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 andrew@str8ts.com.
Post a Comment or Question here...
Thursday 21-Apr-2016
... by: Eyad Khalil, Palestine
Hi Andrew: Can the Daily Puzzles listed on your site be automatically imported into the Solver - without having to enter the 81 digits one at a time? Thanks,
Andrew Stuart writes:
In most cases yes, look for the "Load in Solver" button
Thursday 21-Apr-2016
... by: Bert, Netherlands
problems with '5 daily puzzles' on ipad.
last few days: unable to update kakuro puzzles:
Could't reach the Internet to pick up the puzzles.
No problems with the other puzzles.
Andrew Stuart writes:
Yes, I see the problem. My tracker said I had puzzles up to 2019 but its not true, only certain grades. I will add some more today
Appreciate the alert!
Tuesday 19-Apr-2016
... by: Chris Green, Australia
I'm new here and I'm not sure if this is the place for comments about the Unsolvables.
I just solved Unsolvable #196, using a simple program I wrote (to save laborious work) plus a trial and error approach. I guess such an approach is not considered very valid since Andrew seems to be looking for more logic-based approaches.
Anyway, it took me 22 tries but when I eliminate the erroneous tries, it can be solved in 6 tries. (Of course, that's a case of 20/20 vision in hindsight!) I chose to start with Box 9 because it has the least number of possible candidates at the beginning.
The sequence I chose was: J8 =3, J2 =1, E1 = 1, B1=5, J4 =6 and finally, J3 = 5 for completion.
-1[G2]+4{G2?E2} I don't understand this strong link in the puzzle, could you please explain it?
Andrew Stuart writes:
You puzzle has 25 solutions so it's not a valid puzzle. But, to explain that link ffragment, have a look at AICs with ALSs
Saturday 26-Mar-2016
... by: Uhm, Netherlands
Crazy Extreme (Windoku)
Your solver does not find easy boxline reductions or pointing pairs, instead it finds x-cycles which does slowly the same reduction.
As you documented in: http://www.sudokuwiki.org/Windoku_Strategy
A windoku is just a double sudoku, 1 constraint extra for each cell. The first x-cycle it find is at 2 in column 5, which is just an boxline reduction from that cell. So you can exclude more which an easier strategie, those x-cycles are apparently hard for your solver.
Andrew Stuart writes:
Quite right, now I've appreciated the overlapping nature of the windows and boxes more work needs to be done. I was wondering why there are so many short cycles. Cheers
Edit 27 March: I have added in proper Pointing Pairs and Line/Box Reduction for 'windows'. I am recalibrating the grades as a result and will provide new examples in the solver example list.
Friday 25-Mar-2016
... by: David Munson, Los Angeles, California, USA
I have been noticing problems with the Sudoku solver. I'm working from Los Angeles, California, U.S.A. I cleared all my browsing data in Google Chrome, this includes "cached images and files". I have noticed that the Solver is no longer finding X-Cycles and it seems to also not be finding AICs.
Here is an example position; it is from your Jan 2016 Diabolical Puzzle Pack puzzle #14. From this position, after the Sudoku Solver gets through the basic 6 rules, the Solver finds an X-Wing that cancels two 9's.
Next after going thru the basic 6 rules, it finds an XYZ-Wing. But if XYZ-Wing is de-selected, then the solver should find an X-Cycle for 8's which cancels one 8. But it does not. Instead it finds an XY-Chain. Here is the X-Cycle for 8's: +8[B4]-8[B7]+8[C9]-8[H9]+8[H4]-8[B4]. This is Nice Loops Rule 3.
Also, after canceling the 8 on B4, then if we de-select all the rules from XYZ-Wing to Finned Sword-Fish, now the Solver should find an AIC. But the Solver proceeds on to Digit Forcing Chains. There are probably many AICs in this position, here's one of them: +3[B4]-3[B6]+8[B6]-8[B7]+8[C9]-8[H9]+6[H9]-6[H4]+8[H4]-8[G4]+3[G4]-3[B4]. Another Nice Loop Rule 3.
A month or so ago things were working fine. This happens for other puzzles too. I tried the above position in Windows Explorer as well as Google Chrome, same result. I still have Windows Vista Ultimate; don't know if that makes a difference.
Thanks for your time,
from Dave Munson
Andrew Stuart writes:
You are quite correct. So I had a look in the code, and lo and behold, there was still a debugging line in there from my last testing and update. It had the tragic effect of focusing the AIC search for X-Cycles and other AICS to just one candidate. The candidate in a test puzzles I was trying to fix 20 days ago. Bit of an oops. Which is why I don’t work on mission critical system. Thanks bringing to my attention with such a clear example. Solver updated.
I tried to solve a sudoku with your solver, but it failed when applying the "Alternating inference chains" technique. The solver excluded digit 3 from J6, which is actually the correct value. As a result, the solver gets stuck later and finds no solution. Could there be a bug concerning this technique? I don't understand it myself, so I can't tell what the problem is...
But I can assure that this sudoku has a solution, as I solved it manually and verified it!
JB
Andrew Stuart writes:
I made a tiny, tiny change a week ago, on this, to capture eliminations in a box where there are grouped cells like that. Good to have a counter example, I'll go back and see what I did wrong
Thanks for sharing
Edit: 29th Feb 2016: Fixed. Put in a check in the correct place and now this false positive does not occur.
Wednesday 17-Feb-2016
... by: Wolfgang Lentner, Bavaria
Hallo Andrew,
I checked a SUDOKU with your solver and saw something strange?!
SUDOKU: 000798354030514620000632108000006503020351400543987216370069845000800700080070900 has 83 solutions
using on this SUDOKU your strategie 26: "Altern. Inference Chains", we get ...
SUDOKU: 000798354030514620000632108000006503020351400543987216370069845000840700080070900, which has 58 solutions?!
How can the solutions get less, after a correct conclusion? Doesn't this mean, that the "new" 4 in H5 isn't a correct conclusion?!
greatings from bavaria Wolfgang
Andrew Stuart writes:
Only by accident. Any puzzle with more than one solution (or none) is not a real Sudoku puzzle - there is no 'correct' solution - and cannot be solved logically - so none of the logical strategies applies. They will mechanically chew on the puzzle but the results are meaningless.
Did you miss a clue or mis-type a clue?
Sunday 14-Feb-2016
... by: dimanche, france
Your internet pages are very interesting. Many solvers, congratulations. Solutions clearly explained...
What about an application which should built a sudoku at the level of your choice ?
For instance, suppose you build a scale with 30 different levels.
The user could choice level1+level5+level25, click on a button and a moment later the computer shows the wanted grid...
If this idea has some interest for you, mail me...
Andrew Stuart writes:
It's impossible to know what grade a puzzle is until you make it. So one cannot target a grade during the creation process. You have to make several and then find out which is the first in your target. This will be quicker for easy puzzles as they are more numerous. Some people do have click-button instant puzzle generators but their grading must be weak
Friday 12-Feb-2016
... by: S K JAIN, INDIA
I have written a sudoku solver software that can solve the sudoku puzzles from pure logic. I have implemented most of the widely used sudoku techniques (including the very advanced techniques). It can solve the hardest and toughest sudoku from pure logic (with the help of Adrianne's thread technique when the sudoku is stuck and all other techniques fail). You can download the source code for free for the sudoku software for free from my website. The website is : sites.google.com/site/skjgeek
Andrew Stuart writes:
Well done, programming a solver is the most satisfying endeavour. I've really enjoyed that aspect of it. You say "Adrianne's Thread". Do you mean "Ariadne's Thread"? I remember that strategy when Michael Mepham first published his Sudokus in 2005. It was the "all logic fails me" strategy. Since then the community has expanded the logic space to we need to resort to it less, but I class it as a "trial and error" strategy and therefore outside pattern based logic.
Thursday 21-Apr-2016
... by: Eyad Khalil, Palestine
Can the Daily Puzzles listed on your site be automatically imported into the Solver - without having to enter the 81 digits one at a time?
Thanks,
Thursday 21-Apr-2016
... by: Bert, Netherlands
last few days: unable to update kakuro puzzles:
Could't reach the Internet to pick up the puzzles.
No problems with the other puzzles.
Appreciate the alert!
Tuesday 19-Apr-2016
... by: Chris Green, Australia
I just solved Unsolvable #196, using a simple program I wrote (to save laborious work) plus a trial and error approach. I guess such an approach is not considered very valid since Andrew seems to be looking for more logic-based approaches.
Anyway, it took me 22 tries but when I eliminate the erroneous tries, it can be solved in 6 tries. (Of course, that's a case of 20/20 vision in hindsight!) I chose to start with Box 9 because it has the least number of possible candidates at the beginning.
The sequence I chose was: J8 =3, J2 =1, E1 = 1, B1=5, J4 =6 and finally, J3 = 5 for completion.
A New Metric for Difficult Sudoku Puzzles
which follows on from
Arto Inkala Sudoku
Thursday 14-Apr-2016
... by: UL, Hong Kong
Load Sudoku: CLICK TO LOAD-1[G2]+4{G2?E2}
I don't understand this strong link in the puzzle, could you please explain it?
Saturday 26-Mar-2016
... by: Uhm, Netherlands
Your solver does not find easy boxline reductions or pointing pairs, instead it finds x-cycles which does slowly the same reduction.
As you documented in:
http://www.sudokuwiki.org/Windoku_Strategy
A windoku is just a double sudoku, 1 constraint extra for each cell.
The first x-cycle it find is at 2 in column 5, which is just an boxline reduction from that cell. So you can exclude more which an easier strategie, those x-cycles are apparently hard for your solver.
Cheers
Edit 27 March: I have added in proper Pointing Pairs and Line/Box Reduction for 'windows'. I am recalibrating the grades as a result and will provide new examples in the solver example list.
Friday 25-Mar-2016
... by: David Munson, Los Angeles, California, USA
Load Sudoku: CLICK TO LOADI have been noticing problems with the Sudoku solver. I'm working from Los Angeles, California, U.S.A. I cleared all my browsing data in Google Chrome, this includes "cached images and files". I have noticed that the Solver is no longer finding X-Cycles and it seems to also not be finding AICs.
Here is an example position; it is from your Jan 2016 Diabolical Puzzle Pack puzzle #14. From this position, after the Sudoku Solver gets through the basic 6 rules, the Solver finds an X-Wing that cancels two 9's.
Next after going thru the basic 6 rules, it finds an XYZ-Wing. But if XYZ-Wing is de-selected, then the solver should find an X-Cycle for 8's which cancels one 8. But it does not. Instead it finds an XY-Chain. Here is the X-Cycle for 8's: +8[B4]-8[B7]+8[C9]-8[H9]+8[H4]-8[B4]. This is Nice Loops Rule 3.
Also, after canceling the 8 on B4, then if we de-select all the rules from XYZ-Wing to Finned Sword-Fish, now the Solver should find an AIC. But the Solver proceeds on to Digit Forcing Chains. There are probably many AICs in this position, here's one of them: +3[B4]-3[B6]+8[B6]-8[B7]+8[C9]-8[H9]+6[H9]-6[H4]+8[H4]-8[G4]+3[G4]-3[B4]. Another Nice Loop Rule 3.
A month or so ago things were working fine. This happens for other puzzles too. I tried the above position in Windows Explorer as well as Google Chrome, same result. I still have Windows Vista Ultimate; don't know if that makes a difference.
Thanks for your time,
from Dave Munson
Thursday 25-Feb-2016
... by: JB, France
Load Sudoku: CLICK TO LOADI tried to solve a sudoku with your solver, but it failed when applying the "Alternating inference chains" technique. The solver excluded digit 3 from J6, which is actually the correct value. As a result, the solver gets stuck later and finds no solution. Could there be a bug concerning this technique? I don't understand it myself, so I can't tell what the problem is...
But I can assure that this sudoku has a solution, as I solved it manually and verified it!
JB
Thanks for sharing
Edit: 29th Feb 2016: Fixed. Put in a check in the correct place and now this false positive does not occur.
Wednesday 17-Feb-2016
... by: Wolfgang Lentner, Bavaria
I checked a SUDOKU with your solver and saw something strange?!
SUDOKU: 000798354030514620000632108000006503020351400543987216370069845000800700080070900 has 83 solutions
using on this SUDOKU your strategie 26: "Altern. Inference Chains", we get ...
SUDOKU: 000798354030514620000632108000006503020351400543987216370069845000840700080070900, which has 58 solutions?!
How can the solutions get less, after a correct conclusion?
Doesn't this mean, that the "new" 4 in H5 isn't a correct conclusion?!
greatings from bavaria
Wolfgang
Did you miss a clue or mis-type a clue?
Sunday 14-Feb-2016
... by: dimanche, france
Many solvers, congratulations.
Solutions clearly explained...
What about an application which should built a sudoku at the level of your choice ?
For instance, suppose you build a scale with 30 different levels.
The user could choice level1+level5+level25, click on a button and a moment later the computer shows the wanted grid...
If this idea has some interest for you, mail me...
Friday 12-Feb-2016
... by: S K JAIN, INDIA
sites.google.com/site/skjgeek
You say "Adrianne's Thread". Do you mean "Ariadne's Thread"?
I remember that strategy when Michael Mepham first published his Sudokus in 2005. It was the "all logic fails me" strategy. Since then the community has expanded the logic space to we need to resort to it less, but I class it as a "trial and error" strategy and therefore outside pattern based logic.