Solver App for Android
Strategies for Popular Number Puzzles
Page:2024 2023 2022 2021 2020 2019 2018 2017
2015 2014 2013 :1 2 3 4 5 6 7 :2012 2011 2010 2009 2008 2007 2006

Feedback and Questions

I've received a lot of interesting comments and questions from Sudoku fans over the last few years and this page is where I try to answer them. I'm also directing Str8ts feedback here. Please feel free to drop me a note on the side of the page. Or you can email me directly at .

Post a Comment or Question here...
Your Name or 'Handle'

Part of the World you're from

Email Address (optional) so I can reply directly if necessary (it will not be displayed here)

Any Sudoku you want to publish here for easy loading into the solver - 81 characters, use '.' or '0' for unknowns.

public - comment/question will be added to feedback column
private - email comment/question directly to Andrew Stuart, don't display here

Comment, question or feedback:

Enter these letters
Please enter the
letters you see:

Many thanks to all the people who have helped improve the solvers and strategies with their feedback!

Friday 20-Dec-2013

... by: Ray, Tulsa, OK

I have enjoyed working Sudokus since the year they came out in Dell and became acquainted with the strategies they posted in the center of their magazine. I have noticed that all of the strategies and commentaries except the ones used at the beginning of a puzzle revolve around the unsolved candidates. My question deals with the opposite side of the coin: Are there advanced strategies that can be used exclusively with the solved candidates and their placement? Are there patterns of solved candidates to look for that will help in solving cells? I have never seen any, but maybe this question could get some heads scratching! (Obviously, I am aware that such strategies would have to allude in some way to the unsolved candidates, but I keep thinking that maybe such strategies could either solve, eliminate, or narrow the list of unsolved candidates rather than having to spend all that time listing them in the puzzle boxes.) Thanks Ray

Andrew Stuart writes:
Hi Ray
There is only one strategy that uses solved cells and it is very interesting. Have a look at:
Avoidable Rectangles

Monday 16-Dec-2013

... by: Amb, New Zealand

I've noticed for your solver that the sequence in the tough and Diabolical strategies tend to use harder to spot sequences before using patterns that humans could more easily use.

For example:
X-Wing (once understood) is much easier for a human to find than simple Colouring.
Unique Rectangles, WXYZ Wing are both easier than X-Cycles and XY-Chain.

Could the strategies that are more like trial and error be moved to a later sequence in the moves list?

ie: For tough
Sword Fish
X-Y-Z Wing
Simple Colouring.

ie For Diabolical
Unique Rectangles
Extended Unique Rectangles
Hidden Unique Rectangles
Others (eg X Cycles, XY-Chain, 3d Medusa etc)

These strategies are ones that can be found by learning what to look for and why, and are better than the other trial and error feeling ones. Quite often when I have stepped through a sudoku and turned off X-Cycles, XY-Chain, Jellyfish, I've ended up finding the WXYZ step was available and this seems to be a more elegant solution to the sudoku.


Andrew Stuart writes:
To a certain extent the choice of the order of the strategies that are checked is subjective. I think everyone will want to order them differently if only slightly so. I have moved them around over the years since I started them in 2005. However the choice of putting a strategy next to a grade adjective is even more subjective (and is really just a guide) as the grade is determined by a lot of factors. I'd like to offer the solver the ability to rearrange the order in the solver but its already quite a complex user interface. I agree that a different order will often lead to a different solve path and it's in my long term plan to work out if some ordering give better scores and are more optimal.

Monday 16-Dec-2013

... by: Ron, US

I have upgraded to Windows 8.1. When I display the Daily Sudoku Puzzle, the "Load in Solver" button does not appear. What setting / feature is controlling the "Load in Solver" button?

Andrew Stuart writes:
Looks like that might be a browser zoom. Try resetting that. The player is in a window which does not get expanded when the text does. That's usually the problem

Friday 13-Dec-2013

... by: mwalimu, Illinois, USA

For the "Import a Sudoku" feature on the jigsaw Sudoku pages, have you considered adding a way of importing any arbitrary board configuration (or 'shape', as you refer to it on the page)? A simple way to do this would be to have the user enter a second 81-character string, which must consist of the digits 1 through 9 nine times each (and nothing else except whitespace), in the format of the box numbers diagram on the bottom of the right column of the Jigsaw Solver page. Entering the 'shape' in this manner should be optional and should probably default to the current method.

It seems to me the trickiest part of programming it would be validating that the shapes are contiguous (although even if they're not I suspect all the solving strategies would still work and the solver wouldn't care, even though it would be very haphazard and hard to follow for a real person to follow). Consider allowing one box to be non-contiguous, which would allow for something like a puzzle where one "box" consists of cells B2, B5, B8, E2, E5, E8, H2, H5, H8 while the other eight boxes have to be contiguous as usual.

Andrew Stuart writes:
Yes. I do want to do that and its in my job queue. I'll probably take something from the Killer Solver as a starting point. And an 81-string color map would be its definition as well. The current Jigsaw shapes are stored internally and reference by number. I'd have to change that but no biggie

Wednesday 4-Dec-2013

... by: Jeyaraj, India

Load Sudoku:

It is good to learn


Wednesday 27-Nov-2013

... by: Chug, Boston, USA

Hi Andrew,

I'm an avid consumer with a buglet to report. I loaded KenKen "#865, November 26" into the solver and solved it. However, the solution doesn't make sense but when I click Take Step it says, "That's it, all done". The board I entered clearly does not solve the KenKen. See

I got all the rows and columns to have 1..6 but they don't satisfy the cage constraints. Is this a bug?


BTW, thanks for your efforts in this work. I have really improved my game on your account. -C

Andrew Stuart writes:
You are quite correct, I wasn't handling the situation where a user entered a valid set of numbers (that conformed to the rows/columns rule) but didn’t match the clues. I've re-worked it now so that it checks values against clues when the solver on the server is called (after the basic strategies) and once more at the end when every cell has been filled in. Your example now comes up with messages.

Refresh the page a couple of times and make sure you can see version number 2.08

Monday 18-Nov-2013

... by: PeteTy, ohio, usa
cant seem to find todays killer....

Andrew Stuart writes:
Fixed! I didn’t quite get the data import right yesterday. Thanks for the alert

Tuesday 12-Nov-2013

... by: Gerrit Altena, The Netherlands

Load Sudoku:

After some cleaning it comes to:
Alternating Inference Chain
AIC on 3 (Grouped Discontinuous Alternating Nice Loop, length 10):
- Contradiction: When 3 is removed from E6 the chain implies it must be 3 - other candidates 1/4/6/7 can be removed

Although the end result is correct, the path is not:
after -5[B4] it should be: +7[B4]-7[B9]+3[B9]-3[B6]+3[E6]

The reason is that with +3[D4|E4] the 3 in B4 is already eliminated.

Great program though!

Andrew Stuart writes:
This is an interesting issue. Does that fact that we are setting +3[D4|E4] imply we cannot set any other 3 in the column to ON as well? If it does then whenever we make a chain we must cross out all X in every row, column and box that is used as we construct the chain. This changes the chain from a chain into a 'net' and it would be arbitrary to stop looking for the effects of such an ON at one step. One would have to have a full tree branching search - Bowmans Bingo works in this fashion.

I have built into the algorithm for chains the rule that the same candidate can't be used twice - mainly to stop endless loops appearing. But no rule exists for this type of check. However, if there was a problem doing this then I would have found an example where it creates a mistake. In the zillions of tests I've done this situation hasn't created a false positive, so I conclude it is harmless. I think the way to look at it is that these candidates are being turned on "potentially" - they have no reality to them since we don't know if they will eventually be the correct solutions to those cells.

Do you think there is always an alternative chain that doesn't violate this rule?

Monday 11-Nov-2013

... by: Amb, New Zealand

You have some pre loaded examples for solving techniques, eg Swordfish? Do you think you could add some daily ones or archives with further examples? I'm after some that require X-Wings or Swordfish to complete, but no forcing-chains.

Andrew Stuart writes:
I am making some new stock for 2014 and I'm going to make enough to find examples such as these. I'd like to get about twenty representative puzzles that really highlight a particular strategy and don’t rely on any hard strategies. Obviously the hard the strategy the more intermediate ones will appear but for these mid range ideas I think I can get a good sample set together. I'll publish the links on the strategy pages.

Sunday 10-Nov-2013

... by: Florent Chayet, France

First of all, a great thanks for the great job!

Just one small remark: either the auto clear is on, and then when a solution is added all the previously manually suppressed candidates are back, either it is off and to suppress the candidates after adding a solutio is quite a job. The dream would be that auto clear on results in suppressing the candidates due to the added solution without changing anything else on the grid.

Thanks again!

Andrew Stuart writes:
I think I know what you mean. But it might be complex to realise. The solver has been designed more to show strategy steps with minor tinkering of the candidates. It sounds like what you are asking for is more in the line of a 'player' in which the user takes full control of the solving process but candidates are cleaned off when a big number is entered. You can simulate this by entering a number in the small board and pressing "take step" once which will update the candidates. But that will reset any personally edited candidates. Since the solver/player doesn’t know what are personally edited candidates it doesn’t respect them if they should be removed. Quite a difficult issue
Page:2024 2023 2022 2021 2020 2019 2018 2017
2015 2014 2013 :1 2 3 4 5 6 7 :2012 2011 2010 2009 2008 2007 2006
Thank-you everyone for all your questions and contributions.