To add to the score sheet. Thomas had actually completed the entire sheet in 103 minutes with 17 minutes to spare!!!! But a rare typo error happened when he was transferring the digits to the text boxes. However the error got corrected with 4 minutes to spare and I think that is what cost Thomas his First position. This is just trivia.

Rishi

Yes, just trivia :(. We all make mistakes from time to time. Regardless, congrats to WaterlooMathie on the victory.

I had one other comment which comes back to the entry system. It was not the reason for my mistake this time, but has gotten me on two previous sudoku tests and still does not seem to be ideal for paper solvers. In particular, I did not find the new text boxes to be any real change/improvement to the system. If anything it added more confusion to the visual applet without any new advantages. The challenge for paper solvers is still that one has to click "show cells to fill", then enter said cells, then click submit, and must do this on all puzzles. It takes longer than the entry on the puzzle tests and is more error-prone as well. I'm sure I'm not the only one to forget to hit "submit" on a puzzle they had finished (as happened on the Qixi Festival Test), or who got confused about what to enter by not realizing there is a "show cells to fill" button, as Calavera reports here, and as happened to me my first time too.

The puzzle test entry is much more friendly, and should be mirrored here. I'd recommend putting a separate tab at the top of the page labeled "paper solvers". That tab goes to an answer system exactly as on the puzzle tests, with just two boxes for each puzzle entry. There is a single submit button at the top/bottom that enters all entries when clicked and reports how many solutions have been entered and the time remaining. Having a more compressed set of answer boxes lets one constantly view the clock, and also makes it much friendlier to enter a single answer to a single puzzle. Currently, one must scroll a fair bit, and the puzzle labels don't pop out to me for some reason so much as the red text "can't see sudoku" which I always think is the variant name and not a help button. My favorite part of the puzzle system is that it tells how many solutions have been entered and the time remaining. There is no global report on the current sudoku test, and that makes me waste a minute each time before checking anything else to just see that all the submits were hit.

The one sacrifice that is necessary is to be willing to label rows and columns on the test itself. I still don't exactly understand why these need to be kept such a dear secret until halfway through the test when they tend to be very well chosen, and not having the marking on the paper makes it easier to copy the wrong row/number which seems bad too. Making a clearly demarked entry system for paper solvers and cleaning up the existing one for online solvers without the extra boxes should be to the benefit of every solver. Just my opinion, but I felt it was worth sharing since you have so often taken the time to improve things and have been getting much better.

Thanks for your inputs.

First some quick stats : We received total 800 answers in this test, out of which 300 are submitted using text box. That is a whopping 37%, considering that text boxes are introduced for the first time. This goes on to say that paper players definitely don't like the existing flash submission system, and we must do something better.

The alternate submission system, the replica of puzzle submission system, is rather easy to develop. But as you mentioned (I also noted here), we need mark the rows / columns in the pdf itself. Without marks in pdf, it is never going to be easy for paper solvers.

As more new players take part in LMI tests, it looks like they are confused with 'Show Cells to Fill' button or they simply don't know to use this feature. Several members have reported this in the forum.

Even if we don't provide the alternate submission system, we must have a single Submit button. This has been reported in forum earlier, but it is not done yet because of our technical inefficiencies.

I'm leaving it open right now for further discussion / inputs, but we'll definitely have to improve it for the next Sudoku test (26/27 March).

Indian Results Summary -
Rohan continued his good performance, standing 1^st among Indians, overall 6^th.
Gaurav is deservedly 2^nd, while I believe he is still little rusty.
Jaipal is in good form, and looks all set for this weekend's Hyderbad's regional round.
Well done guys.

Sumit needs to do bit of catchup, but I'm sure he will.
Harmeet looks like in good form too, he mentioned that he was 'focussed' this time.
Rakesh probably decided that he will only do the high pointers ( and he achieved that, but with an overall low score)
Akash continued his trademark not doing anything in last 30 minutes :-)
Ritesh and Amit are mysteriously missing.

Thanks for putting this up Deb... My miss on mentioning this when declaring the results.

Rishi

it was a great set of sudokus. started with 1st one but dont know how , i was stuck in that for a good 15 minutes and ultimately got it wrong. then did the 4 killers . then did the jigsaw diagonal. aroud 30 minutes were left and i started with disjoint diagonal . and as deb nicely put it "Akash continued his trademark not doing anything in last 30 minutes :-) " i dont know how i got it wrong . so wasnt able to do anything in the last 30 minutes . someone please help me get over this serious problem.

Is there a neat and elegant approach for solving the 150 pointer? Can someone share on the forum? I got stuck after a while...and had to resort to trial and error to complete it.

One solution could be ...to start the test 30 minutes after you click on "Start"

funniest answer :-)
on a serious note, please help him if you can....

Thanks Akash, I think what you should do is, if you get stuck in a puzzle or are wrong, put it aside and solve others and go back to the wrong one later with a fresh mind.. atleast that is what i do when i go wrong in a puzzle. If I start that wrong puzzle again immediately, the previous numbers keep popping into my mind and stop me from thinking clearly.

Rishi

I will share that Rakesh but I would love to see Motris and Rohan post their approach. I am not aware of Snyders exact time for that puzzle but I know Rohan cracked it in under 7 mins which made a mockery of the puzzle. It was awesome the way he solved it.

@Rohan, maybe you can share your approach with ppl here.

Rishi

First of all, nice set of puzzles Rishi. Got completely mixed up with the variants in the middle of the test, but managed to come out clean :D

The 150 pointer was the Diagonal-Antiknight Sudoku. I started the test with this sudoku. I got a couple of numbers and was then stuck. A little guesswork made me progress with a couple of more numbers. By this time 4 minutes were gone. Then it struck me. Observe the given numbers. All 4s and 8s are symmetrically opposite. All 2s and 5s are symmetrically opposite. Similarly, 1s and 7s, 6s and 9s. Also, the 3s are symmetrically placed. Once you realise this, I think the puzzle is not very difficult because the entire grid will have these pairs in symmetrical opposite cells.
Even after getting a few numbers, you will observe the pattern holds. Hence, I directly placed a 3 in the middle cell (R5C5). Then I solved the puzzle in about 3mins using this logic.
Solving the toughest puzzle in 7 minutes does give you a boost in confidence :-) But I still couldn't complete all 12 :|

If I found it, I'm sure many of the top solvers must have used this logic too :-)

Rohan.

Thanks for the inputs Rohan.... You are right about the digits being symmetric. In a competitive environment it definitely helps in noticing such finer details which enable you to solve hard puzzles faster. At times in competitions where time is of essence, guesswork and T&E d o help us solve faster than actual logical methods. But for benefit of all I will explain below the intended path using the two constraints of Diagonal and Anti-Knight as was the intent of the puzzle.

On a lighter note, the putting of 3 in R5C5 does not require the symmetry property and comes by default sudoku rules as we have 3 as a given in Box2,4,6 and 8.

To begin with, we can place all the 3s in the given grid once we have the 3 in R5C5. I have pencil-marked all the other cells. 8 in box 7 will be in either R7C3 or R9C1 and no where else in the UR-BL diagonal. similarly, 4 will be in R1C9 or R3C7 for the same diagonal. the number 1 will be in either R2C9 or R1C9 in box 3 as per the given image below.

Rishi





Attachments
----------------
AKDiag1.png (33KB - 2 downloads)

Now let us look for the options of number 1 in the Tl-BR diagonal in the earlier image. We have the possible places of R1C1, R2C2, R7C7, R8C8 and R9C9. R9C9 is eliminated because of 1 being in either R1C9 or R2C9. in box 6 the possibilities of 1 are either R5C8, R6C7 or R6C8. This forms an L shape ( as explained in my blog the L Technique ) . Whichever cell 1 is in, it eliminates 1 from R8C8, as shown in image below. so that leaves only cells R1C1, R2C2 and R7C7 for the number 1 in UL-BR diagonal.







Attachments
----------------
AKDiag2.png (45KB - 2 downloads)

Now let us look at possibilities of 1 in Row 4 and column 4. These have been marked by rectangular boxes. In R4, we have options of R4C2,R4C3 and R4C5. If 1 was in R4C2 or R4C3, it eliminates 1 from R5C3 since it is the same box and if it is in R4C5 then it is at a knights step from R5C3 and it eliminates 1 again. So 1 can be removed from R5C3 marked by a circle.

Similarly the possibilities of 1 in Column 4 are R5C4, R6C4 and R9C4. marked by rectangles. If 1 was in R6C4 it eliminates 1 from R7C3, if 1 is in R5C4 or R9C4 it again eliminates 1 from R7C3 since both these cells are at a knights step from R7C3. So in col 4 whichever cell 1 is in, it eliminates 1 from R7C3 marked by a circle. So we have a revised options for 1 as in image 2. Now we have only R1C1, R2C2 and R7C7 for the digit 1.







Attachments
----------------
AKDiag21.png (46KB - 1 downloads)
AKDiag22.png (32KB - 1 downloads)

Now let us look at Col 3 for the remaining possibilities of number 1. We have R1C3, R2C3 and R4C3, marked by rectangles. If 1 was in R1C3 or R2C3, it eliminates 1 from R2C2 (marked by a circle ) as it is within same box, and if 1 is in R4C3 then also it eliminates 1 from R2C2 as it is at a knights step. so whichever cell in Col 3 the number 1 comes in, it is eliminated from R2C2. Now that leaves us only R1C1 and R7C7 as a possibility for 1 in the TL-BR diagonal, marked by rectangles. Now whichever cell in this diagonal 1 falls in, it cannot be at R7C1, marked by a circle so that eliminates 1 from R7C1 and we can safely put 6 at R7C1. (shown in the second image.







Attachments
----------------
AKDiag3.png (46KB - 0 downloads)
AKDiag31.png (32KB - 1 downloads)

Now that we have got a 6 at R7C1, by using the constraints we get all the 6s in the grid as shown below. Post this I think we can proceed normally and solve the entire grid.

This was my interpretation of this puzzle and the reason I gave it such a high value.

But in a competition the end justifies the means so you dont get marks for methods used but only for the correct solution.

Hope this helps.

Rishi





Attachments
----------------
AKDiag4.png (30KB - 1 downloads)

Thanks Rishi for sharing your approach. I don't think my approach was much different from this. But, probably, I had got stuck somewhere where I could have used the anti-knight constraints to eliminate candidates, and had to guess for placing a 4 between two cells. I'll try to solve it again from the point you have left it here.

Paper solvers had it simplified this time with a simpler answer entry system (by and large, with a few exceptions). Of course, there can be further improvements which can be discussed and sorted out.

I am in the minority who are solving online. And there is one thing which I would like to mention, from an improvement perspective in future tests - the representation of the diagonal lines on the grid. This time we had five grids which had a diagonal constraint. The example on the site also had diagonal lines and was represented using gray lines. But in the test four out of the five grids had black (dashed) lines running diagonally. I am attaching an image which brings out the differences clearly. The pencilmarks in the gray one are clear, but the other one requires a lot of eye exercise while solving the sudoku - especially the numbers 1,3,5,7 and 9 - each of which interfere with one or more of the diagonal lines.

We can surely avoid this by choosing slightly light colors for these lines.





Attachments
----------------
diagonallines.png (54KB - 1 downloads)

I already sent Rishi my approach to the last puzzle. On things like Diagonals I often do minor coloring/trial steps to make eliminations. I certainly took longer than "7 minutes" but it was maybe 15 to do this:

While I didn't instantly notice it, I caught on quickly that the puzzle had rotational digit symmetry (the 2/5 would be placed in symmetric spots, the 1/7, 6/9, 4/8 as well - 3 is self symmetric).

Most of the solve involved considering the most limited digits (first 2s and 5s) and how certain forcing placements cause contradictions. I'll describe steps for "2" but same applies for "5" on other end

After placing the 3's, 2 can be in three positions in box 4, but if it is in R5C2 it leaves no spot for a 2 in column 3, so a 2 is in R45C3, and in C2 in box 7. If the 2 is in R7C2 (pushing the resulting 2's), there is no possible place for a 2 on the LL-UR diagonal, so R8C2 = 2.

Next, either there are two 2s at R4C3 and R3C4 or three 2s at R5C3, R4C4, R3C5. The latter case (more tempting to try) led to an immediate contradiction in the lower-right box as you can't put in 2's to avoid knight constraints or diagonal constraints. So 2 = R4C3 and R3C4.

Then, 2 is in R6C56 but if in R6C5, then no way to put a 2 on the UL-LR diagonal. So 2 in R6C6.

Then (describing 4, but 8 is symmetric), there must be a 4 somewhere in column 6. You know the upper-right has a 4 in either R3C7 or R1C9 and these form (from the anti-knight constraint) a perfect elimination of a 4 from R1C6. This is my favorite step. Anyway, other digits eliminate all but R9C6 for that 4. This gives several more 4's, and all the 2's as a result.

Last necessary step comes from thinking about R5C7 spot. It contains a 7 or a 9. Whichever digit is in this cell must also be in R6C6 because of the eliminations in the center. But a 7 is already forced in C2 in R45, so you can't have the two 7s in the other spots, so this is a 9.

From there it is just entering digits and copying the symmetric logic, but very interesting and challenging puzzle. I think the Anti-Knight Non Consecutive was my favorite, but this was a close second.





Edited by motris 2011-02-14 10:44 PM

You did indeed send me your methodology Thomas and in hind sight your approach appears much better than what I have put up. The only reason I mentioned in the forum for you and Rohan to post your methods was that others would be genuinely interested in knowing how you approached the puzzle. :)

Rishi

Thanks Rakesh for bring this up.. I have received a lot of comments on the use of lines as diagonals and will avoid it in the future.

Rishi

Thanks for the great puzzles. I'm not usually much into Sudoku, but these were great fun. (I did poorly on the test, but solved the rest later.)

By the way, the Anti Knight Disjoint is solvable without the "disjoint" constraint. Yes, finding this out was an accident.

Cheers
Rob

I didn't see the symmetry of the 150 point puzzle but I was able to solve it logically. For instance you can see from the first diagram of Rishi's, that R2C8 is a 5 because the anti-knight moves cancel out the possibilities for 5 being in any other spot on the diagonal.

Edited by WaterlooMathie 2011-02-15 8:43 AM

Hi, Is there anywhere password to get the PB?

regards