Reviews

Systematic Sudoku applies human engineering principles to a game, Sudoku solving, with the goal of making it a more pleasurable and more widely used mentally therapeutic activity. Techniques of basic candidate identification, and advanced value placement were described in detail in the first year of the Sysudoku blog, and now in an accessible set of pages, the Guide. There is a systematic process order, the Sysudoku Order of Battle (SORB), designed to concentrate human attention in a generally efficient way.

To further this goal, Sysudoku reviews both the writings of Sysudoku  authors and collections of Sudoku puzzles, as they become available.  The expert reviews attempt to evaluate claims and theories about human Sudoku solving from a human engineering perspective.  In Sudoku, as in current life, this involves judgement as to what is practical for human solving, as opposed to measures only practical in solving by computer coded algorithms.

The objective in collection reviews is to rate collections in order of difficulty, and to provide realistic examples to show readers what is to be gained in playing the Sudoku game efficiently and at more advanced levels.

Ratings by Sudoku publishers are subjective and unreliable.  Terms like Very Hard, Extreme, Fiendish, Diabolical, or Nasty have become meaningless.  The Sysudoku Order of Battle does provide a fair way out. It is to report a Sysudoku solution path in a way that can be accurately followed. The collection’s difficulty can then be judged by the reader by how far a representative set of puzzles get before collapsing.

That sounds good, but in order for readers at every level to make comparisons, the reports must be accessible at a glance. That is reason for the

Review Table.

The drag off page is a list of collection review tables for all the puzzle reviews. How far you would go from left to right depends on your level of difficulty. Here is a current table:

The left, highlighted column identifies the puzzle number or difficulty rating within the collection.

Under

Gv  is  the number of givens, of interest to some readers, but not in itself a measure of difficulty.

BY  is a number of primary interest to beginning sysudokies, the number of cells reserved in the bypass for specified values, i.e. for subsets, including clues. In any column, the code – – – signals a collapse. Two review puzzles collapse in the bypass in the sample table above.

Box is the number of slinks and aligned triples, then after the slash, the number of  cells reserved in box marking.

Line is a tally of the line marking. It’s a set of lists, separated by commas.  Each list is headed by the number of free (unreserved) cells on the line, followed by a colon.  Lines are marked in increasing order of free cells, taking easier markings first. Changes in a marked line are followed up immediately, and can change the order of marking of remaining lines. List heading numbers can therefore decrease, but generally decrease.

Significant events are signaled by short event codes. For example, in the first puzzle, a rare hidden triple occurs among the lines of 5 free cells. In the 5: list, the code ht counts as one line. Three such lines are marked before the hidden triple and two after. Six consecutive lines of 5 free cells were marked.

When a line marking results in slinks or aligned triples and clues, this result is recorded in parentheses as (slinks, cells reserved) or (cells reserved), without an event code or immediately after one. If the hidden triple above had resulted in three slinks and two clues, the 5 free cell list would have been written 5:3ht(3,2)2. If instead, it had produced 4 clues and no slinks or aligned triples, the list would read 5:3nt(4). If it was 2 slinks and no clues, the list would be 5:3ht(2,0)2.

Event codes used include:

bxl -box/line

np, nt, nq, hp, ht, hq – (naked, hidden)( pair, triple, quad)

hdx – hidden dublex

xw –  W-wing

ER – by empty rectangle,  an X-chain weak link

And under the heading BV is the number of bi-value cells (bv) left by line marking.  Early advanced methods depend on the bv field.

The rightmost column contains brief descriptions of advanced and extreme methods employed for the solution, and any trial results. The order of these methods change in response to results, but generally follows the Order of Battle stages:

Bv scan:  unique rectangle (UR), hidden UR); Sue de Coq; XY-chains; XYZ-wings; APE;  BARN; XZ_ALS, ALS toxic sets

X-panels: X-chains, grouped X-chains, regular, finned, kraken fish;

Medusa coloring and bridging as available.

Small ALS grid scan: XZ pair match, ALS APE.

X-panels: freeform pattern analysis.

AIC building: AIC ANL, boomerangs, 1-way, complex 1-way.

Lite coloring: Completing extensive but stalled clusters.

Suset ALS enumeration: ALL ALS in three tables

ALS mapping: ALS node AIC, ALS_XZ, ALS-wings