Thank you to everyone who participated and to those of you who watched from the sidelines. If you are in the latter category, please do consider joining in next time.

It really was another great season.  We look forward to seeing you in the spring!

Best,

Helen

The minutes ticked by, and sure enough here comes Nick with pf3pushmixfull. It’s a significant improvement, and he’s got a few more right behind. But wait… hold the phone! Could it be? Yes, here comes Alfonso with a killer punch. BAM! lastttry01. Look out! Here comes another. BANG! lasttry02. And the final blow. KAPOW! lasttry03.

Alfonso has set a new standard in terms of dominating the end game. Not one, but two of his final entries improved on the previous leader by more than half a percent. Which is, as we say here in New England, wicked hard to do. So hats off and three cheers for Alfonso Nieto-Castañón, our latest champion. And, as always, thanks to everyone who played. The code that wins carries with it a little bit of all of you.

See you in six months!

The Grand Prize Winner will be declared after all entries submitted before noon (Natick time) are processed.  You can check the queue at the Vines Contest website to see how your entry is doing.

Helen

Curses!! I tried to hold my SOC code for as long as I could, but I was going to be away from the computer for the last hour before the deadline and I knew I hadn’t really kept up with the leader developments since I forked from “m” to develop my SOC solver, so there were a lot of tweaks that could probably quickly improve my code. I went so far as to make sure I could do the submission at the last minute from my phone. I submitted from my phone at 8 min before the deadline because I wanted to have time to log back in to matlabcentral, find my entry, and resubmit with the SOC tag, in case my phone had mysteriously logged out or I was in a dead zone or something. I was hoping 8 min was not going to be enough time for someone to find and insert a winning tweak. I should have known better :)

Congrats to Alex P, who only needed 7 min to find something to improve and resubmit it, winning by improving the time by 0.16 seconds!

His closing comment shows that Michael is a good sport and that he understands the brutal reality of the contest: tweaks can happen at any time. Alex P., who has narrowly missed on prizes of his own, deserves our recognition too. But I can’t resist showing you the one line out of 1675 lines where the code differs. On line 478, where Michael had

for n1=1:1e2,

Alex had substituted

for n1=1:22,

And that, my friends, is all it takes. Assuming your answer is actually better, which is harder than it looks.

There is one exception to the Iron Law of Tweaking, though, and it will soon be upon us. Once we get so close to the end of the contest that your code can’t be scored until after the deadline has passed, nobody can see and modify your code before the end. All the hardcore competitors (and who else is left by now?) are aware of this. They’ve reached deep into the shot locker for their last tricks. Now they are slowly circling the submit button, waiting… waiting for the moment to blaze away. I’m picturing the ships of the line at Trafalgar drifting slowly toward each other, almost within range. The slow match is burning. All is silent.

Good luck!

Congratulations Amitabh on your second prize for this contest!

Remember that midnight tonight (UTC) is the deadline for the Square One Challenge. Mark your entry with the naming sentinel SOC to be eligible. For more details, refer to the earlier blog post.

I ran a small analysis over the solvers that were among the first five positions at the end of Twilight. The leaders at that time were Nick Howe (Revenge of Antichaff), Alfonso Nieto (test05_tweak10), Robert Macrae (The Fragrant Honeysuckle), Kurt Janssens (v 11), and Martin F (Sixth shot).

First, let’s look at the solutions given by the solvers to the original Fifteen puzzle. My initial configuration placed all odd number at the top and even numbers at the bottom. I had to change the original puzzle a little because Kurt’s solver choked if I passed an initial board with a zero. I replaced that piece with a 16. Nick’s solution was able to interleave the even and odd numbers properly. Notice that there was a lot sliding and few pieces were blown out. Alfonso and Kurt seemed to concentrate on creating their vines and cared less about moving pieces. Robert and Martin preferred to move all the high value pieces to a given region before building a simple vine.

Let’s scale the problem a little bit. In my second board, I placed random pieces in a slightly different configuration. Higher value pieces are toward the bottom of the board. Note the excellent job by Alfonso’s solution – it appears to have the right balance between moving pieces and growing the vine. In a board like this, on one side you need to do a lot of sliding in order to bring the monotonically increasing pieces together. On the other side, you want to minimize sliding because “easy” sliding requires blowing pieces out that could eventually be part of your vine. More difficult sliding, like the one that actually occurs in the real fifteen puzzle where you have restricted space, is much harder to program. In this situation, Nick’s solver changed mode and operated more like Robert’s and Martin’s solutions. It moved pieces to the edges to build the vine. Note that Kurt’s solver did not move any piece at all, yet his strategy of focusing on growing a high value vine permitted Kurt to be among the leaders.

Finally, let’s see how the solvers behave with a random board with only ones and twos. Nick’s solver apparently changed strategy again. The new operating mode refuses to move the pieces. The solution is very similar to Kurt’s solver. Again, note that non-mover solvers are doing fine in several of the problems. Alfonso’s solvers seem to have found the right combination between moving and growing the vine, while Robert’s and Martin’s move all the high value pieces to the edges again. Notice the best scoring solvers left most of the twos out of the vine.

In closing, here’s a quick look at results of the current leader, at the time of writing this post.

First, we’ll do a Best Result contest. As usual with this prize, there is no penalty for time. If you can get it to run without timing out, then it qualifies. We’ll be looking at the “Result” column only of the “Best Result” table on the Statistics page. Deadline for this contest is midnight tonight, UTC, or 8 November 00:00:00 UTC.

Alan had a number of interesting, one might say poetic, suggestions for other mini-contests. My favorite: “Solvers that when viewed in the MATLAB editor form ASCII art, yet are still able to run.” I’m not sure how to score that one, so I’ll go with his first suggestion. We’ll call this one the Square One Challenge. You’re trying to get the best score possible given the constraint that you always have to start with square one (row = 1, column = 1, index = 1). As with the No Bogus Words Challenge in the last contest, you’ll need to signal your intent so we can determine the winner. Put the three letters SOC at the beginning of your entry’s name (e.g. “SOC Killer Kudzu”). We’re not changing the contest scoring code, so we’ll be checking your code after the fact. When the deadline has passed, we’ll look at the leading SOC entry and check it. Beware: if any of your vines don’t start on square one, you will be disqualified from winning the Square One Challenge on any other entries, and we’ll move on to the next potential winner. The SOC winner needs to be submitted before midnight tomorrow, 9 November 00:00:00 UTC.

For the second time in a row, Amitabh wins the Saturday Leap prize (see his badges here) with his Vines & Vines, which kept the lead for almost two hours. Amitabh’s remarkable Saturday contributions are especially obvious on our “Percent Improvement” chart. His name appears three times in the stretch following 11/05, or midnight UTC.

Sergey Y. (look at all these badges) takes the Sunday Push prize. You have to appreciate the extraordinary focus that Sergey brings to the Sunday Push. He knew what he wanted to do and he set about it very methodically, racking up a total of ten leading entries, mostly in the small hours of Sunday morning. His focus paid off. Nobody else came near him for the Sunday Push prize.

Well done, and congratulations Amitabh and Sergey! More MATLAB-themed booty will be headed your way. And thanks to all of you who spent your valuable discretionary weekend time trying to squeeze a few atoms of efficiency out of MATLAB program. It’s been fun to watch!

I was going to say “Oh Sebastian, you waited a little too long to submit your entry.” But, once again, close inspection reveals that his entry was actually based on Nick’s winning entry, Cheap Reverse, which wasn’t even scored until 20:58:50 UTC. So, far from being slow, I have to commend Sebastian for his nimble typing. He modified and resubmitted his runner-up (here’s the diff) in about 90 seconds. Nick’s entry was based on Pori 13 by Abhisek Ukil, which was based on -1 by Alex P. which was based on … well, you get the idea. A lot of people have contributed ideas by now.

Nick left us a nice comment in his winner:

This seems like just a tweak (and it is), but I only realized how effective it might be after spending two hours trying something much more complicated that in the end didn’t work nearly so well!

Funny how often that turns out to be the case. And that reminds me: be sure to look at the comments from time to time. There’s a lot of good stuff in there, and some of it’s very funny. You can leave comments on any entry at any time. You can ask questions, give congratulations, or propose new ideas. It makes the game more fun.

• Early Bird prize: Best score for an entry submitted before 21:00 UTC (5 PM Natick time) today
• Saturday Leap: Single best percent improvement for the entire day of 5 November (midnight to midnight, based on UTC time)
• Sunday Push: Most aggregate percent improvement for the entire day of 6 November (midnight to midnight, based on UTC time)

On a more functional level, we have another account that gives some indication of how healthy the queue is. Following @ContestStatus will tell you how long the queue is and how long it’s been since the last entry was processed.

What is the secret to his winning entry? You’ll have to wait until Daylight to see what tricks he’s using. In the meantime, you can work toward the Twilight Prize (Prince of Twilight?), which will be awarded tomorrow at 16:00 UTC (noon Natick time). For those of you looking ahead to other prizes, we’ll have an Early Bird prize on Friday afternoon (21:00 UTC), as well as our regular weekend prizes, the Saturday Leap and the Sunday Push. Have you got ideas for another mini-contest? Please tell us! We’re open to suggestions.

This season’s challenge is going to remind many of you of a favorite childhood game – the game 15-Puzzle! In this contest, use your MATLAB skills to create vines on a grid. Read Ned’s rules to learn more.

If you are new to our contests, take a look at If this is your first contest, you might want to read the Contest FAQ.  To be truly inspired, take a look at our Hall of Fame .

This contest runs from now through noon Boston time Wednesday November 9. If you have questions or comments, you may post them here or reply to this contest’s thread in Newsreader.

Good luck to everyone!

Helen and the Contest team

If you’ve never participated in a contest before, check out some of our previous contests.  You can also find out about folks who have won previous contests in our Hall of Fame.

Hope to see you online!

I think I can, without hyperbole, say that I have never seen anyone so thoroughly dominate the closing minutes of a contest. Look at this scoring cliff.

Amitabh’s Sneaker was on top for a surprisingly long time, given the closing onslaught. I thought maybe that tough old Sneaker might go the distance. It was a timing game: no one was resetting the result of 6877358, but the race was on to shave seconds from the clock. And then along came Nick, with one fist of iron and one fist of steel, smashing the score lower with each blow. When the dust settled, Rapid Weight Loss May Be Harmful was high atop a list of 2713 ranked entries.

Well done Nick. We send our congratulations to you (and your wife).

And well done everyone! Thanks for joining in the fun and making this contest such a pleasant experience. It’s a lot of fun for us at MathWorks to watch the pros go to work. We are amazed by the display of skill.

As always, now is a good time to reflect on how the contest. What did you like, and what would make it run better? How can we get more people to play? How can we get more people to watch? Some of these questions are already being discussed on the newsgroup. Join in the conversation there or leave a comment here.

Albert, who had a higher scoring NBW entry that came in just six minutes too late, deserves a prize for the most heartbreaking title: NBW just realized there was a deadline …. Ouch! We feel your pain, Albert. That’s a hard way to go. Albert’s entry was an interesting outlier in that it took almost 6 times longer than Alan’s winner but compensated with a substantially better result.

The queue is quiet, and yet I see the end of the contest is a short hour away. This is the calm before the storm. Algorithmic ammunition, long held in reserve, is being readied for the final battle.

This is our first darkened queue endgame, which should significantly change the dynamics. Soon the event horizon for inspecting competitive code will close. Then the floodgates will open and dozens of innovations will do battle for ultimate supremacy. Good luck everyone!

Enjoy!

Crossword Contest Commentary

Here are some snapshots from the latest contest.

Contents

Darkness Winner

Let’s start with Alfonso’s Darkness winner, tryagain01. We’ll run it against problem 41 of the test suite. It produces this picture.

tryagain01

Bogus!

Right away we notice that the output is mostly red. This reflects the fact that solvers are rewarded for packing in lots of short words as long as they create a smaller number of long crossing words. In retrospect, we made the bogus word penalty too low, since the result is aesthetically a little unappealing. But the rules are what they are, and within this framework, we see plenty of innovation. There’s nothing wrong with red if it gets you a good result.

Sample Words

Since these letters are meaningless, I’m going to switch to some real words to make it easier to follow what’s going on. Let’s consider the words in the first paragraph of the rules. I realize it’s a little unfair to compare codes based on an example they never encountered in the test suite, so you can take this exposition with a grain of salt.

YOU ARE THE NEW CROSSWORD PUZZLE INTERN FOR MAJOR MATH 

SOFTWARE NEWSPAPER THE NEW YORK MULTIPLY ALL THE NEWS THAT 

BITS AND INTS YOUR EDITOR HAS GIVEN YOU LIST WORDS 

USE THIS WEEK PUZZLE KNOWS YOU CAN POSSIBLY USE ALL 

THE WORDS HIS LIST GAVE EACH WORD WEIGHT AND TOLD 

YOU THE BEST YOU CAN YOU HAPPEN KNOW THAT GAVE 

THE SAME CHALLENGE ANOTHER INTERN YOU BETTER GOOD JOB YOU 

WANT KEEP YOUR POSITION MAYBE SOME SOFTWARE WOULD HELPFUL HERE 

Naturally, the character of the answer depends on the weights and the penalties. For this problem, let’s say that the weight is the length of the word.

Penalty = 0

For zero penalty and grid size of 15, Alfonso’s code returns this.

Penalty = 4

For a penalty of 4 (the median weight of the words in our dictionary), we get this.

Penalty = 9

The highest weight word is 9. When the penalty is 9, there are no more bogus words. Legal crossing words are completely absent in all three solutions.

Twilight Winner, Penalty = 0

Let’s move on to Nick Howe’s Twilight winner, Star Crossed X. This is the zero penalty solution.

Star Crossed X

Yow! That’s a jam-packed grid. Here’s what happens when we boost the penalty to 9.

Twilight Winner, Penalty = 9

Now there are no bogus words again, but still no usage of crossing words anywhere on the grid.

No Bogus Words

The other mini-contest winners from Saturday and Sunday that I looked at returned very similar results to those shown above. From the point of view of looking at “interesting” results, we need to skip ahead all the way to the No Bogus Words entries. For NBW entries, the penalty is effectively infinite.

This is NBW Weighty matters from the cyclist. We’ve finally got some crossing words! Three, to be precise, but that’s more than none.

NBW Weighty matters

Scrabble Scores

Suppose instead of rewarding solvers for the length of each word, we instead rewarded them for their Scrabble scores. The longest words used to be 9. But now the highest value word is PUZZLE, with a score of 26.

The Last Example: A Recent Leader

This is Oli’s NBW Hoola Hoop. It’s the highest ranking NBW entry as of Tuesday morning, Natick Time. We’ve still only got a couple of crossing words.

How Do the Pros Do?

As a quick taste of what is possible with these same words, here is the output from a free crossword grid construction program, EclipseCrossword. Here we are fitting 30 words into the same space where the leading contest entry can only fit 20. Again, it isn’t a fair comparison, since these two programs are working on different optimization problems. But it’s fun to compare.

Published with MATLAB® 7.12

And sure enough, he swooped in with seven tightly packed entries between 15:44 and 16:00 UTC, and now he’s at home counting his prize money. Or he would be if there were any prize money. Congratulations (again) to 1000 Node challenge winner Sergey Yurgenson. I have a sneaking suspicion that he’s really good at tuning up small cars, fixing watches, and maybe building sailing ships in bottles. In any event, he sure knows his way around a few hundred lines of MATLAB code. Here is his winner, CrossCat7. Nice work Sergey!

First of all, we’re bringing back the venerable 1000-Node Challenge. The 1000-Node Prize will go to the entry with the best score that

1. has 1000 or fewer nodes, and
2. is submitted before 16:00 UTC time tomorrow (19 April)

Remember, you don’t need to shorten your variable names or remove comments to win this challenge. That’s exactly why we made it a node-count prize and not a character-count prize. This is how you calculate your node count.

t = mtree(filename,'-file');
length(t.nodesize)

Sergey is the acknowledged master of the Short Form. Can you beat him?

Next, as suggested we’re going to try a No Bogus Words Challenge. If you want to enter, you need to signal it to us by putting the three letters NBW at the beginning of your entry’s name (e.g. “NBW My Super Duper Entry”). We’re not changing the contest scoring code, so we’ll be checking your code after the fact. When the deadline has passed, we’ll look at the leading NBW entry and check it. Beware: if it actually does use any bogus words, the author will be disqualified from winning the NBW prize on any other entries, and we’ll move on to the next potential winner. The NBW winner needs to be submitted before 21:00 UTC on 19 April.

(If we had it all to do over again, we’d probably have made the bogus word penalties higher. But you never know how things are going to go until you start the contest and see what entries come in.)

Finally, there’s the Six Million Dollar Man prize. If you can break 6,000,000 for your “result”, regardless of any other consideration, you win the prize (note that you don’t actually win $6,000,000). Naturally, you can’t exceed the time limit, but other that that, all that matters is the result. My friend Dave on the contest team says it can’t be done. I realize it’s a stretch goal. We’ll just throw it out there and see what happens.

Oh no! Contest time.
I have lost my spouse again.
His mistress: MATLAB.

Here at the Global Contest Headquarters we have a special place in our hearts for the MATLAB widows and widowers out there who tolerate our twice-yearly shenanigans. So it was delightful to hear from one, in poignant verse form no less! Perhaps it is no surprise that, in the name of marital harmony, we have been asked at various times over the years to keep the contests short and infrequent. Thank you for your patience Susannah, and now you’ll have a prize of your own to put on the mantel next to Nick’s.

The judges also send an Honorable Mention to Alan Chalker for his prodigious poetical outpouring, and especially his versification in MATLAB.

So many choices
It seems a shame to say that
Only one can win.

The Sunday Push prize goes to Fel for his “Dead End Strategies” series. Notice that Fel’s best entry for the day would have won the prize all by itself. The mysterious Fel revealed himself to be David Felguera in the last contest where he also did some weekend work, winning the prestigious Sailing Home Saturday Leap prize. Now he’s got both weekend prizes for his trophy room. I think that might be a first… Congratulations to Fel for the largest aggregate improvement for all of Sunday!

Watch the statistics page for information showing who’s doing the best as the weekend progresses.

We will also be awarding a prize to the author of the best MATLAB Programming Contest Haiku. Whereas most of the contest is a purely objective competition that rewards steely logic and intellectual firepower, the Haiku Challenge judging will be completely subjective and arbitrary. It’s just a fun exercise for your neglected right brain. The prize is certain to be cheap. Submit your entries in the blog comments.

Have a great weekend!

Nick had some good competition from James White and Victoria (not to mention Darkness winner Alfonso). But in the end, Nick’s Star Crossed X had more gas in the tank. Congratulations Nick!

His name seems familiar somehow… where have I seen it before? Ah yes! Alfonso was our Grand Prize winner for the Fall 2009 Color Bridge contest. Congratulations and welcome back to the winner’s circle Alfonso!