My puzzle will be a No Ten Sudoku. It is similar to Non-Consecutive and XV Sudoku, where no two neighbouring cells add to 10. It's a bit of a troubling type to design by hand as it's hard to judge in advance if you can still create a valid solution from this point on. This is also the reason it's almost impossible to design a Nonconsecutive Sudoku by hand. Usually they are designed with the help of a computer program, to make sure your currently set up Sudoku still has a valid solution. Too bad I didn't have a program to verify a solution for this variant. Although I usually design my puzzle progressively, I did create a valid solution first for this puzzle and then worked out how to turn that solution into a unique and logically solvable puzzle.
A little bit more on the league. We're using this Google Doc to keep track of the solving times. You can request access to it, even if you're not on facebook. You can add your name to the excel sheet and post your own times here for each puzzle. You can still add your times till a week after the end of a week to be considered for the Hall of Fame.
At the end of a puzzle week, Tom will create a Pdf with all puzzles of the week. This Pdf will contain the puzzles, the best solving time and an estimated time for intermediate and novice solvers. The Pdf of Week 1 is available in the Facebook group or as a Google Doc. The Pdf for week 2 is currently only available in the facebook group.
In case you're interested in creating a puzzle for the Daily League, feel free to let us know in the facebook group. More constructors are always welcome.
Rules for Sudoku
In this Sudoku no two adjacent digits may add to 10.
 |
| Click to enlarge |
Haven't had a go at the puzzle yet, and probably won't until I'm feeling a little better. However I was interested by your comments about making non-consecutive and no tens puzzles with computers. I sympathise entirely, and I think I'd probably add anti-knight sudoku to that list. There are probably others that I can't think of right now.
ReplyDeleteI recall with this puzzle (http://tcollyer.blogspot.co.uk/2009/08/friday-puzzles-10_751.html ) - which was, and still is as far as I'm aware, the first no tens puzzle I'd come across - that the construction was exactly as you've outlined here. That is to say I first spent many hours tweaking solution grids with perhaps one or two adjacent pairs of tens before then laying out a design and then seeing whether it led to a valid solution. Eventually I got something rather easy out.
There are variations on this process. You might start with a cool break into a puzzle and then fill the rest of the grid up and then add the rest of the clues symmetrically as required. Or you might lay down some clues, like where things are going, but then tweak some of the other clues to make things easier/harder as required and get a subtly different grid.
I suppose this is in marked contrast to making something like nurikabe, where you can just keep throwing down the clues until your puzzle has a solution. Generally this is not a good approach for sudoku (or at least sudoku with symmetry) because you almost always end up with some surplus clues, making the puzzle much easier than you intended. An interesting exercise having finished a puzzle like this is to play with removing pairs of clues and seeing if you still get a solution. Sometimes you get lucky!
I suppose I'm not so worried about rambling on giving my current state of fever, but returning to designing non consecutive and no tens, I think the former is certainly possibly (albeit very difficult) to design without having a solution first, and you can do some very cool stuff with the variant. It is my feeling that no tens grids are more rare and I don't think I'd be able to make such a puzzle without having the solution grid to hand to refer to all the time.
Perhaps anti knight is a little easier than non consecutive, but I still have trouble without having a solution grid. I admitted as such for the LMI beginners test I put together, where essentially I took the solution grid from a previous LMI anti knight and then added my own arrangement of clues. I suppose what helped here is a general intuition as to which are good place in the grid for clues - there's a sort of geometric intuition for a designer there which isn't quite as obvious (although there still is some!) for non conseuctive or no tens.
I suppose rishi and fred might be better people to ask here, they have certainly made more and better no-conseuctive/anti knight puzzles than i have!
Isosudoku (the hexa kind) is definitely one to add to the list too. I once started making one by hand and just threw in 12 clues in a nice pattern. Couldn't get valid sudoku, as long as I tried. Then went to a program, set up the constraints as an Isosudoku has and the computer was so nice to tell me that after the 11 of the digits I placed, the puzzle already had a unique solution, but the 12th didn't fit in with that solution. But logically there was no way to deduce that.
DeleteFor this one at first I looked very much at the solution to check where I could get patterns of different digits in a nice pattern. I wanted to avoid having the same digit adjacent. That how the four 3 cell v-patterns on the outside came about. I then tried solving a bit, got nowhere and checked where else I had to add digits. A few attempts failed, but this one worked out. I was aiming for 20 or less digits as I figured any more would definitely be redundant. The max 20 clues would make sure that the no 10 part came into play a lot.