This post highlights the updates on the 2019 review of the KrazyDad Insane collection on www.krazydad.com. In 2019, the collection continues to stand out as a challenge to human solvers, and a source of innovations. Continued congratulations to Krazydad’s dad, Jim Bumgardner.
If a Sysudoku program of updates in place continues, it might become rare to see grids presented as they were in 2013. Here is a grid in Brush Script MT, as opposed to the current Bradley Hand ITC, with multiple copies of multiple slink partnering candidates.
It’s actually a graphic demonstration of a trial contradiction, the extra term of a Single Alternate SdC removing all 4-candidates from the South box by misplacing one digit.
That isn’t repeated in the update, because in 2019, the review is a rare opportunity to create examples of pattern analysis as a follow up to a thorough AIC building phase. Now, KD Insane 4×5 finishes with a pattern slicing trial, completed with coloring.
When it comes to very hard puzzles, the update shows the value of using a human methods oriented solver to back up methods exploration. In this update it gave me examples relevant to systematic Sudoku that I would not have discovered, which will enrich my Sudoku experience to come. This is evidenced by a comparison to the original and updated review tables.
The updated review table has a new column for bypass placements, but overall, basic results are unchanged.
For that comparison, here is the 2013 review table.
Pattern analysis was an intended focus in the original and updated reviews, but the updated review used the solver to uncover available humanly practical method prior to patterns. Only three of the ten were solved without pattern analysis. The book 8 puzzle was originally solved by accidental error. Corrected, it required a pattern trial, in two different paths.
The 2013 LPO was a blanket term for any overlay method, but now is limited to pattern elimination by conflict. Pattern slicing and orphan detection are separate pattern analysis methods requiring freeform enumeration Some differences are terminology refinements. The ER in ER XYZ refers to the empty rectangle used to attach wings to hinge or victims to Z candidates. These are now iXYZ, the “i” being “inference chain” or “irregular” XYZ.
A unexpected, and welcome theme of the Insane review update was the boomerang, which showed up so frequently it’s name is hereby shortened to “boomer”. The name describes the spotting technique and the phase of solving in which the method fits. A boomer is found by examining slinks that start chains, and keeping the chains going by any possible means, while constantly checking for a return to the starting point. Starting points are cells, boxes and ALS value groups.
The boomer’s natural phase of solving is AIC building. This phase starts when the puzzle survives the systematic options of UR, SdC, XYZ wings, XY chains, X-chains, fish and coloring have their shot, and its time to throw in any possible means to keep a chain going. It’s a license to doodle and throw slides out.
Pattern analysis after AIC building is a return to systematic discipline. One reason why it techniques succeeds at this point is the focus patterns of values on the thin side of a candidate imbalance. When the imbalance is extreme, pattern analysis can discover orphans, candidates belonging to no patterns. A demonstration is coming up soon .
I made Jim’s guessing issue as the review began. The arbitrary guess of one cell value is demonstrated to be unnecessary, and inadequate, for the Insane collection. It would fare poorly with the Insanes because, quite frequently, an arbitrary guess will be indecisive and require backtracking over successive guesses. Cherry picked guesses guided by computer solutions are misleading.
The review also reveals the rich treasure of logical truth in the collection that is hidden by arbitrary guessing. Those who guess their way through hard puzzles and claim to be solving them might as well just look up the solution and copy it.
I believe the Insane collection does require trials, but not guesses. I hope the review made the case that trials of cluster colors and long inference chains should be deferred while reasonable alternatives remain. But clearly, labels of “trial and error” or “guessing” do not apply to trials which are logically constructed as a last resort, even though such trials will occasionally reveal a solution.