dailysudoku.com

[tcdev]

Would it be possible to indicate target times with the Daily Sudoku Puzzles?

I'm a relative beginner but have, as far as I can tell, progressed to the level where I can solve - eventually - all of the Daily puzzles.

However, I'm fully aware that I'm only starting out, as my rule base is *very* basic and each of my solutions is covered with candidates scribbled into the left-hand side of just about every square!

What I guess I'm saying is that whilst I no longer find just finishing the puzzles to be challenging - I'd like to know how long it should take a moderately experienced solver to finish. That'd give me something to aim for and gauge my progress - and perhaps try to learn more advanced techniques to improve my time.

Anyone else in the same boat?

Regards,
Mark

[David Bryant]

Hi, Mark! Welcome to the forum.

I've been working Samgj's puzzles for almost a year, so I guess I'm getting to be an experienced player. I always try to complete the Daily Sudoku puzzles without making any extra marks (besides completed digits, I mean). I don't always succeed, but usually I do -- on maybe 20% of the puzzles I have to mark the candidate lists in some cells to obtain the solution.

The easy and medium puzzles usually take from 5 to 15 minutes. Hard and very hard are usually more like half an hour. But these times are highly variable, in my experience. Sometimes I notice something important right off the bat and zip right through the puzzle. Other times I overlook something that ought to be fairly obvious, and get stuck for a long time.

In terms of advanced techniques, you really don't need them to do the puzzles on this site. Probably the trickiest thing I've ever noticed in Samgj's puzzles is a "hidden triplet." One device Sam's very fond of is the "hidden pair" -- learning how to spot those will definitely help you solve his puzzles more quickly. dcb

[tcdev]

Thanks for taking the time to reply!

Today's puzzle (medium) took me 24 mins. I'm finding that after a quick scan of rows and columns, I immediately resort to pencilling in candidates - on the most promising squares first. Today after a few such squares a number of others were immediately obvious, so that helped speed things a bit.

The way I look at it, by starting on candidates immediately, there's no need to revisit some squares twice (or more) because you've forgotten something you discovered earlier. I do concede that it isn't the fastest way to complete the puzzle - just the easiest!

However, it would seem that most people place kudos on not using candidates (or other markings) to complete a puzzle. Clearly in order to 'advance' I need to focus more on getting further before I resort to marking up the puzzle. My only tactic thus far would appear to be identifying pairs - but they don't jump out at me (yet) until I start marking up. I guess I should persevere further with this???

Anyone else got some 'typical' times for easy/medium/hard/very on this site?

Regards,
Mark

[David Bryant]

[tcdev]

That's a good tip for the pairs. I've just realised that I'm only considering single squares in isolation, and that I should start looking for pairs to speed up the process.

I found some nicely graded techniques on brainbashers.com. In particular, what they called 'locked sets' - many of which I had come across but not realised I could use them to eliminate other candidates! I took a quick look at XY-wing out of curiosity (the reason I found the site) but my brain started to hurt! ;)

My plan is to learn one new technique at a time and continue practising until it becomes 2nd nature, then move onto the next technique.

Anyway, today's (medium) took me 14 mins which I'm happy with, although I didn't get to use 'locked sets' at all in this one!

[alanr555]

> The way I look at it, by starting on candidates immediately,
> there's no need to revisit some squares twice (or more)
> because you've forgotten something you discovered earlier.

This is a valid concern.
It was in order to reduce the "load" on short-term memory that
I devised the "Mandatory Pairs" system of recording interim
information. It allows the human solver to concentrate on the
logic rather than the memory (although it is advisable that one
remembers some features that are not recorded on discovery).

The difficulty with deriving the candidate profiles at the start is
that it transforms the solution from a logic puzzle to a puzzle
of "spot the patterns in the profiles". Computers need to do it
that way - humans DO NOT!

> It would seem that most people place kudos on not using
> candidates (or other markings) to complete a puzzle.

For some puzzles this is almost impossible (and actually impossible
for the majority of human solvers!). However SamGJ does give us
a grading for each puzzle. From experience, it should be possible
to solve every MEDIUM (or Easy) puzzle without pencil marks. The
archives have many examples and spending some time on the
Medium puzzles should enable the honing of one's logic skills.

For the Hard/V.Hard, about half can be solved using only Mandatory
Pairs whilst the remainder need some assistance from Candidate
Profiles in order to reveal their secrets.

One technique that I am exploring currently is what I call "Missing"
profiles. This involves writing at the top of each column and to the
right of each row the set of digits not already resolved in the col/row.
However, there is a slight enhancement in that the objective is to be
able to split that set of digits into subsets of pairs/triples/quartets etc
(which I distinguish by surrounding each subset by parenthesis).

This makes it much easier to compare the sets to narrow down the
possibilities for intersecting cells - without making pencil marks - and
to test for "sole position". The latter is the situation where there is only
ONE cell in a line where a particular digit can be placed as all the other
unresolved cells are impossible for such digit. I have not found any
other system (other than scanning candidate profiles) that has an easy
method of finding 'sole position' cases - less easy to find than the
"sole candidate" cases.

In another post our Colorado friend has suggested testing rows/cols
when the number of resolved cells reaches five. I tend to work first
with the higher numbers. Eight filled forces the ninth. Seven filled
very often reveals the other two and six filled can be productive
(sometimes if only to expose a pair/triple more easily) and whilst
five filled does not reveal significant information much of the time, it
does do so more often than one might expect - occasionally solving
one cell which leaves a six-filled which in turn leaves a seven-filled
and eventually a completely resolved line. Thus it is worth scanning
from five-filled - and even for four-filled if the rest of the techniques
have not revealed anything useful. The use of the "missing" profiles
at the edge of the grid does speed up these checks - and also the
setting of candidate profiles eventually if they become essential.

My standard method for Hard/V.Hard is now to start off with Mandatory
Pairs. If this gets stuck, I progress to Missing Profiles and only after
exhausting that do I move on to Candidate Profiles. Of course, this
will never be a 'clock-beating' approach - but it retains the satisfaction
of relying upon "logic" as long as possible before resorting to pattern
recognition.

Quite often, it is possible to revert to one of the earlier methods once
Candidate profiles has revealed what was needed to "break the back"
of the puzzle. When I remember, I use a different coloured pen for
each stage - minimising the reliance upon candidate profiles.

+++

Many of us may feel compelled to challenge ourselves against
the clock. However, this is an artificial challenge as there can never
be a "standard" time for a puzzle and no two puzzles can be said
to have the same "expected" time - even for the same person. The
world champion arising from one puzzle may have great difficulty
with another and even averages over a range of puzzles would not
be a totally equitable comparison.

Really, the only satisfactory challenge is to complete the resolution
of the puzzle - however long it takes - by working the logic. It may
be fascinating to predict how many "easy" puzzles one can do in a
supermarket check-out queue before having to give attention to
unloading the groceries but that, surely, is not the real point. The
real challenge is to reach the solution without errors and using the
minimum amount of written assistance. Incidentally, candidate profiles
seem, to me, to be inordinately error-prone - another reason to avoid
deriving them until absolutely necessary!

> In order to 'advance' I need to focus more on getting further
> before I resort to marking up the puzzle.

There is the approach of using the grade to decide when/how to
mark up a puzzle.

> My only tactic thus far would appear to be identifying pairs
> - but they don't jump out at me (yet) until I start marking up.

Use of the Mandatory Pairs recording system can reveal pairs
fairly painlessly - less painfully even than deriving Candidate
Profiles, as they emerge as part of the processing rather than
by human fallible scanning of strings of digits.

> Anyone else got some 'typical' times for easy/medium/hard/very
> on this site?

No, they are irrelvant - as postulated above.

+++

Alan Rayner BS23 2QT

[samgj]

[rgbrown]

I certainly wouldn't go as far as calling target times 'irrelevant', but I have to agree that they're not hugely helpful. Not only is the variance huge among solvers of a given puzzle, but the variance for the same solver of puzzles of a given difficulty is also quite large. Also, while we're all essentially using the same tool -- logic -- to solve the puzzles, each of us has a different set of techniques and method of employing them, so a puzzle that plays well to your way might keep me stymied for a long time, or vice-versa. Also, I find that, unlike solving crossword puzzles, solving sudokus relies heavily on one's instantaneous brain state. By this, I mean that how sharp you are while working one, relative to another day or time of day or circumstance, is a lot more important. I find, for example, that I frequently have to spend a fair amount of time early in the day, finding out why I can't get a difficult monster to come together, when I started it late the night before. Somehow, in my stupor, right before my clipboard falls out of my hand, I managed to enter a '3' in the wrong row, or perhaps an 'M' in the wrong column, in an extreme (and probably fictitious) example. Also, I find it much more difficult to make headway when, for example, the rabbit is out of his cage, and I have to make sure he doesn't chew through all of the cords on the back of the computer. You know, when, "3, 5, 7, 9, B, E, F" becomes, "Get out of there! OK, that was 3, 9, B, oh, drat!"

Also, remember that while difficulty can be relative, relatives can be difficult.

Russell