by **Lunatic** » Wed Jun 11, 2008 6:14 pm

Ok, I decided to take my solver to the next level, implementing a lookup routine for ALS-xz.

At first I was thinking to collect all possible ALSs on the grid, and then to handle all possible pairs of ALSs looking for possible common restricted candidates, and if found one then looking for a non restricted common candidate that could lead to eliminations. Sounds good in theory, but i think i will end up with a huge amount of ALSs to collect and compare, mostly leading to nothing and wasting time. So, i need a different approach. I searched the "programming sudoku" forum but to no avail.

Thinking it over, i came to the conclusion to look for possible restricted common candidates first, and then lookup if i could place them in ALSs, and if so, look for a non-restricted common candidate that leads to eliminations. How about this ? If anyone knows a better approach, please let me know...

Just a few quick thoughts...

- Could it be that any naked subset in its whole is not useful as part of an ALS ?

- I think that mixing parts of two naked subsets together for making a single ALS is not very useful either. Which would mean that any useable ALS is a subset from a naked subset (or from the whole set if no subsets are present)

Can anyone confirm this ?

I asked the same questions at the programmers forum, but didn't got much replies...

Tnx,

Marc.

Marc

~~~<><~~~<><~~~<><~~~<><~~~

The Lunatic is on the grass...

...under the sun everything is in tune,

but the sun is eclipsed by the moon...