Hi jco,
Sorry about the late response.
Your comment on hitting backdoors in manual solving using GEM (and on GEM in general) is (are) very interesting for me.
Glad to hear that! I'm happy to help in any way I can.
Many times I struggle to solve difficult puzzles manually. Still don't have a firm grasp on GEM. I know Medusa Colouring very well and when a puzzle is weak to Medusa, I solve it quickly.
Medusa is a perfect starting point for learning GEM. In fact, the core cluster of par-candidates at the beginning of a GEM-coloring is a Medusa. However, as you must have noticed, it rarely gets you far -- hence the GEM-extensions.
When I start a GEM-coloring I almost always use a seed that produces the largest par-cluster, because it has the biggest chance of having backdoors and being productive in other ways. Solving it (i.e. finding a contradiction for one or the other parity) might solve the whole puzzle or at least a big part of it. Yet, as I've said, the primary goal is to find trap eliminations instead of a contradiction.
I studied your previous presentation
Do you mean
this?
In difficult puzzles, I end up spending a lot of time to find good AICs (when Medusa does not help). Even after marking all strong links for each digit, since I end up using ad-hoc approaches to find AIC from the links, and it is not rare that some AICs found are not productive (meaning does not help advancing the puzzle).
I wouldn't worry about that too much. You're clearly on the right track. The more AICs (productive or not) you find, the more you learn. It's also good that you've learned to find them by marking and using the strong links directly, instead of relying only on coloring.
So, to solve manually a difficult puzzle I need many non-basic steps (like 10 to 15).
That's not dangerous per se. When I started learning AICs I tried to find as many of them as possible for a given puzzle state. I didn't even try to look for backdoors or weak points to attack. Sometimes I even delayed hitting the most effective eliminations on purpose to prolong the fun. Eventually it wasn't so much fun anymore, and then it became more interesting to find shorter solutions. All the gained experience helped with that, too.
Back to GEM, when a GEM solution is shown, it is always presented in the final configuration, making hard to someone trying to learn it to arrive at that configuration (this does not happen with Medusa).
Did the step-by-step instructions in the previous link help at all? I know it's a pretty complex example, but it should demonstrate the basic principles.
I always searched for GEM solutions presented in steps, showing at least some possibility matrices of partial steps towards the final configuration (pictures/images are no good for me. Lots of colours and gets very confusing to follow, if grids where used instead it would having been easier IMO).
Ah, in that case the linked example probably didn't help much. Does that mean you'd prefer GEM examples with David's notation? They're pretty tedious to produce.
Your GEM solutions presented in the section I understood completely (except for one).
Glad to hear. Which one did you not understand?
However, when trying to use GEM in a puzzle twice I reached a false elimination (other times of course helped finding good eliminations and move on), which indicates still poor understanding on that part on 'promotion'.
Promotions are probably the likeliest situations to make mistakes. They're not needed very often in normal puzzles, but they're important to find (continuous) loops.
Anyway, this is off-topic for this section, but I really wanted to emphasize the importance of your comments on GEM for people like me struggling to fully understand GEM.
Well, I'm glad to hear there's at least one person even wanting to learn GEM, and even happier if I've managed to help!
It is a shame that no GEM or Medusa Colouring solution has been presented anymore to solve a puzzle in this section.
I guess I'm the most likely person to produce them, but I'm pretty much retired from solving. Besides, most of the currently trending puzzles in this section are oriented towards software solvers. I don't have time or interest to solve them manually or to filter out ones that I'd like to. Fishy puzzles aren't well-suited for coloring solutions anyway.
While I appreciate that there's now more variance and learning opportunities than before, I have to admit I miss the consistently fun daily puzzles by Dan and tarek. Many of them didn't even need coloring, and the rest were perfectly suited for GEM. I also miss seeing others' creative solutions, which are less common and more scattered these days.
