Solving for speed

Everything about Sudoku that doesn't fit in one of the other sections

Solving for speed

Postby 999_Springs » Sun Mar 29, 2009 5:51 pm

250th post!

I don't think that solving against the clock is something that has been discussed at length in the last few years. But since all solving competitions just about only involve solving for speed, I thought I would bring up this topic.

When people solve against the clock, I suppose that they all start with cross-hatch scanning for singles, and perhaps locked candidates, until they get as far as they can. There is going to be a point where people start deciding to search for naked singles, and later start to clog up the grid with pencilmarks. However, pencilmark-writing is a tedious process, and so it might not be ideal to fill up the entire grid with pencilmarks at the same time, perhaps only one part of the grid to see what is happening, then another. How would you do this?

As the puzzles get harder, it would be faster to at some point abandon logical techniques altogether and just guess one or both ends of a bilocation/bivalue cell. Obviously, this isn't how puzzles are supposed to be solved, but since you have a horrible constantly ticking reminder beside you, you'll just have to guess and see how that comes out. Of course, I tend to avoid solving not-all-singles puzzles when it comes to speed solving, as it gets rather annoying. However, what would you say be the point at which you start guessing, and would you guess one end of a bivalue, or both simultaneously?

And is timed solving something that needs to be practised frequently in order to keep up one's times, or can one have a three-month break and expect to solve as quickly as before? (I've just had a three-month break...)

I have found a bunch of puzzles which I am interested in seeing how you would solve them against the clock. Any feedback appreciated.
Code: Select all
..5|7..|...
.1.|93.|7..
..6|...|..4
---+---+--
..8|...|.5.
...|.9.|...
.5.|.7.|..2
---+---+--
.3.|...|.4.
.87|...|...
.4.|.68|31. (tarek)
all singles, 39 steps needed to solve

.4.|731|.9.
1..|...|...
9.7|8..|...
---+---+---
486|357|912
752|914|836
319|268|547
---+---+---
5.3|182|4.9
..1|...|...
.94|673|.5.
4.0 puzzle from "maximum number of clues..."

...|.47|296
9..|.8.|457
47.|5.9|831
---+---+---
..4|...|789
.9.|..8|164
...|.94|523
---+---+---
549|876|312
638|412|975
721|935|648
8.3 puzzle from "maximum number of clues..."

48.|3..|...
...|...|.71
.2.|...|...
---+---+---
7.5|...|.6.
...|2..|8..
...|...|...
---+---+---
..1|.76|...
3..|...|4..
...|.5.|...
#4 from the top95

.6.|5.1|.9.
1..|.9.|.53
9..|..7|...
---+---+---
.4.|8..|.7.
...|...|5.8
.81|7.5|.3.
---+---+---
...|.5.|2..
...|...|...
.76|..8|...
#12 from the top95 (argh!)

.45|.67|.89
6.7|8.9|4.5
89.|45.|67.
---+---+---
4..|578|.96
5.6|.94|7.8
789|2.6|54.
---+---+---
968|7..|.54
.54|68.|9.7
17.|945|86. (Eioru)
Time taken to rate puzzle using SE: 5 MINUTES AND 22 SECONDS!!!
Hidden Text: Show
Once upon a time I was a teenager who was active on here 2007-2011
999_Springs
 
Posts: 367
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

Postby StrmCkr » Mon Mar 30, 2009 10:33 am

i speed solve for fun,

i find the best and time reduceing is to focus a group of moves off the same locals that way you dont have to write anything down and its all one chain. thus the knowledge to remember is the cells you started with and what caused it.

but if needed all you need to write in is the key digits used to make the move work in the first place that way later process simply delete or removed parts.

mind numbing easy puzzle can be made difficult as well if you wish it to.

dont look for singles any more, find the quads or combination moves that make a quad (2 pairs, tripes+single):

these are key for any puzzle in timed matches to beat the clock.

idenitfication of ur patterns is also huge and time saving.

(x-wings locked pairs etc are idenifiers)

heres my solutions using this.

Code: Select all
..5|7..|...
.1.|93.|7..
..6|...|..4
---+---+--
..8|...|.5.
...|.9.|...
.5.|.7.|..2
---+---+--
.3.|...|.4.
.87|...|...
.4.|.68|31. (tarek)

 all singles, 39 steps needed to solve


heres a rapid solution: (4 steps i use the ur as 1)

naked triple (125) r7c45,r9c4 =>
R8C5<4>, R8C4<3>, R8C6<9> R8C1 <1>

xyz wing
259: r7c3,r9c14 =>
R9C3<9>, R7C3<2>,R9C1<5>,R9C4<2>,R7C1<6>,R1C3<4> R2C6<6>,R1C6<4>,

hidden pair (28)R2C18 =>
R2C9<5>, R8C9<6>, R8C8<2>,R8C7<5>,R5C7(7),R2C8<8>,R2C2<2>,R1C2<9>,R3C2<7>

(note this is 3 corners of a ur)
naked pair (13)R56C3 => R5C1<4>,R6C1<9>,R4C1<7>
naked pair (13)R6C36 => R6C8<6>,R1C8<3>,R3C8<9>,R1C9<1>,R3C7<2>,R1C7<6>,

from here out its all a short singles chain.
want a solve time too? its around (1:43min )
Last edited by StrmCkr on Fri Apr 03, 2009 9:08 am, edited 14 times in total.
Some do, some teach, the rest look it up.
User avatar
StrmCkr
 
Posts: 647
Joined: 05 September 2006

Postby StrmCkr » Mon Mar 30, 2009 11:02 am

Code: Select all
.4.|731|.9.
1..|...|...
9.7|8..|...
---+---+---
486|357|912
752|914|836
319|268|547
---+---+---
5.3|182|4.9
..1|...|...
.94|673|.5.
4.0 puzzle from "maximum number of clues..."


ur: (58) R12C3 +(58)R12C9 => R3C9<5>

singles to the end. ( < 30 seconds)

Code: Select all
...|.47|296
9..|.8.|457
47.|5.9|831
---+---+---
..4|...|789
.9.|..8|164
...|.94|523
---+---+---
549|876|312
638|412|975
721|935|648
8.3 puzzle from "maximum number of clues..."


"paired propagation." R1C12 possible pairs 8+1,

R1C12 {18} => (532)R123C3 + (6)R2C2 + (32)R12C4 => R5C34 <7>{contardiction)
R1C4 <1>, R2C2<1>, R1C12<>1
singles to the end. (<45 seconds)
Some do, some teach, the rest look it up.
User avatar
StrmCkr
 
Posts: 647
Joined: 05 September 2006

Postby 999_Springs » Tue Mar 31, 2009 6:59 pm

StrmCkr wrote:heres a rapid solution: (4 steps i use the ur as 1)

naked triple (125) r7c45,r9c4 =>
R8C5<4>, R8C4<3>, R8C6<9> R8C1 <1>

xyz wing
259: r7c3,r9c14 =>
R9C3<9>, R7C3<2>,R9C1<5>,R9C4<2>,R8C1<6>,R2C3<4> R2C6<6>,R1C6<4>,

hidden pair (28)R2C18 =>
R2C9<5>, R8C9<6>, R8C8<2>,R8C7<5>,R2C8<8>,R2C2<2>,R1C2<9>,R3C2<7>

(note this is 3 corners of a ur)
naked pair (13)R56C3 => R5C1<4>,R6C1<9>,R4C1<7>
naked pair (13)R6C36 => R6C8<6>,R1C8<3>,R3C8<9>,R1C9<1>,R3C7<2>,R1C7<6>,

from here out its all a short singles chain.
want a solve time too? its around (1:43min )

Wow. How do you locate a naked triple and xyz-wing in a position like that in less than 2 minutes!? Also, with regard to your pair in r2, there are only 3 empty cells left in it when you found it. So 5r2c9 was just a single in r2!

StrmCkr wrote:ur: (58) R12C3 +(58)R12C9 => R3C9<5>

I see how that gives r2c9<>58 but I don't see how that forces r3c9=5. Am I missing something here?

StrmCkr wrote:singles to the end. ( < 30 seconds)

How do you fill in that grid at the rate of over 1 cell per second!?

Lastly, what would you recommend that I do when I get the staring-at-the-puzzle-without-being-able-to-find-anything feeling (is it called solver's block?) while doing a Times puzzle?
Hidden Text: Show
Once upon a time I was a teenager who was active on here 2007-2011
999_Springs
 
Posts: 367
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

Postby StrmCkr » Wed Apr 01, 2009 12:27 am

hidden pair (28)R2C18 =>
R2C9<5>, R8C9<6>, R8C8<2>,R8C7<5>,R2C8<8>,R2C2<2>,R1C2<9>,R3C2<7>

.i like the pair a bit better it back fills the row easier. plus other bonuses.

every move i used stem from the same regions.

its a focused search ignoring the easy stuff (singles) unitl im in the same box.

a slightly more complex move will still solve them.

i use a counting:
tripes and quads
. start with the box reduced the clue that cant form a triple or quad.
then count then its a quick scan of intersections.

box 8.
has the most amout of differnt clues interscting the box.

25 is missing and the key set to check.
9 combinations less givens.
which is only 1+25,

from here i checked box 7 and 9:
if the triple i found could be bent, (xyz){to bend it i used the bivavle in r9C4) (which requires at least 1 cell to work in box 7){few to check}

the hint for this is 7's are singles: in box 7 and 9... which leaves the bent pair option as box 7.

now the really easy way to find the triple was looking for identical given clues intersections:

with the case of 125,

R9C4 has givens 1346789 covering it.
any other locations that have 3467891

or spots with all but 1 of these digits(has to be the same in both cases) are also the triple.

3479 are on rows and columns with 68 being in the box. your looking for 4 given clues.

which are seen in R7C45

so all three locations are covered by the same givens.
346789 => 125 naked.

ur: (58) R12C3 +(58)R12C9 => R3C9<5>

the real way to use this is R2C9 <59>

next:
me i find the best way to avoid the rectangle was to complelty destroy it useing the limtis of the column. and box.

which was a good guess as either the 8 or 5 breaks the ur... only 1 is correct.

problably a wrong application i'll double check it after work. yip wrong application.{ i pretty much guessed.}


for me that point where your stuck is to look at a limit of sequences for a digit. if its two points in a box on the same row.
i'll attempt to form pairs from the pms.
this can create a quick backdoor. or remove lots of incorrect moves.

i call that paired propagation and i sorta show it in the third example.
i havent released much deatails on it as its kinda like trial and error and not at the same time. but im not going to argue that it isnt either.

i have a few users on here assisting with the development to prove it out befor i write too much on it.

How do you fill in that grid at the rate of over 1 cell per second!?


try and loop the placemetns off the same move that way more clues are removed and less referenceing(checking) is required.
basically a flow.
..but its random if it works or not...
. just like if you happen to find the speed moves or not.

i think speed solving should allow the user to preview the puzzle for a few mins.
set it down. then go like mad.
just like they do for speed rubix cubing.
takes away the random part of trying to figure out what approach works best. and gives you a few mins to ponder on the appoarch.

rather then blind solving.
Some do, some teach, the rest look it up.
User avatar
StrmCkr
 
Posts: 647
Joined: 05 September 2006


Return to General