The New Sudoku Players' Forum

[denis_berthier]

.
A few comments about the few above posts.

1) SER
First, I've completely stopped using the SER as a rating. It has been useful as a shared measure of difficulty, but it now seems to have too many problems (not stable under isomorphisms, dependent on uniqueness, irrelevant when some pattern is present in a puzzle...).

There's a persistent misconception that there could be a unique rating. Adding the simple tridagon rule to SER would in no way change the problem. I've shown that we'd need a whole set of rules (Trid-ORk-chains or anything similar). But then we'd also need Imp630-ORk-chains and DP-ORk-chains and any chains based on any pattern that could be discovered in the future.

My stance is, there are various ratings, adapted to different purposes. For a puzzle with a tridagon the W+ORk-W rating has proved useful.

2) Tridagons vs T&E(3)
There has never been any "tridagon search hype" - because there has never been any tridagon search at all.
What there has been is a search for T&E(3) puzzles. This has been extremely successful, with mith, coloin, Paquita, Hendrik Monhard... finding millions of them.
More recently, there has been a search for high BxB puzzles. It has also been very successful, with BxB upto 14 (the previous bound was 7 - with only 3 puzzles).
That all these puzzles have a non-degenerate tridagon is an additional result. It says something about high T&E or BxB levels. But no one can measure the real value of this result if they think there has been a search for tridagons.

Very recently, there has been a misguided search for solution grids having a very special type of "tridagsn-solution". As far as I can see by the current results, it's a failure. I'm not saying it wasn't a good idea to try; it was.
But, upon seeing the intermediate results, the failure shouldn't be a surprise:
- more than half of the solution grids have this pattern (implying that the pattern is not selective enough);
- the pattern excludes possibilities of other tridagons in the puzzles (so that the pattern in the solution grid is neither selective nor inclusive enough);
- having a solution grid, all the work of finding minimal puzzles - the hardest part - remains.
.

[champagne]

The tridagon (thors hammer) pattern is known since at least 3 years, and it was a sensation, that puzzles containing it, could have the highest ratings. But many of these puzzles could easily be solved manually using this pattern (e.g. the first TE3 puzzle was manually solved in 10 minutes).
So the ratings don't say anything about, how hard a puzzle is for a manual solver. In my eyes they are of no worth (what should a rating for a solver program say, if any puzzle can be solved in a millisecond ?)
Looking at a list like coloin's above, i have to try each one manually to see, if it is hard or not. E.g. the first one isn't - tridagon, RT and solved - nice for the puzzle section, but my Sukaku Explainer takes one and a half minutes to spit out a rating of 10.6. This is just wasted time. I'm faster to see the tridagon extra candidate and then get the rating in a second.
So this is the morbid situation since 3 years.
[Added:] And yes, i repeat this 2 times each year, as long as this tridagon search hype lasts.

Hi eleven,
I now see your point and I feel quite comfortable to answer.

First of all, I always considered the players view as important. The title of the forum is "Sudoku Players' Forum".
I have never been a skill solver, but from the very beginning, using a computer to help me, I tried to stick to paths that a manual solver would like to use. This pushed me to give up with my first set of rules in "full tagging" and to work more with the set of rules used in Sudoku Explainer.

This forum has a special section "puzzles" for the players, but is open to other views in the sudoku field.
a Huge power has been spent to prove that we have no "16", something of poor interest for a player,
I later spent as much power to check that we know all "17" again no interest for skill players, but more power had been used here and there to find them, usually easy puzzles.


Defining what puzzle can be of interest for the manual players is a hard task with as many answers as we have players, and the response changes over time.
Having spent much time on exotic patterns, I would say that a boring sequence like in the very old "snail" puzzle has small value for them;
Today when a puzzle reach an ER around 7, some players stop telling "now not too hard".

So I could say that generally speaking here to-day :
The puzzle must rate high with the usual tools
It must contains some properties leading to a relatively short path.

Easy to write, but how to find such puzzles is another story.

You are one step after (or before), not exactly on the solver path, but just trying to select a puzzle of interest in a file. You are doing this in a file containing puzzles having the tridagon pattern somewhere in the resolution path, leading to the high rating, so yes, this should not be done using the tool leading to the high rating.
From what I experimented with the exocet pattern, I could say that if I wanted to do the same in your situation, I would use a tailor made solver code clearing from the file all the puzzles "solved" (where a short path has been defined) This is what I have done years ago with the file of potential hardest

Your complain is posted in a thread looking for "non degenerated tridagon", the worst situation to look for puzzles of interest in your situation.
The "loki" puzzle has the same property, but after a lot of vicinity search, we have a huge file where players got puzzles giving a lot of fun working in large "impossible patterns". The file is likely large enough to cover their needs in the tridagon pattern
BTW, finding puzzles that could be of interest for players is a hard task, and players need files containing such puzzles,
and, unless they do it themselves, they need a code to extract in these files puzzles of interest.
Where we are, we have no idea of the next pattern of interest.

[eleven]

Hi champagne,

first of all i have deep respect for your sudoku work.

Sorry for hijacking your thread for my regular lament, the relationship is just that your outcome is another big bunch of tridagon puzzles. I really suffered, when the hardest thread was flooded with tridagon puzzles with extremely high ratings, which were so easy to solve manually.
And none of the programmers still had the mercy to integrate the tridagon pattern into a solver with trustful ratings for manual solvers, as Sudoku Explainer had been for 15 years.
There had been a small number of exceptions, where new solving techniques made puzzles essentially easier. But now in the ER 10.5+ field, these "exceptions" are the majority, all because of one simple and easy to spot pattern, which is hard to resolve for chaining algorithms. And no public solver is available, which gives us useful ratings for a bunch of tridagon puzzles.

[denis_berthier]

no public solver is available, which gives us useful ratings for a bunch of tridagon puzzles.

Except SudoRules.

[champagne]

Hi champagne,

first of all i have deep respect for your sudoku work.

Sorry for hijacking your thread for my regular lament, the relationship is just that your outcome is another big bunch of tridagon puzzles. I really suffered, when the hardest thread was flooded with tridagon puzzles with extremely high ratings, which were so easy to solve manually.
And none of the programmers still had the mercy to integrate the tridagon pattern into a solver with trustful ratings for manual solvers, as Sudoku Explainer had been for 15 years.
There had been a small number of exceptions, where new solving techniques made puzzles essentially easier. But now in the ER 10.5+ field, these "exceptions" are the majority, all because of one simple and easy to spot pattern, which is hard to resolve for chaining algorithms. And no public solver is available, which gives us useful ratings for a bunch of tridagon puzzles.

Don't worry "eleven", I understand your feeling of an overflow of tridagon puzzles.
BTW, I thought that this pattern had been brought by mith as tridagon, with the way to solve it...
I just learned that it is a by product of the T&E(3) search.

I understood the huge amount of interest on it at the very beginning as a dream to break the 11.9 SER. I never took the SER rating as more than a reliable and understood idea of the potential complexity of a puzzle;
When I was looking for potential hardest puzzles, I used rating >=10.0 as seed for the next vicinity steps.
This is likely what I could do here if I wanted to add puzzles with the tridagon pattern in the path.

I just wanted to check if we had unknown high ratings outside the "loki" family, assuming that this would be as productive as the loki seed.
But this leads to billions of puzzles to sort in each run. After filters, this is still thousands of puzzles per million of starts rating skfr >=10.5, so yes, this thread will continue to feed the flow of puzzles useless for manual solver.

But I'll stop the scan as soon as a new skfr 11.8 will come (if any)

Finding new patterns of interest for players will be tougher and tougher. The file of potential hardest could have other puzzles of interest. I updated in the past a sub file of "no exotic pattern known". Not exactly what you are looking for, but not so far.

[denis_berthier]

BTW, I thought that this pattern had been brought by mith as tridagon, with the way to solve it...
I just learned that it is a by product of the T&E(3) search.

This pattern has been found by mith while he was looking for high SER puzzles. Loki, the puzzle that started the whole story, happened to be the 10th one with SER 11.9. I noticed it was also the 1st one to be in T&E(3).
The production by millions occurred when mith turned his search criteria to being in T&E(3).
All this is perfectly clear in the "hardest" thread.
.

[champagne]

BTW, I thought that this pattern had been brought by mith as tridagon, with the way to solve it...
I just learned that it is a by product of the T&E(3) search.

This pattern has been found by mith while he was looking for high SER puzzles. Loki, the puzzle that started the whole story, happened to be the 4th one with SER 11.9. I noticed it was also the 1st one to be in T&E(3).
The production by millions occurred when mith turned his search criteria to being in T&E(3).
All this is perfectly clear in the "hardest" thread.
.

We are closer to what I had in memory

[champagne]

Here's one where the "odd box" in the magic square changes from box 5 to box 2:

Code: Select all

tridagon digits: 1,2,3
magic square: B1245
                              type 2
+-------+-------+-------+    +-------+-------+-------+
| 4 5 . | . . . | . . . |    | 4 5 1 | 3 7 8 | 2 6 9 |
| 6 . 7 | 9 . 5 | . . . |    | 6 2 7 | 9 1 5 | 8 4 3 |
| . 8 9 | . . . | . . . |    | 3 8 9 | 4 6 2 | 7 5 1 |
+-------+-------+-------+    +-------+-------+-------+
| . 6 4 | . 8 7 | . . 5 |    | 1 6 4 | 2 8 7 | 9 3 5 |
| . . . | 6 . . | 4 . . |    | 5 3 8 | 6 9 1 | 4 2 7 |
| 9 7 . | 5 . . | . . . |    | 9 7 2 | 5 4 3 | 6 1 8 |
+-------+-------+-------+    +-------+-------+-------+
| . . . | 8 . 6 | 1 7 4 |    | 2 9 5 | 8 3 6 | 1 7 4 |
| . . . | . . 4 | 3 . 2 |    | 8 1 6 | 7 5 4 | 3 9 2 |
| 7 . . | . . . | . 8 . |    | 7 4 3 | 1 2 9 | 5 8 6 |
+-------+-------+-------+    +-------+-------+-------+

{-7r7c8,+9r5c6} -->
                              type 1
+-------+-------+-------+    +-------+-------+-------+
| 4 5 . | . . . | . . . |    | 4 5 3 | 7 2 8 | 6 1 9 |
| 6 . 7 | 9 . 5 | . . . |    | 6 1 7 | 9 3 5 | 2 4 8 |
| . 8 9 | . . . | . . . |    | 2 8 9 | 4 6 1 | 7 5 3 |
+-------+-------+-------+    +-------+-------+-------+
| . 6 4 | . 8 7 | . . 5 |    | 1 6 4 | 2 8 7 | 9 3 5 |
| . . . | 6 . 9 | 4 . . |    | 5 3 8 | 6 1 9 | 4 2 7 |
| 9 7 . | 5 . . | . . . |    | 9 7 2 | 5 4 3 | 8 6 1 |
+-------+-------+-------+    +-------+-------+-------+
| . . . | 8 . 6 | 1 . 4 |    | 3 2 5 | 8 7 6 | 1 9 4 |
| . . . | . . 4 | 3 . 2 |    | 8 9 6 | 1 5 4 | 3 7 2 |
| 7 . . | . . . | . 8 . |    | 7 4 1 | 3 9 2 | 5 8 6 |
+-------+-------+-------+    +-------+-------+-------+

45.......6.79.5....89.......64.87..5...6..4..97.5........8.6174.....43.27......8.
451378269627915843389462751164287935538691427972543618295836174816754392743129586
45.......6.79.5....89.......64.87..5...6.94..97.5........8.61.4.....43.27......8.
453728619617935248289461753164287935538619427972543861325876194896154372741392586


Thinking of a possible canonical morph of a puzzle giving an easier way to see if we can have a bridge to the loki family, this example does not push to optimism.

assuming that the canonical magic square is in a "type1" or "type 2" in four boxes, here, with a -1 +1 vicinity single move , the new puzzle belongs to the family, but to get the square in the expected morph, we have to exchange columns 4<->6. then we have the magic square, but top down.

For me the best process could be to apply the known vicinity steps and to see if one new puzzle hit the loki family.

[champagne]

Loki, the puzzle that started the whole story, happened to be the 4th one with SER 11.9.

Late check, but this seemed to me very low.

status of the data base in November 2020

Code: Select all

98.7.....7.....6....6.5.....4...5.3...79..5......2...1..85..9......1...4.....3.2.;11.90;11.90;11.80;GP;champagne dry;1;22;
98.7.....6.....87...7.....5.4..3.5....65...9......2..1..86...5.....1.3.......4..2;11.90;11.90;11.60;GP;kz0;11523;23;
12..3....4....1.2...52..1..5..4..2......6..7......3..8.5....9....9.7..3......8..6;11.90;11.90;11.30;elev;second flush;2;23;
.......39.....1..5..3.5.8....8.9...6.7...2...1..4.......9.8..5..2....6..4..7.....;11.90;11.90;11.30;tax;Golden-Nugget;3;21;
.2.4...8.....8...68....71..2..5...9..95.......4..3.........1..7..28...4.....6.3..;11.90;11.90;9.90;elev;3;4;22;
........1....23.45..51..2....25...1..6...27..8...9......42....7.3...6...9...8....;11.90;11.90;9.90;dob;12_12_03;248078;23;
12.3.....4.5...6...7.....2.6..1..3....453.........8..9...45.1.........8......2..7;11.90;11.90;2.60;elev;1;5;22;
5.6...7...1.3.....8...5.9.....1...2.....8.6.7.....2.4.7...9...6.3...42....5......;11.90;1.20;1.20;OW;2015_08;1744614;22;
..3..6.8....1..2......7...4..9..8.6..3..4...1.7.2.....3....5.....5...6..98.....5.;11.90;1.20;1.20;elev;2;6;22;


[denis_berthier]

.
My mistake; I corrected it.
Loki was the 10th with SER 11.9. So, not even very so remarkable according to the prevailing (if not unique) search criterion of that time.
It was the 4th not in B7B.
And it was the first in T&E(3) - which is what really started the big bang of the search for T&E(3) puzzles, as I predicted it here: http://forum.enjoysudoku.com/the-hardest-sudokus-new-thread-t6539-1048.html
.

[mith]

Very minor correction to the previous:

The trivalue oddagon/tridagon pattern was first pointed out by marek stefanik in response to the high SER/high clue count puzzles I was finding at the time. As discussed elsewhere, it is present in a handful of puzzles prior to my joining the high SER search (dobrichev contributed the first few), but I started generating them in bulk as my scripts found higher and higher SER for this high clue count minimals. There was a bit of discussion about it here (and a bit more elsewhere, including a youtube discussion of the pattern), well before Loki was found by me and discovered by Denis to not be in T&E(2) - though that discovery certainly exploded efforts to generate these specifically.

(As it happens, I'm at a sudoku convention and presenting something tomorrow on this topic, so I've been digging up old posts for that... Still planning to post some updates in the T&E(3) thread at some point, life is hectic.)

[champagne]

I have to apologize, but I was not very active on the forum at that time and I missed the discussion with "marek stefanik".

[champagne]

The chunk 2 000 001 - 3 000 000 has been closed with 14562 puzzles having a rating >= 10.5
Coloin has the file for a better analysis.

The best ratings are

Code: Select all

2.........6.3.....9.....5....687..31...5.68.7....1365.67..31.85..87.51.6...68..7.;11.3/11.3/2.6
.......2659...6..4....8.......64..32...3.27.8....7846.6...23.8....8.46.33..76.24.;11.3/11.3/6.6
.3..4..7.5........9.....4.....32..64.2.6.48.7.6..8723.....73.82..84.2..6...86.74.;11.5/11.1/7.1
.9....6..24....8.....78.......31..65..36.57.8..5.7813.....53.81...1.75.6...86..7.;11.6/11.6/2.6
9.....6..24....8.....78.......31..65..36.57.8..5.7813.....53.81...8.7..6...16.57.;11.6/11.6/7.6


Next chunk 3 000 001 - 4 000 000 has been started with the same parameters. So far the yield of >=10.5 is similar, but the best skfr is only 11.2

[champagne]

Noting special here, but I think that this friendly post is worth to be saved "as it has been written", protected against cyber attacks..
I keep a copy of it in my computer

.
A few comments about the few above posts.

1) SER
First, I've completely stopped using the SER as a rating. It has been useful as a shared measure of difficulty, but it now seems to have too many problems (not stable under isomorphisms, dependent on uniqueness, irrelevant when some pattern is present in a puzzle...).

There's a persistent misconception that there could be a unique rating. Adding the simple tridagon rule to SER would in no way change the problem. I've shown that we'd need a whole set of rules (Trid-ORk-chains or anything similar). But then we'd also need Imp630-ORk-chains and DP-ORk-chains and any chains based on any pattern that could be discovered in the future.

My stance is, there are various ratings, adapted to different purposes. For a puzzle with a tridagon the W+ORk-W rating has proved useful.

2) Tridagons vs T&E(3)
There has never been any "tridagon search hype" - because there has never been any tridagon search at all.
What there has been is a search for T&E(3) puzzles. This has been extremely successful, with mith, coloin, Paquita, Hendrik Monhard... finding millions of them.
More recently, there has been a search for high BxB puzzles. It has also been very successful, with BxB upto 14 (the previous bound was 7 - with only 3 puzzles).
That all these puzzles have a non-degenerate tridagon is an additional result. It says something about high T&E or BxB levels. But no one can measure the real value of this result if they think there has been a search for tridagons.

Very recently, there has been a misguided search for solution grids having a very special type of "tridagsn-solution". As far as I can see by the current results, it's a failure. I'm not saying it wasn't a good idea to try; it was.
But, upon seeing the intermediate results, the failure shouldn't be a surprise:
- more than half of the solution grids have this pattern (implying that the pattern is not selective enough);
- the pattern excludes possibilities of other tridagons in the puzzles (so that the pattern in the solution grid is neither selective nor inclusive enough);
- having a solution grid, all the work of finding minimal puzzles - the hardest part - remains.
.

[coloin]

The chunk 2 000 001 - 3 000 000 has been closed with 14562 puzzles having a rating >= 10.5

As before ...
usually the bxb5 = 11.1-11.3 , and bxb6 11.5-117 , TE3 11.6-11.7 [ without uniqueness ]

This time only 2 puzzles [ the last 2 11.6s] with BxB6, no TE3 puzzles...

One would expect to see a TE3 puzzle expand in one in every 70,000 grids

A vicinity search on one minimal from each of the top 100 expands gave more BxB6 puzzles in 6 furthur grids. no TE3.

Code: Select all

1.........567....3.8.1.3..5...6...9....8..24.8.7......53..18...6....7....7156.... # minimal from original 6
1.........567....3.8.1.3..5...6..9.....8..24.8.7......53..18...6.15.7....7..6.... # minimal from original 6
1...5..8...67.....78...3.6..376.8.....5..1....1.57....3.....9.25.......4...31.6..
.23......45...9...7...23.4.2.4.7.....35.94...9..5..3.......5..8.....76.1.97......
..345.7.......9...7.9.2.5..24..75...3.59.4....9..3......4....1....3.2.....2....68
..34....9.5.7.9..27...2....29..75...3....4....47.3....5.4....9....5..81....3..6..
..3.5.7.9...7.......9.3254....39..75...2.73.4....4..2....5.4...6.....2..81.......
..3.5.7.9...7.......9.3254....39..75....4732....2....4...5.4...81........6....2..