You are here

Sudoku round 6, Liar sudoku


I think that the layout of liar sudoku isn't convenient for solving.
Is it possible to make the clues much smaller and to place them in a corner of the cells?
Big wrong clues taking the whole place in the cells make trouble to me while solving.


Hi Fred76,

I am sorry that you find the presentation inconvenient. It is based upon the same presentation as seen for the potentially false clues in "Pinocchio Sudoku" as seen in the 2012 World Sudoku Championship. It is important to note that most of the clues are not lying and so do provide useful information - it might well be that this information is harder to see if presented as smaller clues in the corner of cells.

Unfortunately I am unable to change the presentation for the CB. If there is further feedback from other solvers then of course I'll review this next time a similar variant comes around.

I'm sorry it is not possible to be of any assistance to you this time,

Tom Collyer
Competition Director,
WPF Sudoku GP 2016

For what its worth, I would prefer the presentation Fred suggested from the next time. Didn't really think about it much beforehand which is a lack of preparation on my part, but this was inconvenient. In Pinocchio there are just 3-4 potential liars to keep track of, whereas here, while you are correct that the digits will mostly be proven true and useful, its still a lot of potential liars to keep track of while solving.

I have some experience in this as we are preparing a pinocchio contest that will be held on LMI. Basically, there are 3 possible designs for pinocchio/liar clues: grey digits (or different color in general), contour digits, small digits in the corner.

For pinocchio givens I always used grey numbers of the same size and font as the standard givens as it is very easy to design it in Excel. Even though I always tried to make it as brightest as possible, quite often I had troubles in testing as I confused a pinocchio given with a standard given or a filled digit. Thats because I use grey pencil for playing and I think most other players use it too.
It is not necessarily true that in Pinocchio sudoku there are just 3-4 potential liars to keep track of. If you use all 9 digits for pinocchios, then you have 27 pinocchio givens (in the competition Liar Sudoku there was 37 clues).

Though it was inconvenient for me to play pinocchio puzzles with grey clues, I had no problem with the graphics of this GP round. When the middle of the digit is white, it is comfortably distinguishable from non-pinocchio digits and when you blacken it with pencil it looks different (when you have a grey pinocchio clue and you blacken it with pencil it still looks quite the same).

There are three important things to note:

1) In Pinocchio Sudoku, after you fill the cell that contained a pinocchio clue, you need to know what was the original digit as it has impact on the other two digits from the pinocchio set. In Liar Sudoku, the original digit has impact on the respective row, column and box. And all you need to know is whether the clue was correct or not and you can easily mark it for example by circling a false clue and you can ignore the clue that was proved correct.

2) In Pinocchio Sudoku, the clues of the same set (e.g. 111) are usually placed close to each other or follow a symmetric pattern so you dont have a problem locating all clues of the same set (hence you dont need large digits and small clues in the corner may suffice).

3) In Liar Sudoku, the value of a clue is sometimes not the most important information it gives. Since there is exactly one false clue in every row, column and box, the location of the clue may be more important as well as the location of all clues. They form a geometric figure and sometimes you need to see the whole figure so you could prove a clue false or correct.

Moreover, if you have a puzzle that has only pinocchio/liar clues and it has symmetry like the puzzle in this contest, it doesnt look that nice with small digits in the corner.

Based on that, I think the most suitable design is small digits (possibly coloured) in the corner for Pinocchio Sudoku and standard sized contour digits for Liar Sudoku (the design used for this contest). But it is only my subjective opinion and any other opinion will be useful for our upcoming LMI contest.

I remember I made a few stupid mistakes in the pinocchio round during WSC2012, due to the fact I kinda forgot that pinocchio clues cells could contain a different digit.
I think my speed solving is also based on quick recognition of blank cells pattern. With this layout, it is really hard for me to solve it fast and safe (without mistake).
Although I also created and designed pinocchio sudoku in the same way as WSC 2012, last year I've seen a pinocchio sudoku in a national qualification tournament (with more than 3-4 pinocchio clues), that was designed with small clues on the corner, and I immediately thought it was a brillant idea and a better layout.
Also I don't think there is negative effect such as « information that is too hard to see » with smaller clues. The clues can also be designed as in previous sudoku GP round Roman Numeral sudoku, i.e. clues on upper half of the cell. With this layout, I don't see in which way it can be hard to see informations given by right clues. There are ways to make it clear with enough room to write the right digit in the cell.
I think the fact that in the IB, you had to erase the wrong clues to draw the solution is a sign that it is not a perfect layout.

I think the « rule » should be that only clues that are part of solution should be drawn occupying the whole cell, and I think I'll apply this in my future designs.
I know it is subjective, and I know that for me it's important not to have « visual » obstacle in my solving. Perhaps I'm the only player in this case.

I remember a few cases in the past where designs were discussed :
In WSC 2011, the design of false sudoku was discussed on the WPF forum before the IB was published, if I remember correctly. The decision was to design small clues on a corner of a cell (rules were different as here, though).
On LMI, there was a discussion about marking or not marking the diagonal for antidiagonal sudoku, A player asked for a change when the IB was published, and a poll was made to ask the players what they prefer (the result was against the player request, so it wasn't changed, but there was room for change based on the result of the poll).

I understand that a change can't be done base on request of only one player, but I think if there is a change request there should be enough time to invite players to tell which design they prefer , by a poll, or simply an open discussion in the forum during 3-4 days. Of course it's not possible if IB is published only 2-3 days before the beginning of competition. That means there is no way for a player to see a design change from IB to competition booklet. I don't think it's good and I think this is a reason the IB should be published at least one week before the competition.



Yesterday, I was planning to post that the clues as they were (full size, white inside) were fine for me, enough different to not consider them true, easier for noticing the constraints on wrong placement than small ones, and circling the wrong ones was enough to notice which one was to change.

But then, with the resuls out, I now know that I failed to solve correctly liar sudoku, and checking my grid, it was because I considered a clue I had circled (so a wrong one) as true (as it was, I did manage to find all the right wrong clues, but then failed to solve the resulting sudoku).
So what I thought as a sufficiently different clue at the beginning ended being too normal in the middle of solving. Also, maybe circling wrong clues was not a good idea after all, considering "circling" is usually associated with "true".

I don't have a lot of experience with other layouts, so I can't exactly sayif another one would have been better, but I can telle that this one, despite my initial good opinion, didn't work for me this time.
(but I broke three different grid in this round, so it might be that I wasn't in my best shape while solving).

Hi Kithyane,

Circling the wrong clues in the solution is a way of quickly identifying them in order to highlight the liar rule to a beginner. By circling them in the solution it is clear there is exactly one per row, column and 3x3 box for the given clues.

Clearly, circling all the correct clues is far less efficient for this purpose. I can't say I personally associate circling with "true" (I might associate it with "different") - but nevertheless I think you are confusing the presentation of the solution with a recommended solving notation. It is up to the solver to decide how best to solve the puzzle, and part of that process is using your own convenient notation. Different solvers will doubtlessly have different notations!

I hope that makes things clearer,

Tom Collyer
Competition Director,
WPF Sudoku GP 2016


I realise I've not been very clear in my previous post. Don't worry, I didn't take the presentation with circles as the recommended notation for solving !
I did consider a few others but chose circles in the end, essentially because I found it easier to check the "one per row, column and box" constraint, and I thought that it would be important to keep track on that. Now that I know which mistake I made, I'll try another notation next time.
(I wouldn't have said I associated circling with "true" before, or I wouldn't have picked this notation, but it seems despite my best effort it is the case. Something to work on, I guess)

Hi Fred76, jzverina,

Thank you for your long and thoughtful posts. I apologise in advance for not being able to reply in as much depth, but I think you both raise some interesting points both for and against.

To Fred76 in particular: I'm sorry that you were not able to reach your usual high solving standards this round, but it seems to me that you have made your mind up and that we are going to disagree with each other. The IB is scheduled to come out 1 week before the contest so I am of course sorry that there was a delay of a couple of days. However, there is no way that it is practical for me to put issues of presentation to a public vote even with a month before the contest - there are simply too many other tasks with a higher priority. Sometimes you have to accept the decision of the director.

On balance I am happy with the presentation as used for liar sudoku, particularly with regards of highlighting clue symmetry. I accept that solvers may have struggled with it, but it is also clear to me.from the results that many did not. It is not yet clear to me that your alternative presentation offers only advantages without disadvantages, but you are of course free to pursue your own design preferences and I look forward to seeing them.

Let me finish by saying that although we disagree, discussions like these are very useful and will of course inform decisions made in the future. So thank you!

Tom Collyer
Competition Director,
WPF Sudoku GP 2016