The New Sudoku Players' Forum

[mith]

Having only 1 guardian is NOT the most favourable case, as it allows one placement and the pattern becomes as inexistent. Tridagons with more guardians are often much more useful to reduce the puzzle complexity, because they allow the use of chain rules based on them.

The context is "a manual solver having heard of the TH pattern can solve it easily". Having a single guardian is certainly no guarantee that the puzzle will be "easy" afterward, but it does guarantee that someone knowing about the pattern is going to be able to trivially place that digit. (All such placements currently known do immediately reduce the T&E depth, of course.)

That said, I'd be interested if you have stats on guardian count vs. your rating.

Likewise, the T&E depth itself just tells us that these puzzles resist a T&E process.

I have a strong objection to this.
The T&E-depth can't be compared with SER, the value of which is the result of obscure computations that have never been described in plain English.
On the contrary, being strictly in T&E(BRT, n) means the puzzle needs to use contradictions arising from n candidates when only Singles are used to get such contradictions. This is clear structural information about the puzzle.

My point is only that "hard" always requires a set of rules to measure against (in the context of "the highest SE rating... for a puzzle that is in 4-template"). I'm not equating the "worth" of T&E depth vs. SER. (And I agree it would be valuable to have a clearer explanation of how SE works.)

As it currently stands, knowing a puzzle has depth 3 implies a lot more about the puzzle (since all known examples have the trivalue oddagon pattern); but we don't know that will always be the case. I'm not sure knowing a puzzle is depth 2, say, tells us much more with respect to another rating system or human solvability than that it is depth 2 for the chosen set of resolution rules. (Depth 0 is of course different; pretty much any reasonable rating system is going to rate a singles-only puzzle as easy.)

[JPF]

Hi mith,

Thank you for posting your database of 11.8+, depth = 3.
Here's a batch of 11.8+ minimal puzzles that I didn't find in your database. However, I don't have the tools to test the depth.

The puzzles are in gsf-minlex , PGXExplainer format.

minimal puzzles 11.8+: Show

Code: Select all

...............123....45..6.......3...6.172.828..6..17.127..3...7.....8.3.86..7..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.216......7....8..3.87...6.   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.216.....3.87.....67....8..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.217.....3.86...7.67....8..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.217...6..7....8..3.86...7.   ED=11.8/1.2/1.2
........1.......23...2456.......6.....247.8..45.8..7....47.8.6..7652....28..64...   ED=11.8/1.2/1.2
........1.......23...2456.......6.....247.8..45.8..7.6..47.8....7652....28..64...   ED=11.8/1.2/1.2
........1.......23...2456......56.....24.78..45.8..7....478..6..765.2...28..6....   ED=11.8/1.2/1.2
........1.......23...2456......56.....24.78..45.8..7.6..478.....765.2...28..6....   ED=11.8/1.2/1.2
........1.......23...2456.....6.4.....275.8..45..8.7....5.78.6.7.64.....82...6...   ED=11.8/1.2/1.2
........1.......23...2456.....6.4.....275.8..45..8.7.6..5.78...7.64.....82...6...   ED=11.8/1.2/1.2
........1.....2.....3....24....5.....2.67...88.62.37...657.....3.28.....78..36.5.   ED=11.8/1.2/1.2
........1.....2.....3....24....56.....72.38...2.78...6.758.....3.26.....86..37.5.   ED=11.8/1.2/1.2
........1.....2..3245..........67...6..84..7.8..2.5....8....6..47...68.55.6..82.7   ED=11.8/1.2/1.2
........1.....2.3.....3.4.5..261.....17.23.8.63.8.7.1..61.8....2.3.76..878.......   ED=11.8/1.2/1.2
........1.....2.3.....3.4.5.16.23.7.7.281....83...6.1..81.7....2.3.68..767.......   ED=11.8/1.2/1.2
........1.....2.3..45...6......45.....27..8...6.28.7...54.68.7.7.642...88..5.7...   ED=11.8/1.2/1.2
........1.....2.3..45...6......45.....627.8...2.8..7...54.67.8.7..5.8...86.42...7   ED=11.8/1.2/1.2
........1.....2.3..45...67.....45.....28..7...6.27.8...54.6..8.7....8...8.642...7   ED=11.8/1.2/1.2
........1.....2.3..45...67.....45.....28..7...6.27.8...54.67.8.7....8...8.642....   ED=11.8/1.2/1.2
........1.....2.3..45...67.....45.....627.8...2.8..7...54.6..8.7....8...86.42...7   ED=11.8/1.2/1.2
........1.....2.3..45...67.....45.....627.8...2.8..7...54.67.8.7....8...86.42....   ED=11.8/1.2/1.2
........1.....2.34....56.....1.257...5.6.18..6.278.1...16..7...52..68..78.7......   ED=11.8/1.2/1.2
........1.....2.34....56.....1.257...62.8.1..57.6.18...87......1.6..7...25..68..7   ED=11.8/1.2/1.2
........1.....2.34....56.....56.17...17.258..62.8..1...61..8...5.2.67..878.......   ED=11.8/1.2/1.2
........1.....2.34....56.....762.8...867.51..25..817...25.68..761...7...7.8......   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8..7.12.6.7..5.72..8..68..751..   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8..7.1726.8..5.28..7..68..751..   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8..7.8726.1..5.28..7..6...758..   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8.17.12.6.7..5.72..8..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8.17.1726.8..5.28..7..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....2..1.....65..8.17.8726....5.28..7..6...758..   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72........1.26.8..5.28.17..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72........12.6.7..5.72.18..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72........8.26....5.28.17..6...758..   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72..1.....1.26.8..5.28..7..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72..1.....12.6.7..5.72..8..68..75...   ED=11.8/1.2/1.2
........1.....2.34....562....1..7....65..8.1.72..1.....8.26....5.28..7..6...758..   ED=11.8/1.2/1.2
........1.....2.34235...6....6.......7..258..82.3..7...87....6.35.6.8...6.2.37...   ED=11.8/1.2/1.2
........1.....2.34235...6....7.258...823..7..56........5.6.8...62..37...7.8....6.   ED=11.8/1.2/1.2
........1.....2.34235...6...27.3.8...8.2.57..5.6.......53.67...6.23.8...87.....6.   ED=11.8/1.2/1.2
........1.....234..23...5.....23......65.78...5...87...62.7....7...23.6883.6.5...   ED=11.8/1.2/1.2
........1.....234..23...5.....23......65.78...5...87...62.8....73.6.5...8...23.67   ED=11.8/1.2/1.2
.......12.....1..3.45...6....6.17....875...6.4..6.8....7..548..6.4......85.1..7..   ED=11.8/1.2/1.2
.......12.....1..3.45...6....6.17...41.6.8...8.75...6..64......58.1..7..7...548..   ED=11.8/1.2/1.2
.......12.....3..4.56...3....4.3.....7.4......8.65..47.67..58..3.578....8..3.67..   ED=11.8/1.2/1.2
.......12.....3..4.56...3....4.3.....7.4......8.65..47.67.85...3.574.8..8..3.67..   ED=11.8/1.2/1.2


JPF

[mith]

Here's a batch of 11.8+ minimal puzzles that I didn't find in your database. However, I don't have the tools to test the depth.

The following are depth 2:

Code: Select all

...............123....45..6.......3...6.172.828..6..17.127..3...7.....8.3.86..7..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.216......7....8..3.87...6.   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.216.....3.87.....67....8..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.217.....3.86...7.67....8..   ED=11.8/1.2/1.2
...............123....45..6.......3...6.27.1818..6.2.7.217...6..7....8..3.86...7.   ED=11.8/1.2/1.2
........1.....2.....3....24....5.....2.67...88.62.37...657.....3.28.....78..36.5.   ED=11.8/1.2/1.2
........1.....2.....3....24....56.....72.38...2.78...6.758.....3.26.....86..37.5.   ED=11.8/1.2/1.2

The rest are in the database, but with expanded forms rating at 11.7. I'd post a list of all minimals of the 11.7s, it would be something like 500k puzzles I'll just wait until I get the full update ready.

[1to9only]

Code: Select all

java -Xrs -Xmx500m -cp PGXplainer.jar sudoku.serate --input=test-in
........1.....2.34.35.......62......7.3.25..885...67...7853....32..68...5.62.7.8. ED=11.9/11.9/3.4
........1.....2.......3..45.16.23...27.81.6..3.87.61...32.67..81.7.8....68....... ED=11.9/11.9/3.4
........1.....2.......3..45.16.23...27.8.61..3.871.6...32.68..718..7....6.7...... ED=11.8/11.8/3.4

PGE gives the same results as above. The 3 Loki-related puzzles are in minlex form, so to SE (PGE) they are different puzzles.

The lksudoku fix is to address (original) SE mis-rating the same puzzle or morphs of the puzzle as observed in the Patterns Game.
E.g. I got an se-mismatch is the Patterns Game for this puzzle. Put the puzzle multiple times in a file, and (original) SE gives these ratings:

Code: Select all

...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.3/10.3/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.3/10.3/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.2/10.2/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.3/10.3/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.2/10.2/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.2/10.2/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.3/10.3/3.4
...6..4...3.....5...5.7...98..4.6.....2.9.......2.71..4....18...5......3..9....7. ED=10.2/10.2/3.4 ED=10.3/10.3/3.4


[mith]

Code: Select all

java -Xrs -Xmx500m -cp PGXplainer.jar sudoku.serate --input=test-in
........1.....2.34.35.......62......7.3.25..885...67...7853....32..68...5.62.7.8. ED=11.9/11.9/3.4
........1.....2.......3..45.16.23...27.81.6..3.87.61...32.67..81.7.8....68....... ED=11.9/11.9/3.4
........1.....2.......3..45.16.23...27.8.61..3.871.6...32.68..718..7....6.7...... ED=11.8/11.8/3.4

PGE gives the same results as above. The 3 Loki-related puzzles are in minlex form, so to SE (PGE) they are different puzzles.

Here are the first and third in solution_minlex form (the first is just the max-expand of the third; it has one extra given):

Code: Select all

1.3.56....571.9.3.69.37......1.93..75.96.73.....51.......96..........4....5...86. ED=11.8/11.8/3.4
1.3.56....571.9...69.37......1.93..75.96.73.....51.......96..........4....5...86. ED=11.8/11.8/3.4

So, still a morph dependency on that first one. Maybe this is a different issue though.

[JPF]

The rest are in the database, but with expanded forms rating at 11.7

Ah, I hadn't realized that your database was for 11.8+ regarding both minimals and expanded forms at the same time.
I can see that your exploration is already quite extensive!

JPF

[Paquita]

Mith,
I am really looking forward to your publication of 11.7 puzzles - and after that, of 11.6. I noticed that if I found a puzzle and you did find the same, you now publish them with your name. I don't mind really but it makes all my current work futile, when I can't tell if the ones I find are in your collection as well.
I prefer to add new puzzles to the total collection. So I need to know which are really new. And in what directions I should search that have not been covered by you.

I do have great respect for what you have done. You found so many new puzzles, you have discovered the T&E(3) collection and increased the knowledge of hard puzzles. Now I wonder if there are many more unknown puzzles out there to be found, or whether your search has been exhaustive.

I still come across puzzles that I think are new. But I think I better wait before I publish any more puzzles.

[denis_berthier]

I'd be interested if you have stats on guardian count vs. your rating..

The closest thing I have are the tables in section 14.12 of the User Manual.
.

[mith]

Yeah, I understand it's a problem; as far as credit, what I've been doing is:

For minimals of puzzles already published in an expanded update, I consider those already found - minimals are easy to generate and this is the reason I haven't published them in the past.
For puzzles added to the database between updates, if someone posts a minimal I will update the credit accordingly even if I have already generated the puzzle, since I haven't published it. (I've done this with a couple of JPF's recently, but I'll scrape the thread at some point to make sure I'm not missing any.)

As far as overlapping work, perhaps with the coming update I will include which scripts have been run on the minimals. The current stats look like this:

1112250 expanded forms. [edit]updated after removing duplicates; really need to find this bug but it's a quick check[/edit] All minimals of these are found and added to the minimal database. All of these puzzles also have been checked by the depth adder script (try adding a given from solution and check if the depth is still 3; this is how the max-expands are generally found, including the current 79 40c depth 3 puzzles) and transformer script (change isolated cycles of givens, which can't affect the depth; this was implemented based on a suggestion by 999_Springs), and are rated by serate.

4634322 minimals. Of these:
4112144 have been searched at {-2,+1}
100585 have been searched at {-1,+2}
37643 have been searched at {-2,+2}

Whenever a puzzle is found by these scripts (exhaustively generated with gsf, don't have to be minimal, just valid and not depth 0), I singles-expand it, check for depth, gsf-minlex, and then check against the expanded database to see if it's new. Reducing is much faster so I've focused more resources there; eventually it may be a criteria for posting an update that the reducing search is done on every puzzle, but I'm not quite there yet. Currently, preparing for an update means stopping these scripts on the minimals and letting the scripts on the expanded forms catch up (it's always the serate script that is last to finish).

I will also aim to provide some scripts on github to make it easier for everyone to check puzzles against the existing database.

The rest are in the database, but with expanded forms rating at 11.7

Ah, I hadn't realized that your database was for 11.8+ regarding both minimals and expanded forms at the same time.
I can see that your exploration is already quite extensive!

JPF

It's actually for minimals of 11.8+ expanded forms; a few of those minimals rate below 11.8 because of SE quirks!

When I publish a separate SER-based update, I will have to denote which ones have been rated as expanded forms and which haven't, or re-run the whole batch (I'd rather do this only if the morph dependency issues are addressed, it's going to be a lot of CPU time). At this point the minimals from 11.6+ depth 3 expanded forms outnumber the other 11.6+ minimals by about 100 to 1. (Though, there are likely more morph or SE version issues like this in the existing ph2010 database anyway; see here for one.)

[mith]

I'd be interested if you have stats on guardian count vs. your rating..

The closest thing I have are the tables in section 14.12 of the User Manual.
.

Thanks, I'll check it out later.

[hendrik_monard]

Hendrik,

Your other puzzles from the 11.9's list do rate 11.8, maxlex, PGE, only this one remains 11.9. I wanted to mention it as the 11.9's are rare and ususally get a bit of attention when one is found. And it should be a 11.9 in the database. (The tenth!)

I apologize for the doubles in my list. I made a mistake in what I did publish already.

Hi Paquita,

With all due respect, AFAIK the current (informal) rule for the rating of new puzzles is that the SER, calculated on a new puzzle as first published, is accepted and should not be changed afterwards.
Of course, rules can change but this should be discussed before they are applied. And if the rules are changed, they are not retroactively applied.
You propose that the SER should be calculated on the maxlex isomorph of a puzzle. This seems to be very arbitrary. The minlex or maxlex isomorph have only meaning for the detection of the novelty of a new puzzle. There is no reason to impose the SER calculation on either of them.
You should also be aware of the fact that your rule would also disqualify 'loki', mith's 11.9.

THIS BEING SAID, I understand your anxiety that rating 11.9 is rare and should not be accepted too easily. Perhaps one could consider to add, only for the 11.9 group, an additional rating that is morph independent. I am thinking of the BpB system of Denis. Anyway, I do not expect that there will be an explosion of the number of 11.9s.
Opinions on this issue from other members, especially those with experience from the early times of this thread, would certainly be most welcome.

[Paquita]

Hendrik

I did not understand why your new 11.9's were barely noticed, congratulations are in place! There can be a dispute about some being 11.9 or 11.8. I did not mean to set a standard, I just mentioned that not all of the new 11.9's are beyond that dispute, and 11.9 in all ratings. So, ok, accepting that the publisher of the 11.9's decides what to use, minlex, maxlex or some other form, and SER rating, you doubled the number of 11.9's available and that is a big thing. I did put them in my list of puzzles but it seemed they were barely noticed by other forum members, not in any 11.8 plus list of new puzzles. I remember that mith said something in some post quite a long time ago about that he rated a 11.9 from you as 11.8 and therefore added it as 11.8. So I thought, well maybe your 11.9's are taken as 11.8s and everyone is ok with that. And in that case, I just wanted to say that at least one should be rated 11.9 anyway, and indeed, miths puzzle would then also be an 11.8.

[Paquita]

Mith,

Fair enough. On second thought and after what you said I changed my mind about waiting to publish. I am publishing what I have from the last days since you published your 11.8 list.
https://docs.google.com/document/d/e/2PACX-1vS3OM2cAJUsObb6TrSr24sMdv49D3lbnXu1qI1n7H10-qG05698T0FjQQA6F72xzGxZe6IQ8BX5CJto/pub

A modest contribution. I rated these with skfr (I found the source and recompiled it, great tool!). I will publish updated ratings later when I have used PGE or SER - that is one of the time consuming processes indeed. (Another very time consuming process is to minimalize puzzles with lots of clues. I wrote my own minimizer but it takes a long long time to minimalize any puzzle beyond 32 clues, especially if there are a lot of singles added. Anyone have a tool that can do it faster?)

And this is the 11.8 part of that file. I think they are new, but these days, I am not so sure...

Code: Select all

987......6..54.............37.82.6..2.86...73.9.........3....62.....67.8.....239. ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...239...3.....2.....67.8 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...27.8.2...639...3.....2 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73..9.......9...27.8.2...639...3.....2 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...6........293..7..67....39.3.8...2.7.92....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....39.3.8...2.7.2.....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....29.2.8...3.7.3.....68 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...39.37....2....2....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...923..78...6......6.8...39.37....2...29....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...29.27....3....3....67 ED=11.8/1.2/1.2
98765.......4..........3...8.57.692..9.2..7...2.....686...2.89.2...6..57.5....... ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9.........3....62.....67.8.....239. ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...239...3.....2.....67.8 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...27.8.2...639...3.....2 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73..9.......9...27.8.2...639...3.....2 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...6........293..7..67....39.3.8...2.7.92....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....39.3.8...2.7.2.....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....29.2.8...3.7.3.....68 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...39.37....2....2....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...923..78...6......6.8...39.37....2...29....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...29.27....3....3....67 ED=11.8/1.2/1.2
98765.......4..........3...8.57.692..9.2..7...2.....686...2.89.2...6..57.5....... ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9.........3....62.....67.8.....239. ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...239...3.....2.....67.8 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73.9........6...27.8.2...639...3.....2 ED=11.8/1.2/1.2
987......6..54.............37.82.6..2.86...73..9.......9...27.8.2...639...3.....2 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...6........293..7..67....39.3.8...2.7.92....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....39.3.8...2.7.2.....68 ED=11.8/1.2/1.2
9876.........958.......4...8.32..6...923..7...6.......67....29.2.8...3.7.3.....68 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...39.37....2....2....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...923..78...6......6.8...39.37....2...29....67 ED=11.8/1.2/1.2
9876.........958.......4...73.2..6...293..78...6......6.8...29.27....3....3....67 ED=11.8/1.2/1.2
98765.......4..........3...8.57.692..9.2..7...2.....686...2.89.2...6..57.5....... ED=11.8/1.2/1.2

[JPF]

Hi Paquita,

For information, 1167 puzzles from your 1751 list are duplicates.
To obtain the minimal puzzles of a puzzle, gridchecker by Mladen Dobrichev is an excellent tool.

JPF

[denis_berthier]

Denis I am using your program to rate my previous 11.8 puzzles from august 13th. It takes a long time, .

I'll soon (in a few days) release an improved version of:
- the T&E(n, n=1,2,3) that allows some gain in resolution time (30 to 50%) by allowing to use a precomputed solution (obtained e.g. by DFS) in order to avoid testing some candidates in it ;
- new "solve-file-knowing-solutions" functions that will allow to do this for full files of puzzles.

Done.
See http://forum.enjoysudoku.com/csp-rules-sudorules-kakurules-t38200-90.html for details