dailysudoku.com

[alanr555]

It would seem that the gradings MAY be set by reference to the need to
use particular computer-oriented techniques (eg the so called methods A
through F). I find that I can solve the "very hard" examples fairly
consistently but I get stuck on the 'Medium' examples.

This seems to be a difference of approach. I use TWO solution grids and
put the contents of these in the cells. The standard grid (123456789 with
impossibilities removed) is at top left and the second grid at bottom left.
Thus, I like to use the LARGE print!!

The second grid consists of marks where a particular number MUST be in
one of two places within a block of nine (ie the other seven cells have
been excluded). The virtue of this is that when a cell containing values
of the second grid is resolved to a different number, it is IMMEDIATELY
possible to resolve the cell containing the paired item, (eg if value 6 must
be in A1 or C3 and value 7 must be in A1 or B3 then resolving A1 leads
immediately to the resolution of C3 and B3 - unless the value in A1 is
either 6 or 7, in which case the new information is the removal of the
pair mark and the usual removal of a possible value in the standard grid).

My methodology is to mark up the pairs first using "look and see" logic
working primarily on groups of three blocks in a rectangle (eg the first
three columns for nine rows). This process is used on the three column
groups, then the three row groups and then all rows/columns with six
or more resolutions are inspected. Then this process is repeated in
greater detail (the first scan is relatively superficial).

Where the pairs are in a horizontal or vertical line, the existence of a
third entry of the same number in the row/column leads to some good
"new" information and a resolution. This is much easier to spot than the
search for combinations in the more complex rules (eg trying to find an
example of F is wearying on a manual scan but fairly clear using the
secondary grid).

Having exhausted the two scans and row/column checks (plus the
consistency of each block), I then revert to the standard method of
determining the remaining possibilities. However the second grid can
assist in eliminating what would otherwise look like feasible numbers.
Very often this will split a row/column into two groups (eg 56,56 and
78,89,79 by shewing say 567 in the first cell to be impossible as the third
and fifth cells form a pair. If 567 were possible then there would be one
big group 56789 rather than two subgroups 56 and 789).

Any cells resolved, I mark by a circled number as otherwise the chart
looks like a veritable mass of numbers.

Can anyone else point me to ways of solution that do not depend on
eagle-eyed spotting of patterns in the way that computers would approach
the solution grid? I marvel at people who can hold vast arays of helpful
information in short-term memory in order to resolve a cell. Am I unusual
in needing a lot of "aide-memoire" marks?

Alan Rayner BS23 2QT

[Smiffy]

I just use a single grid and don't mark in all the candidates. It's just too messy - can't see the wood for the trees.

I start with a scan for "easy meat" by intersecting rows and columns with boxes. I used to just do this scattergun but I found I tended to miss a few, so now I use a methodical approach. I start by picking one of the more common numbers on the grid, and examine each box in turn. If there is only one square the number can be, I mark it in big. If there are two squares I mark the numbers in small, in one of the corners of the squares. If there are 3 or more possibilities I leave it blank. After I have repeated this process for all the numbers, I have a quick root around to see if anything else has emerged.

Often a square will contain more than one small number at this point. What I do is to use different corners for different numbers, so if a square has a 3 in the top left and a 7 in the top right, I know that it does not necessarily mean the square can only be 3 or 7 - this being depicted by the two numbers close together. If it turns out that 3 and 7 are the only numbers, I join them with a line.

Next up I look for rows, columns and boxes with low numbers of missing entries. This may yield some unique numbers (marked in big), and again my policy is to mark small numbers on the grid only if there are two possible numbers for the square. As you all know these doublets can be very useful especially when you get a matching pair (e.g. 56-56).

As well as nearly complete lines I also target those missing numbers which are present in intersecting boxes, as these elminate a lot of squares. If I find only two slots for a number in a line, I tend not to mark it in, rather check that there isn't another number which shares the same two slots (revealing a hidden pair). For boxes, I tend to look for those whose missing numbers are not the same as those in intersecting boxes and lines. Hidden pairs in boxes will have already been revealed by earlier steps.

Only when I get stuck after the above do I resort to looking for triplets. Where I do have to look at larger numbers of remaining candidates, I prefer to write the missing numbers for lines or boxes oustide the grid - again because my eye and brain get fuddled when the grid is full of too many numbers.

Finally, I recently read this http://www.simes.clara.co.uk/programs/sudokutechniques.htm after finding it on this forum. I was quite pleased to see that I have been using most of the techiques already without knowing what they were called

[bird316]