I already posted one of the first appearances of the pattern in the ph databases. It is from the batch of puzzles from dobrichev added and/or provided to champagne on December 3, 2012, which includes the puzzle I posted previously. There are four from that batch, and I don't know which was actually found first (not sure if there was any sorting by SER before they were added; the first two IDs are 11.0, the other two 10.9).
They are in the file above, but here they are again on their own:
- Code: Select all
........1.....2.3....14..5...678...2.18.2467.72.6.1.4..61.7....2.4.18...87.4.6...;11.00;1.20;1.20;dob;12_12_03;534519;32;*
..............1234..2.5.6....3...76..47.3...262..78..3.764....828.....7.3.4..7.26;11.00;1.20;1.20;dob;12_12_03;536614;31;*
..............1..2..3.4..15.341....6.7.38..418.1.6...7.478.....16.4.3.7.3.8.76...;10.90;1.20;1.20;dob;12_12_03;545447;31;*
..............1..2..3.4..15.341.6..7.7.38..418.1.7...6.478.....16.4.3.7.3.8.67...;10.90;1.20;1.20;dob;12_12_03;545448;32;*
Obviously the last two are closely related. All four have single guardian non-degenerate trivalue oddagons.
The 23 puzzles from dobrichev in the above list were all added in Dec. 2012 or Jan. 2013. champagne's 4 were added Dec. 2018 or Jan. 2019. Paquita's were in the second half of 2019, while mine were from 2020.
So there's a roughly 6 year gap between the first puzzles with the pattern and the next appearances. We probably can't establish definitively whether there is a direct line from the dobrichev puzzles to the later ones in ph2010, never mind to the first T&E(3) puzzles (which still seem to be from August or September 2021, though those scripts haven't finished yet). But I will still at some point attempt to trace back using neighborhood searches as a guide.
------
[edit]I should clarify/reiterate that the list of 838844 puzzles linked above is generated without actually checking for trivalue oddagons at all. The first pass filter
only checks for the conditions guaranteeing that it is not a degenerate tridagon by Denis' definition; namely, that there are at least three digits which appear at most once as givens with those appearances confined to a single box. The non-degenerate trivalue oddagon, if it exists, is in the four boxes which do not share a band or stack with the box containing those digits as givens.
This has the advantage of being very fast, but obviously it is possible for a puzzle to meet the first pass conditions and not actually have a non-degenerate tridagon because of the placement of the other digits in the other boxes. For example:
- Code: Select all
2...8.....3...9......7...6....456.....6...7.9.......8..89.4567.4..6.79.8..789..45
would make it through the first pass filter, but box 5 cannot have three cells containing 123 in a diagonal pattern.
That said, the vast majority of the list do have non-degenerate trivalue oddagons (even by my stronger restriction on "non-degenerate") - this may be due to the bias toward high SER, or it may just be more common in unique puzzles with this property, I'm not sure. Once the current script is finished, I'll recheck the ones marked as degenerate (again, by my definition, not Denis').
My current script is working through these and giving the exact details (digits, cells, number of guardian cells/candidates) for the non-degenerate cases, so I'll provide that when it's done (about 5% so far, it'll be a few days).