I agree with all of this actually. A puzzle, once chosen, can have lucky/unlucky results, depending on solver's strengths vs required logic and a whole lot of other factors.

BUT, again, my main point is that this has to do with points distribution, and something that can be noted only after the puzzles are selected and solved and give out unexpected experiences. It is possible there will be some luck in the puzzle selection. It is also possible that there won't be and all the points are actually pretty accurate, at least for most individuals. The point is, as a solver, no one should go in "expecting" a puzzle to be wrongly valued.

The context here means I wasn't saying luck will never be a factor. I was replying against it being a given that luck will come in during puzzle selection. But, I hope you agree, IF the points are assigned right for an individual, then puzzle selection for that individual becomes more about having their own strengths and strategies. And as a solver, one should go in expecting that to be the case.

Because all individuals solve differently, I'll argue it is impossible to assign points "right" for more than a few individuals on a given test even when we do our best as test-setters. With several test-solvers (Grandmaster Puzzles currently uses about 6-8), you can get a view of consistent puzzle times and high or low variance puzzles but your points are really just a best estimate of the mean value and some puzzles have very wide spreads and some have very small spreads even for individual solvers.

As I see tests, any loop puzzle or number placement puzzle will probably be overvalued for me. Any arithmetic puzzle will probably be undervalued. Looking at this test, for example, I took longer on the Tapa ? (over 3 minutes for 30 points) compared to the Different Neighbours (2.5 minutes for 95 points) and Indirect Yajilin (2.25 minutes for 75 points). Some of this was a mix of good and bad luck but there is quite a difference in returns from this scoring. If I had to leave one of these puzzles unsolved at the end, would I make the right choice? On the other hand the Kakuro took me what looks like 15 minutes for 110 points. I actually feel all those four puzzles were about in the right score range for the general solver (Tapa maybe wanting to be 45). But still, the points do not come close to matching my results.

Tom's point, and mine, is that score/time guidelines are useful but by their very nature imprecise things and hoping for "perfect scoring" will always be a challenge. The fairest ranking of players will always be the time it takes them to complete the entire set of tasks before them. That way solvers will all have cleared the same hurdles, the easy and hard ones, and so if you want a good ranking of 10-50 solvers you should try to get 10-50 solvers to finish the test.

That we had 10 solvers finish is very good and first and second are clear. The effects of different penalty systems will change the order of 3-5 but we have the clearest sense of the relative solving results of these solvers. And no, I am not using the final scores for this comparison. Just the raw times.

Changing the subject a lot here ... the puzzles I struggled with the most on this test were Battleship and Different Neighbours, and the Tents I feel like I solved more by guessing/intuition than by really knowing how it was going to work, and none of these were puzzles I ended up submitting during the 75 minutes (I wasted most of my last 15 on failing to solve the neighbors and on failing to count to 6 properly with the snake). Any tips on how to get started on those three puzzles? Once I got a little bit of a start on neighbors (by checking a couple different cases) and on tents (by guessing), I could see how to finish the solve pretty straightforwardly. But that first step was hard/lucky to find.

Back to the original topic, the kakuro took me about the same time as the median of the other puzzles I solved -- in other words, for me it was massively overvalued. The hamle was also some relatively quick points for me. The slitherlink and the tapa took me a lot longer than their point values would suggest, in fact taking almost as long as the kakuro and the hamle took me, certainly way over half the time despite being worth less than half the points. I see, though, that other solvers spent a lot of time on the kakuro, so I'm not suggesting it was valued wrongly, just that there's some combination of strategy and luck in puzzle selection when you're not going to solve all the puzzles.

I used to be able to go into these tests figuring that I wouldn't solve even half of the puzzles, which meant that I could pretty much focus on types I know I'm pretty fast with, and even if it means I end up not looking at a puzzle that's overvalued on average for most solvers, chances are the ones I don't do would turn out to be inefficient for me! Now it seems like I have to plan to at least look at all the puzzles and go for it on the ones that I know how to start quickly, and the strategy/luck is more often in knowing when to give up early vs sticking with a hard puzzle to the end.

Maybe a lot of the variance (luck?) is in the design of the test -- a high-value kakuro is likely to mean some easy points for me, and gains relative to most other solvers, whereas a high-value slitherlink is likely to be a puzzle I don't even try to earn the points on unless I'm in danger of finishing the test. I agree with motris's point that the best measure of my overall solving ability on the test would be the time taken to solve all the puzzles. I may be able to get 2/3 of the points in the allotted time but that doesn't mean I would get all the points in 3/2 of the time!

For example, try to solve different neighbours with letters. At the end you will be able to identify which letter stands for which number.
The "eyes" in the tents have only two possiblities - to form a "square" or "tetramino". It is obvious, that one construction should not be used twice.

Edited by greenhorn 2013-10-31 1:25 AM

Define "good ranking". Are we then saying that, in this test for instance, the top 10 was a completely fair reflection of skill, but the rest of them were all dependent on luck? Maybe in one test if a person has a bad day due to puzzle selection, then you'd say that one performance was on luck, but that is about the same level of luck as someone solving 9 out of 10 puzzles the fastest, and then getting completely stuck on one to the extent that they can't finish. Over a bunch of tests, both of these factors get negated and generally the rankings turn out fair. Its why in the WSC/WPC the preliminary rounds usually end up being a fair reflection of skill, even though they probably have the same finishing level of top few.

In general, I disagree with this point that a bunch of solvers clearing the same set of tasks minimizes the luck factor. There's always luck involved that a person with the most skill overall still might not spot that one deduction necessary to solve one puzzle. There's also then the choice issue, where you know this test is finish-able and should be done by you, but you're stuck, so do you guess, and then you enter into more luck territory. I made this last point because its the same as a slower solver taking a chip at a high pointer even if they're not sure of finishing within time and then choosing to guess through with time running out. In the end, I just feel both scenarios even out for everyone and its just down to the individual and the relevant competitors having a good or bad day, as it is in most competitive environments.

I'm saying "good ranking" in the sense of robust to changing conditions. Let's say two different test solvers were used for the test, and new scores were applied. Would Murat's 13/15 without Akari and Slitherlink still be better than all of the solvers with 14/15 solved? What if the test ran for 70 minutes. Or 80 minutes. How one sets the rules and scores for a test solvers cannot complete can cause large swings in rank position in these ranks. Regardless of how one sets these things when a test can be finished, the top 10 are basically unaffected. That is what I mean by having a good ranking for those solvers. It doesn't matter if player A was a tester or X was the time and not Y.

To Josh's question -- I definitely found the test rewarded me for being able to sense where intuition and logic would work well. I did use letters with different neighbours to link a whole bunch of cells, but then went to intuition and using uniqueness to get the assignments quickly. The Battleship, surprising enough, was really done by logic which is unusual for the type and made it one of my favorite puzzles. There is a really neat consequence of the logic of the two "1" rows and the two columns with three available spots that each need to take "2"s. Once I caught onto that, it was a much different puzzle.

Edited by motris 2013-10-31 3:55 AM

I actually disagree with this too. It may be just a perception thing, but points do play a role sometimes in how players subconsciously approach the puzzle. If I do a 7 pointer, I'm expecting an easy path. That could derail my run of solving if it turns out to be hard, I get frustrated, and decide either its not worth the points or just put up with it. Whether I finish the test eventually or not, it does cause a slight change in my progress over the test.

I'll give an example - I finished TVC XV. After the previous two, I was expecting to be around that level where if I have a good day I finish, else I'll come up just short. One of the puzzles that really got me stuck was a Double Back Tapa. It was 63 points, which was still a good enough sum for that test to be patient and have a look at it. So I stayed at it without ruining my thought process and needing to change strategy. If it was worth say 30 points, it definitely would not have been worth the waiting, and I now have a choice on whether to stay on it or break my progress and go to a new one. And at least for me, that has resulted in a mistake or two later, in the past.

Maybe some people aren't like that, and just stay the course no matter what, in which case your argument holds true. But that can't be everyone's way. Whether a person finishes or not, they could be affected by points distribution, which could in turn affect the rankings.

In the Tents puzzle I assumed that neither of the top 2 patterns could be a square as this would mean a non-unique solution with respect to which tree was connected to which tent.

I can't see where the password is the booklet this contest, which happened while I was on holiday. I'd like to do the puzzles... am I missing something?

I meant to get the password removed from the pdf file, but missed to do it.
The password is now available in the (archived) page - http://logicmastersindia.com/lmitests/?test=M201310P3.

Thank you!