## Diminishing return?

Advanced methods and approaches for solving Sudoku puzzles

### Diminishing return?

Having just gone through many posts in this section including some from last year, it appears no 'super-solver' exists. And I'm still waiting on inventor to share the super-technique that was alluded to over a year ago.
Mike Barber did some nice work on heirarchy of techniques on 25000 random puzzles. In the end though, after listing 49 different techniques, he was still only 92+% successful in solving the puzzle.

The question is: Has anybody done an analysis of how many techniques it would take to solve 75, 80, or even 90% of all random generated puzzles?

Based on his data, every technique that you add has a smaller and smaller impact on the number of puzzles solved.

For those that don't use computer solving programs, knowing they can solve 85-90% of puzzles with the first eight to ten techniques might just be good enough.

Any thoughts.
hrcjcr

Posts: 19
Joined: 19 November 2005

Self taught techniches solve more than half of random puzzles.

The last 10% keeps me active and learning.

The very last 1% keeps programmers working.

Edit: x-wing should be self taught, some people read about it first.
wapati
2010 Supporter

Posts: 527
Joined: 13 September 2006

### Re: Diminishing return?

hrcjcr wrote:The question is: Has anybody done an analysis of how many techniques it would take to solve 75, 80, or even 90% of all random generated puzzles?

I have a terrible memory but, when I first started reading the forums, I believe that I read that most random puzzles can be solved with just a few, simple techniques. In fact, people were posting how fast their solvers ran so they could process a large number of random puzzles in search of interesting puzzles to solve. Now, discussions of random puzzles have been replaced with discussions of pattern generated puzzles. The latter group has a much higher success rate on producing difficult puzzles that need extended techniques to solve them. Of course, any technique statistics gathered from them will be biased because of the method of generation and the hierarchy in which the techniques were applied.
daj95376
2014 Supporter

Posts: 2624
Joined: 15 May 2006