The technique is called crossover.
You can learn it from http://www.sudopedia.org/wiki/Crossover
or search it by google.
Any other question?

Thanks Min. After that crossover step, it does indeed solve out logically.

It will become easy buddy when you spend some time and give the total time till the submission of last answer in minutes instead of time of submission.....

I am happy to see others found the Diagonal to be difficult. I spent many minutes trying to find that first number and simply couldn't find it for quite some time! But once I did I was able to make some progress -- it still was not easy from that point forward however.

The crossover step is perfectly legitimate, and I think this is one of the best Diagonal puzzles I've ever solved. Like many others on the test, it is 5-star.

When it comes to ties -- I'm reminded of the 2009 Semifinals where a time-based tiebreaker for incomplete solutions was used, and it did not seem to work out very fairly. If two great solvers did exactly the same puzzles and didn't finish, maybe they both deserve the credit?

I was not in Zilina, but I think it's a very different situation. Here you have not to choose stop playing before end of time. You can submit answer before the end of test and continue to try to solve last puzzles.
if two great solver who didn't finish have same amount of points, but one reached it after 1h45 and the second after 1h58, it seems fair to me that the first one appears first in the ranking.
I don't say it's not fair now, it's just a detail, not very important for the LMI tests.

For diagonal sudoku, I've quite often found visualizing placements along the extra constraint, such as at WSC2 Prague Final and here, incredibly valuable. Indeed, so much so that I specifically take time in my solving routine to think about the immediate consequences of any number much more so than any other variant, kind of like a Nishio where you see if a few placements down the way you can possibly place all 9 instances of a digit with exactly 1 on each diagonal. This is generally not useful on classics, since a lot of either/or choices in empty space don't do much, but on diagonals the difference between two options is often digits on 0 diagonals, or 2 digits on a single diagonal, or other results that can't work. So I would not say I was looking for a crossover here (I'm not familiar with the term actually). But one of the first things I did was convince myself R1C1 couldn't be a 4. Similar thinking next showed me the 7 couldn't be in that spot either, so one or the other had to be in the center and that became obvious as the 7 (leaving R2C2 for the 4). That's just how my mind approaches diagonals. And I guess I suspected by then, since no Matchmaker could have any digits placed without identifying the extra constraint, that some cool diagonal cycle was certainly the way to start the puzzle here.

Re: Zilina and tiebreaking - the fundamental problem in Slovakia was not grading playoff puzzles along the way, as has been done by virtually all other WPC/WSC hosts in my time competing. Having 36 semifinalists in a single heat made this impractical, but that's a whole other discussion about good/bad planning decisions.

Within your scoring system, you have time-stamps for when a solver has submitted their last "correct" solution which makes it clear when, at a particular score, solvers have reached that point. Imagine a close test that has a grossly undervalued puzzle that is a particular stumper for many people. Maybe no one solves it, maybe someone does. But if 3-4 people are tied without just that puzzle, then the first to have everything but that one solved is a very fair tiebreaker to use. Here, Jakub and I are still pretty close to tied within 2 minutes (not a terribly meaningful amount of time to separate us, but I'd pip him out), but anything approaching a 10-15 minute separation case, as was the case in Zilina, would have let a large number of tied solvers be separated fairly. There, many solvers reached 3 of 4 puzzles done with 10+ minutes left to attempt the last stumper. Slower solvers who eventually reached 3 of 4 puzzles done with 2 minutes on the clock just said done and turned in their papers since they had no chance to finish the last puzzle. Since you'll know when solvers get to a particular score, its not a comparable situation.

As one of the most vocal critics of competition rules, I'd be completely satisfied with using the last submitted time as a sorting method with granularity at 1 minute levels pretty reasonable to view as a tie break.

Edited by motris 2010-08-18 4:02 AM

Using the timestamp of the final correct puzzle is certainly not an unfair way to break ties. I don't think anybody would have a claim that they deserved to be ranked higher *because* they didn't submit until after the other person.

But when there are still puzzles yet to be solved, there are other reasons why someone could be "ahead" (which is to say they solved all but one puzzle more quickly than anyone else). If you are done with *all* your puzzles (and you solved on paper), then you immediately submit them electronically. But if you have one puzzle left...maybe you spend as much time as you think you can on that puzzle? Maybe you leave five minutes at the end of the test to transfer your paper solutions over to the website, so while it seems like you just then solved the second-to-last puzzle, really you reached that point many minutes before and were just stuck.

If this tiebreaker were adopted, since players from all over the world value LMI tests as the top monthly sudoku competitions, they would naturally want to rank as high as possible. But I think that would mean they would go from paper-to-computer-to-paper-to-computer...over and over near the end of the testing period, which is just when they need to focus the most on solving. I think that's needlessly inefficient and doesn't really reflect who's better.

I would prefer weighting the puzzles themselves *slightly* differently. Give an extra 0.1 to the hardest puzzle, 0.01 to the second-hardest, 0.001 to the third-hardest...and so on. So if you match someone's point total, maybe these fractional points will provide the difference. Or, in the case where nobody finishes the test, let the top performers stay tied.

I'm in Zilina too. By the rule of round 2, i came to add-round, more than 70 players had same score with me. Maybe after that, many players use the rule to get previous rank. I think it's not fair to top players, maybe they only use 20 mins to finish 3/4, but the fourth puzzle is very very hard. Other players maybe use 30 mins to finish 3/4, if they submit first, they come to next round.
Online competition is different too, we can't know if one is still working on the puzzles after last submitting. If base on last submitting, it will be more unfair.
I had thought if it's necessary to make every variant in different points, as Ziti said, i can give diagonal 56 pts, and maybe non-consecutive 54 pts. But they are all 55 pts. Because it's unlike with classic, everyone will be good at different type of variants. If your are good at one variant but it has a low score, you will feel unfair to you. And see once again with the diagram, many players maybe didn't find P14 can't be no 3 odd/even, then diagonal, touchy, no 3 odd/even are last remain puzzles. As you see, the hardest step in diagonal is the first step, so i think one will give up this puzzle when he can't find any step.

Edited by Minfang Lin 2010-08-18 6:57 AM

To Thomas:
Do you know the technique Finned X-Wing, Empty Rectangle and ... they are alike nishio, and i think they are good and interesting techniques.

Every variant has its own special technique, and the designer should show it in his puzzle.

To Ziti and others (I'm sure you will like the logic ):
Find the next step of this diagonal sudoku:

(Hint: check digit 9)

Neeraj, Sorry it took longer than I thought I would take because of other commitments. But finally I've done it (2 weeks? after you suggested)

Updated score page - http://logicmastersindia.com/M201008S2/score.html

I've added 2 more columns -
LastCSub - stands for - last Correct Submission
TimeTillLCS - stands for - Time Taken till Last Correct Submission

Examples (for whom LastSub and LastCSub are different )
Gotroch : 6 minutes after he submitted the last correct puzzle, he just submitted P09 without filling anything.
Rohan Rao : 25 minutes after he submitted the last correct puzzle, he just submitted P09, which turned out to be wrong.

Many other examples are there. But I just choose top 2 for illustration.
The table is now sorted by Total Points Descending, then by TimeTillLCS Ascending.

However, I've not modified the Ranks yet, since players are having different opinion about using LastCSub as a tie-breaker rule.

I tend to agree with Ziti, when he mentions that paper players would be at a disadvantage since they might have to do paper-to-computer-to-paper-to-computer...over and over near the end of the testing period.

Lets wait for few more voices before we take a decision.

Very nice example Minfang. When I'm solving sudoku for fun I often run into eliminations like Finned X-Wing or Empty Rectangle which are very much what I'd just call situational thinking - like a Nishio with two steps ahead - but are pretty powerful to catch. When I'm solving classic sudoku for speed though, I rarely reach a stage where I'm searching for any of those things and on hard puzzles I will admittedly bifurcate with underlined digits to see if I can force a solution onto a grid. I'll also use uniqueness and other things that I wouldn't use when just solving logically.

To Ziti: Solving on paper versus online is a choice applied to all solvers evenly (choosing paper means you wait for it to print, but get the advantage of any notes you want, ...). Solvers can do some on paper and some online. Its the solver's option. So I don't think its necessarily unfair to apply a time-based standard that might require paper solvers to enter their answer with say 30 minutes left instead of at the last second, provided the value of time is made clear ahead of time. For this test, I entered all the relays and five of the matchmakers with ~25 minutes left, figuring I should get those pages entered, and then simply came back two more times to enter the last. Of course I missed one submit button (on the easiest respective "nonconsecutive" puzzle), but that's beside the point. I'd say you and others would use a similar strategy on the USPC - at some point after 2 hours make sure your answers are in place and then start trying to pick off 1 at a time. This approach - certainly with solvers knowing how tiebreaks apply - seems like one can adapt how they'd play on paper to not be disadvantaged.

Its also true that the matchmakers weren't evenly difficult, so changing their scores a little would help. More important is probably increasing the relative score of the matchmakers versus the classic relay as the classic relay was certainly a much higher point/minute value than the other which is why I started there. Solvers like Nikola that finished the whole Matchmaker were somewhat hurt scoring-wise by the relative valuation, although different scores for that section would certainly affect my strategy and likely others' as well.

Edited by motris 2010-08-18 9:06 AM

This is fixed now (hopefully correctly)

To Thomas:
We called the method like this "magic step". I think all the top players have their own magic method.

Because many people only play classic puzzles, they don't know how to solve variants, and everyone want to get a better score, everyone will be happy when he get a higher score. It's a relay, so later puzzles also have extra points for all correct previous puzzles. Nikola and zorko did best in matchmaker, i think they can get 100 pts more for the great job. And Nikola show us another method for diagonal, it's also great.

Amit said: The small point classics were really beautiful and I had never enjoyed solving classic puzzles so much before.
So we make 3 "small point" classics(solved from 1 to 9), but you should find the correct position of digits from previous puzzles first.
Just enjoy them!



000210000003000100060004020020000007006105300800000040070400090001000500000032000
000000600010304000600010000000900000009000500000002000000030006020105030004000700
003200000070000430040010005000500006000000100100006000900000020050000070000000300

TTHsieh&Leafcard

Edited by Minfang Lin 2010-08-18 5:41 PM

So, does Nikola move to 790...and, into first place...

Thanks for these additional puzzles Minfang Lin.

By low points classics I had meant the first four, so apart from the first one which went from 1 to 9, I liked the next three too :)

Good work with the increasing number of givens. Though I did not notice it during the test, this is the perfect way to display that the difficulty is not directly related to the number of initial clues.

The hard work that you guys have put in making the puzzles is evident from the quality of puzzles. Thanks a lot.

Lets not change the ranks for this test. We can start from next time if this works out.

Thanks for all the feedback regarding the tie breaking rules.

Ziti had pointed out that the paper players have a disadvantage if we now use the last correct submission time as a tie breaker. Motris also suggested that it would not be a disadvantage if the tie breaker rules are announced before the test.

Going forward we will use the last correct submission for breaking ties between players.

However since we had not announced the rules prior to the Qixi event we will not change the rankings.