangles may intersect but not overlap. Rectangles can’t touch otherwise; in particular, they can’t touch by a corner.