Thanks James and Nil for your feedback, and congrats for good scores.
We are certainly going to repeat the format, so we'll discuss about potential changes in the scoring system in the end.


Message to all players who have stopped solving at 7 or 8 puzzles. Please note that the rank that you see now in the partial score page will take a huge dip, you don't get to at least 9 puzzles. You still have 30 hours to participate further.

Was it really mentioned somewhere before the contest? I haven´t seen it until now. I stopped solving seriously after spoiling first puzzle.

Meh, I'll edit it back in myself later.

Or you can.

I suppose it would be wrong to mention that the Samurai Sudoku uses the digits 1 through 9? (Note I haven't solved it yet, or tried...will this evening perhaps.)

Random question. Why are typical smiley codes not used here?

Cyclone


Edited by Cyclone 2012-01-29 12:47 AM

See the Instruction Booklet "Scoring" paragraph, at the end, "Players who solve all 10 will have their worst puzzle discarded" or something like that.

Edited by prasanna16391 2012-01-29 12:59 AM

It is mentioned in "Scoring" section in the IB. See last line. And it was also mentioned in forum, while answering to kiwijam's questions.

The forum didn't have all smileys inbuilt. We added some smileys, but yes, it is not complete. Some other forums (e.g. phpBB) have a more exhaustive list of smileys.

To "solve" the instructions is always the hardes puzzle for me

Just like you should always solve the example of a new puzzle type, you should also "solve the instructions" of a new type test

Phew! Just finished! It was tough but enjoyable. Thanks to Deb and the team for the huge amount of work done and for a very attractive formula. I notice that in spite of the length of the contest the distances between players are minimal (which means that the 10 minutes I trew away stupidly on one puzzle for not reading the instructions costed me several positions....). Also glad I could end in bonus position on 9 puzzle out of 10 (well, I like to leave space for further improvement for next time....). Congratulations to all those which finished at the top, and relized times which are a bit embarrassing for the average player....

The formula is almost perfect. I feel I can support an improvement among those proposed. I do agree with detuned, when I says



perhaps next time the most difficult puzzles will not be of the same type of this time.... however I feel that life has been particularly cruel for a player which solved one of the difficult puzzles in, say, 59'55" and got 0.1 bonus points, i.e. only .1 point more of a player who solved it in 2 hours, particularly if only a handful of solvers could finish that puzzle with the bonus. so the formula 45/60/75 (and perhaps 90), although arbitary, is a good idea and I support it.

while I do not agree with Motris

The formula relies on the possibility, which many players have used, to stop solving a puzzle, putting it in a corner and solve it the next day or later once the possibility for the bonus has been lost. In this case the concept of normalized time is heavily affected and meaningless because the number of player doing this will vary from puzzle to puzzle, depending on the difficulty. Also, a player needs to know in advance the criteria of awarding the bonus.

I renew my thanks to the organizers and authors and look forward for the next one of these.
Stefano

Stefano, I was trying very hard to not give a specific formula, because I think it is more important to speak to the goals of a scoring formula and match them to the test. As a contest constructor I will say that it is very hard before the test, when you've seen just 2 or 3 solvers, to put a fair value on all of the puzzles, particularly with complicated bonus. Also, when having so many constructors (7), it will be very hard to have consistency of difficulty. To say puzzle X should be 75 or 45 might be better, but it might not be correct. And if the puzzle you made into a 75 turned out much easier than a 60 then you are really screwing up the test results by using a guess, and not a more robust formula.

A simple formula might be "top time + 30 minute window" gets some bonus, like what Nilz mentions, which would still have an effective one hour cut-off. I personally prefer something like this:

Solvers in the top 10th percentile earn bonus. Specifically, the top solver earns 100 + X points, solvers in top 10th percentile earn bonus proportion to time gap from 100th to 90th percentile. Solvers at 90th and below get 100 points. With the results I'm seeing on this test, there still is a stable "one hour is too long, so you can start over whenever" effect, and actually for many puzzles the time is now more like 40 minutes. But now you let the competitors scale the times, not 2 or 3 test-solvers.

I'm a bit frustrated that my best relative puzzle performance on this test is the puzzle I have to drop. I'm not frustrated enough because of how I did on other puzzles, but I want a good system that works in general for this kind of test but maintains the goal of removing pressure after one hour. I probably should not have mentioned croco-puzzle specifically, because I did not mean a 3000 to 0 scale. I meant something more like a 3000 to 2700 scale where every solver below 2700 goes up to 2700 whenever they finish. I hope something like this would meet your approval, because my goal is not to ruin the benefit of having comfort to sleep on a mistake and start again the next day.

Edited by motris 2012-01-29 11:49 AM

Feeling very happy after solving all the puzzles :)
Honestly I did not think I could make this far.
I should have been aggressive and solved few of them on paper (actually solved all of them using MS Paint) at least that would have fetched some bonus.

Thanks to Deb, LMI team and the authors for a wonderful contest.
I hope everyone will complete all the puzzles.

Regards,
Ravi

I really enjoyed the puzzles, thanks to all the authors for the fun! Given the differences in difficulties, I also think it would make sense to normalise the puzzles based on observed performance.

Also congrats Macher, I also used mspaint for everything and it can be annoying at times. I am now trying Paint.NET which is similar but has layers and unlimited undos - useful for when you want to *cheat*!

Good to know there are other people who have to struggle with paint during these tests too

Score page is under maintenance, and will be back online soon.

Submission is now disabled for Puzzle Marathon.

Puzzle Marathon is now over. Congratulations to motris, MellowMelon and Kota for winning.

Individual puzzle wise, there were 6 winners. Volxa (Kakuro), motris (Loop The Loops, Samurai Sudoku, Braille Word Search, Pentomino Areas), xevs (Graffiti Snake), Para (Black And White Loop, Small Regions), ACM (Tapa)

In a turn of events, I have many individuals to thank to for this event.
Thanks to Rohan, Serkan and Zoltan to first listening to the idea and believing in it, and then adding all the required inputs. The scoring system, the puzzles types could have been very different without their suggestions.
Thanks to all puzzle authors : Tom (Kakuro), Prasanna (Loop The Loops), Serkan (Graffiti Snake), David Millar (Braille Word Search and for the logo), Zoltan (Black And White Loop, Small Regions), Vladimir (Pentomino Areas and Tapa). I wrote Different Neighbours and Samurai Sudoku, and unfortunately the difficulty of Samurai was not appropriate.

I hope everyone enjoyed the test.

Link to score page : http://logicmastersindia.com/M201201P/score.asp
If your name is missing in the score page, it is intentional. Please PM me for clarification.

Thanks for some great puzzles. Regarding Samurai: I had figured the IB version was just going to be much harder than the actual one. If it had been mentioned that the two would be similarly hard, I could have pointed out there would be a problem. I'm sure many others could have too.

I find it funny how my individual puzzle rankings resemble my WPC playoff ones. Here I didn't win any of the puzzles but my time was in the top 5 for all of the types I care about (sudoku and word searches are bleh). Similarly in the WPC playoff: on every puzzle I had the 2nd, 3rd, or 4th fastest time of the competitors that finished, but never the fastest. (which reminds me that I got sent the time breakdowns in an email with a promise that it would be posted publically soon after, but I guess that slipped someone's mind)
Apparently I'm better at being consistent than having incredible times on individual puzzles.

One thing I would like to discuss that wasn't in the feedback is answer extraction. Obviously some of these methods are tried and true (Sudoku, Star Battle), and I continue to find the "longest group" method very nice as well (Tapa, Snake, Loops) especially given its advantages in an instant grading system.

I haven't seen the Kakuro and Different Neighbors style before, but that might be one of the cleanest of any mechanisms I've seen. Finding marked squares is often a pain, and in November's test it was even common to miss some I's/J's. This one-per-column system with all of those columns marked solves that issue in an extremely nice way. I hope this one is adopted by all future tests where appropriate.

The Word Search and Pentomino Areas extractions were fine - not amazing but not at all bad. Black and White Loop was the only mechanism that really bothered me. It's a good method for smaller grids, but keeping track of things on the large grid was tricky. I was half expecting to get the red X on my first submission for that one (the time penalty wasn't enough to justify further checking).

Thanks for the test. Awesome, as always.

Funny story: Even though Graffiti Snake was one of my best puzzles, relatively speaking, one piece of logic I'd been using was completely bogus: Since the head and tail were in opposite corners, the snake has to cross each line an odd number of times. So, for example, if there are an odd number of clues in a column, exactly one of the black bars has to touch the edge of the grid, so that there aren't an even number of gates. However! Since a snake could enter the gate and leave on the same side it entered, that wouldn't count as a crossing. Therefore, bogus logic. (An example of a column that fails my test would be the one with 5-3-3.)

I hope I explained that well enough. It just means that I got really lucky.

Edited by figonometry 2012-01-30 8:31 AM

Reg answer key mechanisms :

Kakuro and Different Neighbours - During my solve, I couldn't find any row / column / contiguous cells to be marked as answer key. (Kakuro for being so scattered solve and DN for being DN). When I shared the circled cells with one circle per column idea to Tom, he instantly agreed that this will be a cleaner approach. I just adopted the same for DN.

Zoltan was never a big fan of marking letters inside the grid for B&W Loop. I was not keen on marking "rows with longest length". I have been seeing how players are making mistakes in the longest length mechanism for a while now. It was my decision to mark letters inside the grid, rather than arrows outside the grid - based on this poll results - http://logicmastersindia.com/forum/forums/thread-view.asp?tid=358
Yes, for a big grid, it is not an easy key, but the other way I looked at it, when you are trying to find the answer key, and you don't visit some letters, that means you definitely have a mistake in the solve.

We discussed about some potential scoring systems. I'm attaching the individual solving times (in excel form) so that it will be easier to do any kind of simulations.
For example : bonus should be given only to players with solving time < 30 minutes + best time is captured in the excel.

Link : http://logicmastersindia.com/M201201P/MarathonSolvingTimes.xlsx

Edited by debmohanty 2012-01-30 10:50 AM

For Snake, it was a problem. I found many players submitting "longest black cells", not "longest snake part".
I decided not to penalize players for making this answer key misunderstanding, but their bonus was not adjusted.

That reminds me to add a note that, during Decathlon we put lot of effort to make sure that all possible answer keys are captured in the system (e.g. row / column swaps). I didn't have to do similar effort in this case because of "single puzzle" submission mechanism.

All puzzles are pinned and uploaded as a single pdf without password here

Thanks to all who participated, and to Deb for allowing me to contribute a puzzle and design the logo for this test.

If you have an account on my blog, The Griddle, and solved my braille word search in this test, enter the two missing words in your code entry page to unlock the new "Run For Your Life!" badge: http://thegriddle.net/home/badges/18

I noticed that some of the authors have participated as players too. They have been awarded an arbitrary 100 points for the puzzle they provided. I believe this is penalizing. In a competition in which the difference between players is given by the bonuses (which is good, and is the strongest point of this beautiful competition), it is equivalent to deny them the opportunity to earn a bonus in the puzzle they provided (for which they are supposed to be strong solvers too...). In my opinion it could be fair to award the author of a puzzle a bonus score equivalent to the average of their bonus scores obtained in the remaining puzzles solved (including zeroes, if any).

I agree that 100 points is not fair to all authors (it was not arbitrary though, 100 was assigned based on the assumption that an author should be able to solve his puzzles, and we just took the worst case scenario where there is no bonus). So, yeah, it is bit harsh.

In fact we started "best 9 of 10" to handle authors' cases. But later, 9 of 10 turned out to be very essential for this kind of contest even otherwise. So, we can give some bonus points based on forcolin's formula above. Or, we can say for authors, it should be best 8 of 9, and then multiply the score by 9/8.

Is it a good idea to have an option where a player can choose if he/she needs a hint to solve a puzzle at the cost of some X points depending on the time lapsed from the start of puzzle along with reduced bonus system (if applicable) ?
This might help few solvers to have the satisfaction of solving all the puzzles and create more interest in puzzles.

Regards,
Ravi