Ashish Kumar Regional Coded Snake
Denis Auroux Regional Battleships
Fred Coughlin Turning Fences
James McGowan Summon
Ko Okamoto Hidoku
Matej Uher Nurikabe
Prasanna Seshadri Statue Park
Riad Khanmagomedov Scrabble with Loop
Robert Vollmert Meandering Numbers
Swaroop Guggilam Fillomino

Now,how many did I get right ?! :-p

Three? I assume that you guessed right Denis, Prasanna and Riad.

You got 4/10, one better than last year.

Thank you!.That's precisely the target I had in mind(to better my previous's year guess score).Looking forward to the event this year.Already tuned myself into the marathon mood by having a go at this puzzle

http://www.gmpuzzles.com/blog/2015/03/birthday-loop-prasanna-seshad...

and battling with it for close to 7 hrs!!!though I have to admit that I failed to solve it in the end and called it quits........

The first Marathon in 2012 just added all the times together.
Then in 2013 and 2014 we had a better bonus system, where a rank bonus and a time bonus were added to try to normalize the puzzles more.

It didn't seem to work as well as it should though, so I plotted the bonuses on a chart to say what scores the top 100 solvers got for each puzzle. As you can see, each puzzle has a mostly straight line, and the gradient depends only on how many people solved it within an hour (which was between 33 and 212).

If a solver is roughly 10th in every puzzle, they will discard their Heyawake and Turning Fences times.
If a solver is roughly 100th in every puzzle, they will also discard their Heyawake and Turning Fences times.
But we want to score all these puzzles similarly, so that the worst 2 performances are discarded, not just the hardest two. Which needs the lines to look roughly similar.

The points available for the easiest puzzles are so much higher than for the harder puzzles. As an example, I was pleased to finish 17th for the Heyawake, and made many mistakes in the 4x4 Sudoku to finish 75th, but still my Heyawake bonus was lower and so was discarded.

If anything, this is the reverse to what we'd expect. If only a few people can solve a hard puzzle within an hour, perhaps they should get more bonus than people that solve an easy one? All of the other LMI tests offer more points for solving harder puzzles...


One starting suggestion is to remove the 'how many people finished within an hour' variable from the formula. Make it the same for each puzzle, e.g. 100, or 200, or the max solvers over all puzzles? It doesn't matter if there's now a small jump between a 59 minute score and a 61 minute score, everyone knows 60 minutes is the target.

(this also removes a paradox where I know that I shouldn't finish a hard puzzle at 59 minutes, I should delay submission until after 60 because I'll be boosting the other solvers time bonus and rank bonus, and I'll just be discarding my own score later anyway)

I understand it is late for this year's Marathon, but I think this needs to be discussed at some point.



Edited by kiwijam 2015-03-04 2:34 PM






Attachments
----------------
Marathon2014Bonuses.png (50KB - 3 downloads)

I agree with you kiwijam. Now it seems, that it´s better to solve easier puzzles than harder one, because this puzzles should be more probably discarded. Your suggestion is fine, but it´s hard to predict number of solvers.

And the other problem is, that players doesn´t solve some puzzles in real time. They only print them and solve "how they can". For example, if I do not like scrabble puzzles, then I only print it and solve for a day (not because, i´m not able o do it faster, but only because I know, that I will solve it longer than 1 hour). So my time is not around two hours (as I solve it), but around 24 hours. And if you want to use formulas, where same number of solver appear, my 2 hour time should change the amount of points, if I should solve in real time. So the trouble is with people, whose do not solve "in real time".

What about faktors for every puzzle, to be used to multiplying the result? For example the hardest puzzle should have factor 2 and easiest puzzle should have factor 1. Or anything similar to that. But faktors should reduce the differences between puzzles at least for some players.

You don't need to predict the number of solvers. For example, after last year we could use n=212 for all ten puzzles. Then the rank bonus is the same amount for each puzzle.

Based on the inputs in this thread, we found that the hardest puzzle indeed gets eliminated for a majority. In effect, it is true that the hardest puzzle has minimal say in the rankings, especially near the top of the standings. So we are coming up with a revised structure for points table as below. Do share your thoughts on this. This uses the easiest puzzle as a reference. We'll review post-marathon if this addresses the problem or not. (Changes are in red)

I am not able to understand the difference between:
1. The top ranked player for the particular puzzle; and
2. The best player for the particular puzzle

No difference. They are the same.

That means 100 is the maximum score a person gets (by the person who solves fastest correctly) : 50+25+25. Am I getting it right?

Absolutely.

It looks much improved, Rakesh, thank you.

Good luck everyone, have fun!

Where is the puzzle? I can't find it

Click on on each puzzle to download the pdf.

I took my printout of Fillomino and there are absolutely no grid lines in between except on the borders....Now I need to draw lines with a scale in between

I would advise the LMIans to prefer the one with the solid lines as my printer had a problem with the former and completely missed printing the dotted lines in between.I would stop myself short here by saying that the Fillomino puzzle by itself is absolutely delightful and perfect!!!Good luck to all!

Fillomino seems to suffer the slightest ambiguity at the very end, and not affecting the key (unless I somehow messed up): R13C2 and R13C3 contain some permutation of 20,1.

R13C2 and R13C3 have no ambiguity.

Why is Hidoku listed with 4 players when it hasn't been uploaded yet?

Is there anyway to give up on a puzzle once started? I am finding the Nurikabe impossible to solve logically (certainly not "easy") and trial and error is not much fun.

Finally got it after a lot of erasing. I guess I'm used to Nikoli nurikabes that never seemed that hard. Need to figure out new theorems or guess better.