|
|
motris
Posts: 199
 
Country : United States | motris posted @ 2010-10-18 6:42 AM uvo - 2010-10-18 6:26 AM
Now the contest is over I want to mention that the bonus system is seriously flawed: It is easily possible to lose points by solving a puzzle, which is certainly not desirable. For example, it took me about 12 minutes to solve the Number Skeleton mutant: If I hadn't solved it, I would have scored about 19x9=171 instead of the actual 7x10=70 bonus points - plenty to compensate for just one unsolved puzzle (-30 points).
There are two characteristics of your bonus system, which combined lead to this side effect:
1) Both the puzzles and the time bonus have a similar points per minute factor (1200 points / 120 minutes and 10 points per minute, if all puzzles are solved).
2) The decrease in bonus for not having all puzzles solved is very low (factor 9 instead of 10).
Together, this means that one puzzle where you are slower than the average 10 points per minute is worth less, possibly much less than the equivalent bonus score.
Unfortunately, 1) the mentioned Number Skeleton was not my last puzzle and 2) I didn't think about the bonus system in advance. But even if I had, I would have tried to solve all puzzles nevertheless and not cared about my score. (In a WPC, that might be different.)
The puzzles however were excellent (not that I expected anything less). Funnily enough, for the German Championship four years ago I designed a Battleships puzzle very similar to the hard one here, with the same 2-7-2 clues to work with. And I guessed correctly that the mutant Nurikabe would have twelve pentomino islands, but that didn't help much :-)
Yes, I thought of this scenario too late after we made the change to 9 of 10 earning bonus points. In the first pass the final puzzle should have been worth ~80 points, so with 9 versus 10 points per minute, you would have to take over 9 minutes to solve it to lose points. With the friendlier bonus system to not penalize entry mistakes, the end result is a puzzle you lose points on if you solve it in more than 3.2 or so minutes. This is certainly less time than the last puzzle a solver is likely to face.
The simplest answer - on a test like this which was pretty well-timed given the test-solvers - is to ignore time bonus and use raw clock time to break ties. That, or a much steeper drop off of the partial value was certainly needed. The partial time bonus was included to address the possibility someone might finish very early, but with an error, as has happened on past tests but that was not really a factor here. Certainly worth more improvements before the next time. Edited by motris 2010-10-18 6:49 AM
|
|
Aerion
Posts: 4
Country : United States | Aerion posted @ 2010-10-18 7:08 AM Ha, I see mathgrant made the same answer entry mistake as I did on the Easy ABC Connect -- accidentally started reading the wrong column halfway down. I felt dumb for not just writing the letters directly in the grid to prevent such a thing, but I feel less dumb knowing I'm in good company. ;)
Thanks for a great set of puzzles! Between this and a new issue of P&A (all while keeping an eye on the baseball), it's been quite the weekend. |
|
harmeet
Posts: 85
Country : India | harmeet posted @ 2010-10-18 8:20 AM Gr8 set of puzzles Thomas... it was really fun to solve them...
Deb... I think my minesweeper set of puzzles are not scored... can you please check them?  |
|
keshava.hs
Posts: 10
Country : India | Hi,Greetings!!Th puzzle set was awesome!!It covered all range of puzzlers. Thanks again.Please find the attached answer for Battleship-Hard, I could not find my mistake.3 Cheers,Keshav
 |
|
debmohanty
 
Country : India | harmeet - 2010-10-18 8:20 AM
Deb... I think my minesweeper set of puzzles are not scored... can you please check them? I think they were incorrect. I've your answer sheets, I'll double check. |
|
debmohanty
Country : India | keshava.hs - 2010-10-18 8:21 AM
Hi,Greetings!!Th puzzle set was awesome!!It covered all range of puzzlers. Thanks again.Please find the attached answer for Battleship-Hard, I could not find my mistake.3 Cheers,Keshav Keshav, R5C9 can't be 1x1 submarine. It has to be a part of 3x1 cruiser OR 4x1 battleship. Remember that R5C9 is a square and not a circle. |
|
keshava.hs
Posts: 10
Country : India | Hi Deb,
Thanks a lot for lightning response!
Since i didnot had printer so used paint brush for solving, where i have painted every cell, which has mislead me.
Thanks for resolving my doubt.
Best Regards,
Keshav |
|
rakesh_rai
Posts: 774
Country : India | Thanks for the nice set of puzzles.
For me, the best puzzles were (1) ABC Connect Mutant (2) Skyscrapers Hard and (3) Masyu Mutant.
All hard/mutant puzzles were not of same difficulty (but worth 30 each). So, for those solvers who cannot complete the full set in time, there was some luck/skill in choosing which puzzles to solve. For example, a few mutants - Battleship, Minesweeper, Tents and Tom Tom - were very easy. If only I knew this before the test...Mostly I found that the Hard was tougher than the Mutant.
Some comments on specific puzzles-
ABC Connect Mutant: Tough. But I have now managed to solve this somehow. Liked the way the loops have to be stretched sometimes to make the turns count, especially for 8 and 9.
Number Skeleton Mutant: Tough but solvable logically. Have seen such puzzles on USPC mostly.
Tents Hard: Had to resort to Trial and Error. If someone can show a logical point to start the solve, that would be useful.
Skyscrapers Hard: Very Tough.
Nurikabe Hard: Nice (not difficult). Initially, I thought the 44 was a printing error. But realized that 44 will actually start the puzzle and travel all over the world to finish the puzzle. It looked like a snake puzzle in some ways, where the length is given. |
|
rakesh_rai
Posts: 774
Country : India | I have a doubt. Didn't we decide earlier that the Last submitted correct answer timestamp would be used as a tiebreaker? But I see same rank given in case of ties. |
|
motris
Posts: 199
Country : United States | motris posted @ 2010-10-18 10:18 AM rakesh_rai - 2010-10-18 9:16 AM
Tents Hard: Had to resort to Trial and Error. If someone can show a logical point to start the solve, that would be useful.
This tents puzzle actually started with no number clues in the first construction, and makes use of "too crowded if a tent goes this way" logic. The final version got some numbers to steer some of the bottom, and also to distract from the "no numbers" approach. Maybe this image will help you get started.
 |
|
debmohanty
Country : India | Very interesting use of too-crowded logic. A quick look at the green tree shows the cell the corresponding tent should be in.
The numbers given in the bottom were definitely distracting, because there were of no use to get the start. |
|
rakesh_rai
Posts: 774
Country : India | motris - 2010-10-18 10:18 AMThis tents puzzle actually started with no number clues in the first construction, and makes use of "too crowded if a tent goes this way" logic. The final version got some numbers to steer some of the bottom, and also to distract from the "no numbers" approach. Maybe this image will help you get started. Thanks motris !! The puzzle solves very smoothly once I use this starting point - and it took hardly 2-3 minutes. Earlier, I was focusing more on the "4" column to get started. Perhaps that was an intended distraction. |
|
debmohanty
Country : India | rakesh_rai - 2010-10-18 9:16 AM
Nurikabe Hard: Nice (not difficult). Initially, I thought the 44 was a printing error. But realized that 44 will actually start the puzzle and travel all over the world to finish the puzzle. It looked like a snake puzzle in some ways, where the length is given. In Nurikabe, it was interesting to notice (may be many solvers don't notice during the test ) that the easy puzzle had only 3s, the hard puzzle had only 4s and the mutant, obviously, had only 5s.
The Art of Puzzles? |
|
debmohanty
Country : India | rakesh_rai - 2010-10-18 9:25 AM
I have a doubt. Didn't we decide earlier that the Last submitted correct answer timestamp would be used as a tiebreaker? But I see same rank given in case of ties. Yes, we had decided that.
For this test, we wanted to have a different tie breaker (#of mutants solved, then #of Hard solved, then #of Easy solved ). But we had not announced that before the test. So We removed all the tie-breakers and gave same rank for same points. But the table is still sorted by TimeTillLCS in case points are same.
Btw, I noticed that TimeTillLCS is 78 for you, which is way below others' TimeTillLCS in top 100 |
|
Nikola
Posts: 103
Country : Serbia | Nikola posted @ 2010-10-18 12:11 PM Great puzzles, thanks Thomas! That's why people love your puzzles. 
Pay attention to my time, I think that everything would be different if the test lasted 100 minutes. Mutant number crossword is "worthily" for the rest of the my time.
Nikola Zivanovic |
|
shankarsudhir
Posts: 6
Country : India | I noticed that I've submitted the wrong column in the Hard ABC Connect (instead of column A , it's the one next to A to the right). Any chance to verify if that's correct and acceptable ?
Good set of puzzles. (esp the mutant versions)
-Sudhir |
|
shankarsudhir
Posts: 6
Country : India | ignore my last post. Even with that, another mistake i made while submitting is with "E" instead of "C". No points for such carelessness :-)
Sudhir |
|
vopani
Posts: 739
Country : India | vopani posted @ 2010-10-18 10:11 PM Puzzles were exceptionally good, fun to solve, and in a competitive environment.
I liked Nurikabe (Hard), Tents (Hard), ABC Connect (Mutant), Tents (Mutant) and Minesweeper (Mutants) as the best 5.
I kept going wrong in Number Skeleton (Mutant).
I loved the Hard Nurikabe. It had a nice simple logical solving pattern, but the puzzle was just too cute for me to solve without smiling :-)
Thanks Thomas, for the wonderful weekend :-) |
|
MellowMelon
Country : United States | Boy is it a relief to hear how tough everyone found the Mutant Number Skeleton. It took me about twice as long to do as any other puzzle in the set during my testsolve, and I had to do trial and error. When I was told of the intended solution I figured I had just royally messed up, but it looks like I'm not the only one.
I commented to motris that experience solving nikoli puzzles, especially those that they publish online, would confer a substantial advantage for this test. Not surprising to see H.Jo win, with that in mind. |
|
motris
Posts: 199
Country : United States | motris posted @ 2010-10-19 4:00 AM MellowMelon - 2010-10-19 1:10 AM
Boy is it a relief to hear how tough everyone found the Mutant Number Skeleton. It took me about twice as long to do as any other puzzle in the set during my testsolve, and I had to do trial and error. When I was told of the intended solution I figured I had just royally messed up, but it looks like I'm not the only one.
I commented to motris that experience solving nikoli puzzles, especially those that they publish online, would confer a substantial advantage for this test. Not surprising to see H.Jo win, with that in mind.
It was certainly the outlier in your times, but was actually one of Wei-Hwa's faster solves (under two minutes, in line with the Hard of the same type ). I figured it was the most high variance of the mutants, since the solve is so dependent on noticing a property of the number list and does not solve in any other way like brute force which can be applied on a lot of the other puzzles.
The key insight to the Number Skeleton mutant is to see that all the 6 digit numbers have digits strictly increasing from left-to-right. This means as you follow a path of five-cell entries from top-left to bottom-right, the digits along that entire path must also be strictly increasing, but this places a lot of constraints on the digits at the intersections, particularly the first digit of the entry going down the rightmost column. Along this thinking you should be able to get that the intersecting 3 cell entries are 131 and 615 and from those get the other 3 cell entries which will help complete the puzzle. Edited by motris 2010-10-19 4:00 AM
|
|
vopani
Posts: 739
Country : India | vopani posted @ 2010-10-19 7:05 PM 6 Germans in top 10.
What do you have to say about that with WPC round the corner? ;-) |
|
rakesh_rai
Posts: 774
Country : India | Rohan Rao - 2010-10-19 7:05 PM
6 Germans in top 10.
What do you have to say about that with WPC round the corner? ;-) At the WPC, there cannot be more than 4 in the top 10 |
|
debmohanty
Country : India | rakesh_rai - 2010-10-19 7:46 PM
Rohan Rao - 2010-10-19 7:05 PM
6 Germans in top 10.
What do you have to say about that with WPC round the corner? ;-) At the WPC, there cannot be more than 4 in the top 10 Also, 3 of the four USA participants were involved in creating / testing the puzzles. |
|
forcolin
Posts: 172
Country : ITALY | motris - 2010-10-18 6:42 AM
The simplest answer - on a test like this which was pretty well-timed given the test-solvers - is to ignore time bonus and use raw clock time to break ties. That, or a much steeper drop off of the partial value was certainly needed. The partial time bonus was included to address the possibility someone might finish very early, but with an error, as has happened on past tests but that was not really a factor here. Certainly worth more improvements before the next time.
alternatively, you can simply go back to the original plan and award the maximum bonus only in the case of 10/10 puzzle solved. This also considering how kind you have been in forgiving typing mistakes, this reduces the impact of wrong transcriptions.
stefano |
@ 2010-10-18 6:42 AM (#2253 - in reply to #2252) (#2253) Top