The New Sudoku Players' Forum

[scsnesrud]

I've been learning coniugate pair chains. Could someone explain how to determine where to start them. Thank-you, Steve

[SpAce]

First identify all bi-location strong links for the target digit. Then forget Simple Coloring and learn about X-Chains (and Alternating Inference Chains (AICs) in general). That way you'll learn Simple Coloring as a side-effect, but notice that it's quite useless without upgrading to Multi-Coloring or X-Coloring -- which actually are even more useless because of the added complexity. X-Chains and other AICs are much more versatile and teach you the concept behind coloring techniques too (and much better), but you won't need or want to do coloring once you've learned them. My only exception to that rule is very difficult puzzles where something like GEM (3-D Medusa on steroids) may help to find hiding chain and net deductions, though I haven't used that in a long time either.

PS. It's tempting to start learning about chains with Simple Coloring because, from one perspective, it's the simplest of such techniques. That's why I started with that too, but like I said above, quickly found it was pretty useless in practice. It actually made it a bit harder to learn proper chaining techniques because it's a net (like coloring in general) and only uses strong links.

[JasonLion]

You start them anywhere that results in an elimination. That typically requires trying many possible starting/ending points, possibly through coloring or by already knowing which candidate you want to eliminate.

[SpAce]

If you insist on doing Simple Coloring (or 3D-Medusa) then you should start by identifying the strongly linked clusters, like I already said. Then pick the largest of them, or otherwise one that looks most promising, and color that first. It doesn't matter where you start the coloring within a cluster, because you only have strong links and they produce the same results no matter which way you do it (one nice feature of basic coloring techniques). Then analyze the results and see if the colors reveal any individual eliminations or contradictions that prove one color false. If that doesn't work, pick a different cluster (but use different colors). The results of certain more powerful coloring techniques depend on the starting point if they use weak links as well.

[scsnesrud]

Thanks I'll work on the x-xy xyz and theAIC's.