Thread Subject:

Fall 2010 MATLAB Contest, November 10 - 17th

This season's contest, Sailing Home, is officially underway! You are now a captain and your assignment is clear. You can find out more about the assignment in the contest rules at http://www.mathworks.com/matlabcentral/contest/contests/30/rules .

Please use this thread for questions or discussion about the contest.

Good luck to everyone!

Helen and the Contest Team!

Looks like fun!!

Seems like the problem is exactly solvable in O(n^4) time, which should make it easier than most contest problems, which as far as I can remember usually haven't been poly-time solvable. (Not that solving it exactly may be the most practical.)

hmm.... are you sure it's poly time solvable? I haven't given it a much thought, but the frictionless aspect seems to complicate thinks massively....

"Pete Stobbe" <stobbe@acm.caltech.edu> wrote in message <ibf0uv$9ej$1@fred.mathworks.com>...
> Seems like the problem is exactly solvable in O(n^4) time, which should make it easier than most contest problems, which as far as I can remember usually haven't been poly-time solvable. (Not that solving it exactly may be the most practical.)

"Helen Chen" <helen.chen@mathworks.com> wrote in message <ibejc2$bf1$1@fred.mathworks.com>...
> This season's contest, Sailing Home, is officially underway! You are now a captain and your assignment is clear. You can find out more about the assignment in the contest rules at http://www.mathworks.com/matlabcentral/contest/contests/30/rules .
>
> Please use this thread for questions or discussion about the contest.
>
> Good luck to everyone!
>
> Helen and the Contest Team!


Does the rule: "no function handles" mean we can't use bsxfun?

If we can use bsxfun what are the limitations? Would @times
be okay but not @(x,y)x.*y

Thank you,
-Sean

"Helen Chen" <helen.chen@mathworks.com> wrote in message <ibejc2$bf1$1@fred.mathworks.com>...
> This season's contest, Sailing Home, is officially underway! You are now a captain and your assignment is clear. You can find out more about the assignment in the contest rules at http://www.mathworks.com/matlabcentral/contest/contests/30/rules .
>
> Please use this thread for questions or discussion about the contest.
>
> Good luck to everyone!
>
> Helen and the Contest Team!

Shouldn't there be a link to the current contest page (http://www.mathworks.com/matlabcentral/contest/contests/30) from the main contest page? It's not especially obvious how to get there if you click on the Contest link bar button...

I enjoyed being able to finally put some faces to the names of the contest team with the behind the scenes video. Thanks for putting that up!

"Sean " <sean.dewolski@nospamplease.umit.maine.edu> wrote in message <ibf9o8$8su$1@fred.mathworks.com>...
>
> Does the rule: "no function handles" mean we can't use bsxfun?
>
> If we can use bsxfun what are the limitations? Would @times
> be okay but not @(x,y)x.*y
>
> Thank you,
> -Sean

Hi Sean,
I believe that the restriction on function handles means you can not use bsxfun.

Good luck,
Seth

FYI, I couldn't get the runcontest.m file to execute because MATLAB couldn't interpret the Copyright statement on the first line. Removing it got it to work.

"EBS " <ericDOTsampson@gmail.com> wrote in message <ibfecc$21u$1@fred.mathworks.com>...
>
> Shouldn't there be a link to the current contest page (http://www.mathworks.com/matlabcentral/contest/contests/30) from the main contest page? It's not especially obvious how to get there if you click on the Contest link bar button...

Thanks for pointing that out. I have added a link from the main contest page to the current contest page.

Seth

"Alan Chalker" <alancNOSPAM@osc.edu> wrote in message <ibfprs$aln$1@fred.mathworks.com>...
> FYI, I couldn't get the runcontest.m file to execute because MATLAB couldn't interpret the Copyright statement on the first line. Removing it got it to work.

Hi Alan,
We caught that problem this afternoon and uploaded a new version of the file. I just double checked and the corrected file is availble on the file exchange.

Seth

Hi Seth,
On the contest page, the following is mentioned "This formula results in the lowest (lower is better) score when your ship stops on point A" . Am I correct in my understanding that the ship should STOP at point 'A' when it is returning from B to A? In that case, reverse velocity needs to be applied when the ship reaches point A, else the ship will not stop at point A. In the example given on the contest page, the ship reached point A on 'Turn 8' but does the ship stop there? So, journey is not complete on Turn 8.

I feel the requirement is that the ship should stop at Point A and not just REACH point A on its way back from point B to point A. Please correct me if I am wrong.
-Chintan

"Michael Bindschadler" <mikeREMOVETHISbind@gmail.com> wrote in message <ibemfc$5mn$1@fred.mathworks.com>...
> Looks like fun!!

It looks like somebody put a lot of effort into the visualization, but personally, I find it kind of hard to look at and interpret visually. A simple alternative is to just use a quiver plot along the lines of:

rwind = chart(:,:,1);
cwind = chart(:,:,2);
quiver(cwind,rwind); %note that cols correspond to the x direction and rows to the y
axis ij;
axis equal;

"Chintan " <t_chintan2@yahoo.com> wrote in message <ibft30$5aa$1@fred.mathworks.com>...
> Hi Seth,
> On the contest page, the following is mentioned "This formula results in the lowest (lower is better) score when your ship stops on point A" . Am I correct in my understanding that the ship should STOP at point 'A' when it is returning from B to A? In that case, reverse velocity needs to be applied when the ship reaches point A, else the ship will not stop at point A. In the example given on the contest page, the ship reached point A on 'Turn 8' but does the ship stop there? So, journey is not complete on Turn 8.
>
> I feel the requirement is that the ship should stop at Point A and not just REACH point A on its way back from point B to point A. Please correct me if I am wrong.
> -Chintan

The scoring function only takes into account the distance from the last point the ship visits to point A, so the ship does not have to have zero velocity to get a zero distance at that last point. That's the way it looks to me, anyway.

"Michael Bindschadler" <mikeREMOVETHISbind@gmail.com> wrote in message <ibftlu$dkr$1@fred.mathworks.com>...
> "Chintan " <t_chintan2@yahoo.com> wrote in message <ibft30$5aa$1@fred.mathworks.com>...
> > Hi Seth,
> > On the contest page, the following is mentioned "This formula results in the lowest (lower is better) score when your ship stops on point A" . Am I correct in my understanding that the ship should STOP at point 'A' when it is returning from B to A? In that case, reverse velocity needs to be applied when the ship reaches point A, else the ship will not stop at point A. In the example given on the contest page, the ship reached point A on 'Turn 8' but does the ship stop there? So, journey is not complete on Turn 8.
> >
> > I feel the requirement is that the ship should stop at Point A and not just REACH point A on its way back from point B to point A. Please correct me if I am wrong.
> > -Chintan
>
> The scoring function only takes into account the distance from the last point the ship visits to point A, so the ship does not have to have zero velocity to get a zero distance at that last point. That's the way it looks to me, anyway.

I responded to Chintan's other thread as well, but I'll state it here as well.

I looked through the scoring mechanism, and I believe you are correct. The final velocity is not part of the score, so you are aiming for a final point to minimize the distance from A.

We will update the contest rules to clarify (but that might not be until Thursday morning.)

Seth

On the same note. Does the ship have to end a move at "B"
Or can it just pass through it?
For example starte a move in the square below "B".
And then move 2 squares up, crossing B, but not actually ending a move there. Will this count as having reached B?

I don't think so, but wanted to make sure.


"Seth Popinchalk" <Seth.Popinchalk.nospam@mathworks.com> wrote in message <ibfu6j$goc$1@fred.mathworks.com>...
> "Michael Bindschadler" <mikeREMOVETHISbind@gmail.com> wrote in message <ibftlu$dkr$1@fred.mathworks.com>...
> > "Chintan " <t_chintan2@yahoo.com> wrote in message <ibft30$5aa$1@fred.mathworks.com>...
> > > Hi Seth,
> > > On the contest page, the following is mentioned "This formula results in the lowest (lower is better) score when your ship stops on point A" . Am I correct in my understanding that the ship should STOP at point 'A' when it is returning from B to A? In that case, reverse velocity needs to be applied when the ship reaches point A, else the ship will not stop at point A. In the example given on the contest page, the ship reached point A on 'Turn 8' but does the ship stop there? So, journey is not complete on Turn 8.
> > >
> > > I feel the requirement is that the ship should stop at Point A and not just REACH point A on its way back from point B to point A. Please correct me if I am wrong.
> > > -Chintan
> >
> > The scoring function only takes into account the distance from the last point the ship visits to point A, so the ship does not have to have zero velocity to get a zero distance at that last point. That's the way it looks to me, anyway.
>
> I responded to Chintan's other thread as well, but I'll state it here as well.
>
> I looked through the scoring mechanism, and I believe you are correct. The final velocity is not part of the score, so you are aiming for a final point to minimize the distance from A.
>
> We will update the contest rules to clarify (but that might not be until Thursday morning.)
>
> Seth

Gebert:
Only ending points of each move are considered, for both "A" and "B".
Lucio

"Gerbert Myburgh" <gmyburgh@webmail.co.za> wrote in message <ibg423$q3$1@fred.mathworks.com>...
> On the same note. Does the ship have to end a move at "B"
> Or can it just pass through it?
> For example starte a move in the square below "B".
> And then move 2 squares up, crossing B, but not actually ending a move there. Will this count as having reached B?
>
> I don't think so, but wanted to make sure.
>
>
> "Seth Popinchalk" <Seth.Popinchalk.nospam@mathworks.com> wrote in message <ibfu6j$goc$1@fred.mathworks.com>...
> > "Michael Bindschadler" <mikeREMOVETHISbind@gmail.com> wrote in message <ibftlu$dkr$1@fred.mathworks.com>...
> > > "Chintan " <t_chintan2@yahoo.com> wrote in message <ibft30$5aa$1@fred.mathworks.com>...
> > > > Hi Seth,
> > > > On the contest page, the following is mentioned "This formula results in the lowest (lower is better) score when your ship stops on point A" . Am I correct in my understanding that the ship should STOP at point 'A' when it is returning from B to A? In that case, reverse velocity needs to be applied when the ship reaches point A, else the ship will not stop at point A. In the example given on the contest page, the ship reached point A on 'Turn 8' but does the ship stop there? So, journey is not complete on Turn 8.
> > > >
> > > > I feel the requirement is that the ship should stop at Point A and not just REACH point A on its way back from point B to point A. Please correct me if I am wrong.
> > > > -Chintan
> > >
> > > The scoring function only takes into account the distance from the last point the ship visits to point A, so the ship does not have to have zero velocity to get a zero distance at that last point. That's the way it looks to me, anyway.
> >
> > I responded to Chintan's other thread as well, but I'll state it here as well.
> >
> > I looked through the scoring mechanism, and I believe you are correct. The final velocity is not part of the score, so you are aiming for a final point to minimize the distance from A.
> >
> > We will update the contest rules to clarify (but that might not be until Thursday morning.)
> >
> > Seth

In the Hall of Fame there is “Concours de la Toussaint” contest. What is that?

The new runcontest.m has a very nice built-in mechanism to limit the boards your run against (by feeding a vector of board nums as the second variable). I've found it easier at this point to concentrate on the small boards (which visualize better), so I made the following list that shows the board size / num in increasing order (size is first col, num is second col).

         121 162
         154 77
         165 196
         196 58
         240 123
         247 82
         247 85
         255 37
         289 131
         304 17
         324 166
         336 168
         336 197
         342 107
         352 191
         361 16
         361 49
         380 133
         396 179
         400 20
         400 78
         400 153
         408 27
         420 111
         420 145
         437 73
         440 101
         440 151
         441 100
         441 113
         441 175
         475 200
         480 129
         484 171
         494 188
         500 81
         529 70
         529 143
         540 186
         550 75
         572 8
         576 24
         588 126
         616 46
         650 118
         676 15
         676 19
         682 7
         700 149
         702 154
         729 181
         768 79
         775 156
         784 86
         837 189
         841 33
         858 150
         900 35
         900 121
         928 184
         961 173
        1024 99
        1024 190
        1088 56
        1089 76
        1140 97
        1140 106
        1156 125
        1156 195
        1184 67
        1225 14
        1225 31
        1225 88
        1258 182
        1271 95
        1287 42
        1292 28
        1296 29
        1326 62
        1330 122
        1406 159
        1443 130
        1444 50
        1452 98
        1480 51
        1482 66
        1505 157
        1512 61
        1521 48
        1521 138
        1521 140
        1548 165
        1599 45
        1600 10
        1600 32
        1600 117
        1638 90
        1656 92
        1681 115
        1710 93
        1755 9
        1760 128
        1760 198
        1764 5
        1764 40
        1764 64
        1764 96
        1764 139
        1806 41
        1849 134
        1862 177
        1886 34
        1900 84
        1924 23
        1936 148
        1960 102
        1978 60
        1989 54
        2021 13
        2021 38
        2024 169
        2025 87
        2040 160
        2040 174
        2050 158
        2091 144
        2106 65
        2116 142
        2145 55
        2145 199
        2150 26
        2150 83
        2150 116
        2160 146
        2184 119
        2184 187
        2200 194
        2209 104
        2268 161
        2303 68
        2304 18
        2304 47
        2304 91
        2304 137
        2352 36
        2401 74
        2436 1
        2444 167
        2448 4
        2464 21
        2475 59
        2478 141
        2491 30
        2500 147
        2601 43
        2601 178
        2610 39
        2610 135
        2640 94
        2640 127
        2695 124
        2703 103
        2704 170
        2726 114
        2756 11
        2790 164
        2805 72
        2808 57
        2809 132
        2867 176
        2891 3
        2914 44
        2914 53
        2914 109
        2915 2
        2928 69
        2940 80
        2961 71
        2961 112
        2964 110
        2990 172
        3213 136
        3233 120
        3328 163
        3339 63
        3380 105
        3392 152
        3417 185
        3465 183
        3596 192
        3600 25
        3600 108
        3658 22
        3660 155
        3752 6
        3844 12
        4096 180
        4248 52
        4340 89
        5332 193

Looks like Nicholas is winning Darkness ;)

FYI, in case everyone didn't notice it, there is a significant change to the rules:

"To keep the queue moving smoothly, you are limited to submitting no more than five files every 15 minutes, for an average of three minutes per file. If we find you are creating multiple accounts in order to get around this limitation, we may disqualify all your entries"

Of course I'm a little bummed about this, but oh well. Is there anyway to have this restriction be 'reset' if the queue is completely empty? Of course I understand the logic behind it, but for those of use who tweak a lot during 'off hours' when nobody else is doing anything, it unnecessarily restricts things.

For example, if the average code run time is ~60 seconds (which is typical of past contests), and nobody else is submitting codes to the queue, I could submit 5 entries in a row at say 2:00AM, have them finish running at 2:05AM, and then have to wait another 10 mins to submit anything while the queue is completely empty and nothing is happening.

By resetting everyone's restrictions at that 2:05 mark (with the queue empty), I can then submit another 5 entries (making good use of the contest machinery), and anybody else who happened to come along and submit an entry would only have to wait 5 minutes for it to run (or a max of 15 mins if my entries all timed out)

"Sergey Y." <ivssnn@yahoo.com> wrote in message <ibh17r$qom$1@fred.mathworks.com>...
> In the Hall of Fame there is “Concours de la Toussaint” contest. What is that?

Hi Sergey -

Good eyes! That is an error on the scripting of that page. We are doing some experiments with different types of contests and that should not have displayed publicly. I have removed that from the Hall of Fame.

Thanks for pointing that out. :-)

Helen

"Helen Chen" <helen.chen@mathworks.com> wrote in message <ibh86v$cqa$1@fred.mathworks.com>...
> "Sergey Y." <ivssnn@yahoo.com> wrote in message <ibh17r$qom$1@fred.mathworks.com>...
> > In the Hall of Fame there is “Concours de la Toussaint” contest. What is that?
>
> Hi Sergey -
>
> Good eyes! That is an error on the scripting of that page. We are doing some experiments with different types of contests and that should not have displayed publicly. I have removed that from the Hall of Fame.
>
> Thanks for pointing that out. :-)
>
> Helen


Than you may want to fix another bug:
If one select Player and then go to “See other contests NN has participated in” then one will see current standing even during Darkness phase. (I was able to see that Nicholas was leading)

"Sergey Y." wrote
> If one select Player and then go to “See other contests NN has participated in”
> then one will see current standing even during Darkness phase. (I was able to
> see that Nicholas was leading)

Thanks for telling us this Sergey! We'll fix that.

Alan Chalker wrote
> FYI, in case everyone didn't notice it, there is a significant change to the rules...
> Is there anyway to have this restriction be 'reset' if the queue is completely empty?

This is the kind of rule change where we are very interested in hearing what people prefer. We might well change it again for the next contest, but since we had significant problems with a backed up queue last time, we decided to limit the input rate to the worst possible processing rate (fail after timeout). That way, one person can't singlehandedly clog the machine. So these are the rules for this contest. We won't change them now that we've started. But we will be happy to reconsider how to manage the queue for the next contest. So let us know what you think works best.

I have been able to figure out the scoring formula and am posting it here as I traditionally do. As usual, it’s very similar to the recent contests:

score = k1*result + k2*e(k3*runtime) + k4*max(complexity-10,0) + k5*nodes

Where:

k1 = 0.2
k2 = 0.01
k3 = 1/11 (0.090909)
k4 = 1
k5 = 0.001

It's a little premature to do a leading entry analysis, since we are still in Twilight, but I'll give it a shot:

The current leading entry has a time of 25s, result of 8085, cyc of 40, and nodes of 2446. Here’s a breakdown of the current tradeoffs:

-cyc and score are a 1:1 ratio (i.e. each point shaved off cyc is a point shaved off the score)
-time and score are a 1:0.1 ratio
-result and score are a 1:0.2 ratio
-node and score are a 1:0.001 ratio

The entry isn't anywhere close to the ‘knee’ of the time exponential curve, which is rather flat until about ~60s.

Personally I like this new rule.
It still gives you A LOT of submissions. 480 in a 24 hour period is far more than most people submit in the entire contest time.

It leaves you with enough to work with to do some tweaking, but prevents you from doing silly tweak bombing.

"Alan Chalker" <alancNOSPAM@osc.edu> wrote in message <ibh7o0$bqd$1@fred.mathworks.com>...
> FYI, in case everyone didn't notice it, there is a significant change to the rules:
>
> "To keep the queue moving smoothly, you are limited to submitting no more than five files every 15 minutes, for an average of three minutes per file. If we find you are creating multiple accounts in order to get around this limitation, we may disqualify all your entries"
>

"Gerbert Myburgh" <gmyburgh@webmail.co.za> wrote in message <ibii5f$4ll$1@fred.mathworks.com>...
> Personally I like this new rule.
> It still gives you A LOT of submissions. 480 in a 24 hour period is far more than most people submit in the entire contest time.
>
> It leaves you with enough to work with to do some tweaking, but prevents you from > doing silly tweak bombing.

Agreed. In fact I believe the contest would benefit if it was stricter still. Even a limit as low as a single entry per day need have no real impact on the capacity of parameter tweakers to find a near optimal parameter set. This is simply because there is no limit on the number of entries that you can test against the sample suite. 1, 10, 1e6. Go berserk.

If it is critically important to you that you test your parameter sets against the contest suite rather than the sample suite, then you are clearly not finding the best parameter set, but simply overfitting. Last I remember, we agreed that that is a bad thing.

"Ned Gulley" <gulley@mathworks.com> wrote in message <ibhdun$oi5$1@fred.mathworks.com>...

> Alan Chalker wrote
> > FYI, in case everyone didn't notice it, there is a significant change to the rules...
> > Is there anyway to have this restriction be 'reset' if the queue is completely empty?
>
> This is the kind of rule change where we are very interested in hearing what people prefer.


Please keep this rule!

Nice contest problem.

Nathan

Nice contest problem and good improvements to the contest machinery. Especially seeing whether or not your submissions scored in darkness is great.

Congratulations to the three twilight winners. Great scores and a mile ahead of the rest!

I recieved the following error on submission. Could someone please explain what it means - I can't work it out!

The reported result was "Failed (execution): Error using ==> execfilt at 2 The function GET has been disabled.".

For the record, my code doesn't have any fuction calls for "get" present

Thanks...

"Bradley" <starstrike99@yahoo.com.au> wrote in message <ibjth1$qfs$1@fred.mathworks.com>...
> I recieved the following error on submission. Could someone please explain what it means - I can't work it out!
>
> The reported result was "Failed (execution): Error using ==> execfilt at 2 The function GET has been disabled.".
>
> For the record, my code doesn't have any fuction calls for "get" present
>
> Thanks...

I had the same error in my first submission. Turned out I left a 'hold on' statement somewhere in the file. You must have left in some plotting function that needs to get the current axis/figure.

"Peter van der Walle" <walle@amolf.nl> wrote in message
news:ibk1m5$ms$1@fred.mathworks.com...
> "Bradley" <starstrike99@yahoo.com.au> wrote in message
> <ibjth1$qfs$1@fred.mathworks.com>...
>> I recieved the following error on submission. Could someone please
>> explain what it means - I can't work it out!
>>
>> The reported result was "Failed (execution): Error using ==> execfilt at
>> 2 The function GET has been disabled.". For the record, my code doesn't
>> have any fuction calls for "get" present
>>
>> Thanks...
>
> I had the same error in my first submission. Turned out I left a 'hold on'
> statement somewhere in the file. You must have left in some plotting
> function that needs to get the current axis/figure.

Peter is correct. In this case it looks like it was the "close all" in
Bradley's submission that called the GET function causing the entry to fail.

Major kudos again to the Mathworks folks for coming up with another interesting contest concept! I've been loving it so far, and I think the sailing problem has a lot of potential.

I don't know if people will be interested given that a different piece of code is now in first place, but I am submitting this version of Nautilus with some comments describing what it is trying to do at various points in the code. Hopefully this may be helpful in encouraging people who want to jump in to give it a try.

By the way, congratulations to Bradley for knocking my code off the pedestal! It was very considerate of you to wait until after Twilight was over to submit your entry, which I presume was independently developed even though it was submitted in Daylight. I guess I lucked out a bit on that. I'm curious though -- on the sample testsuite it doesn't do nearly as well. Is this another example of volatility in the real testsuite?

Hi Nic,

Thanks for the congratulations - I mixed up the times and thought I still had a few hours of twilight left. Quite frustrating to write out my first hash at some code, adjust the one free parameter left in clear view of everyone, and find that it hits top spot only to realise that I was an hour late!

Teach me to prioritise my job over MATLAB contests...

Thanks for the posts on the figure error; I had been plotting some images while writing the code and accidentally left that line in.

B

"Nicholas Howe" <NikHow@hotmail.com> wrote in message <ibk36r$aja$1@fred.mathworks.com>...
> By the way, congratulations to Bradley for knocking my code off the pedestal! It was very considerate of you to wait until after Twilight was over to submit your entry, which I presume was independently developed even though it was submitted in Daylight. I guess I lucked out a bit on that. I'm curious though -- on the sample testsuite it doesn't do nearly as well. Is this another example of volatility in the real testsuite?


You are absolutely correct. Bradley’s code shows worst performances when compared with Alfonso’s, Greg’s and yours on test suite. That puzzles me because the difference is significant. Possible consequences – any new ideas need to be tested “alive”, even if they do not work on test suite.

"Sergey Y." <ivssnn@yahoo.com> wrote in message <ibk92c$2q5$1@fred.mathworks.com>...
> "Nicholas Howe" <NikHow@hotmail.com> wrote in message <ibk36r$aja$1@fred.mathworks.com>...
> > By the way, congratulations to Bradley for knocking my code off the pedestal! It was very considerate of you to wait until after Twilight was over to submit your entry, which I presume was independently developed even though it was submitted in Daylight. I guess I lucked out a bit on that. I'm curious though -- on the sample testsuite it doesn't do nearly as well. Is this another example of volatility in the real testsuite?
>
>
> You are absolutely correct. Bradley’s code shows worst performances when compared with Alfonso’s, Greg’s and yours on test suite. That puzzles me because the difference is significant. Possible consequences – any new ideas need to be tested “alive”, even if they do not work on test suite.

I don't think this situation is that unusual. The contest suite usually has a few "gotcha" boards, with weird patterns, that are not present in the test suite. The creation of those is Doug Hull's specialty, according to the Contest Team video.

> I don't think this situation is that unusual. The contest suite usually has a few
> "gotcha" boards, with weird patterns, that are not present in the test suite. The
> creation of those is Doug Hull's specialty, according to the Contest Team video.

I think there may have been some "gotcha" boards in the testsuite as well. I had to change a line of code to pass testboard #119, and looking at that one it seemed to fit such a description!

Bradley

"Bradley Steel" <starstrike99@yahoo.com.au> wrote in message <ibkaj4$bfq$1@fred.mathworks.com>...
>
> > I don't think this situation is that unusual. The contest suite usually has a few
> > "gotcha" boards, with weird patterns, that are not present in the test suite. The
> > creation of those is Doug Hull's specialty, according to the Contest Team video.
>
> I think there may have been some "gotcha" boards in the testsuite as well. I had to change a line of code to pass testboard #119, and looking at that one it seemed to fit such a description!
>
> Bradley

That explains it. The version of your code I am testing does not pass #119. Besides #119 the rest can be attributed to minor variations of map sets.

The code was essentially bugfixed by tradewinds6 or 7...
And within that there's the play-variable "targetScore" (and to a lesser extent maxIter) to optimise between accuracy and speed.

I haven't looked through them all, but a lot of the submissions since then seem to be focusing on speeding up parts of the code, eg ind2sub which is quite slow, and the end optimisation which does nothing if the accuracy is good (ie small value).

B

> That explains it. The version of your code I am testing does not pass #119. Besides #119 the rest can be attributed to minor variations of map sets.

A request for Alan: could you please stop including the "tweaking away" comment in your submissions? Putting that in every entry clutters up the Submission Comments page with multiple copies of the same thing. Thanks!

"Nicholas Howe" <NikHow@hotmail.com> wrote in message <ibkpdh$fg9$1@fred.mathworks.com>...
> A request for Alan: could you please stop including the "tweaking away" comment in your submissions? Putting that in every entry clutters up the Submission Comments page with multiple copies of the same thing. Thanks!

Sure. No problem.

Do we have/will have Early Bird Special, Saturday Leap, Sunday Push ?

Let me clarify my question:
Statistic page shows that we are in the Leap phase. However it was not announced.
What is the deadline if we are indeed in the Leap?
Some new participants may not know want does it mean at all.

"Sergey Y." wrote
> What is the deadline if we are indeed in the Leap?
> Some new participants may not know want does it mean at all.

We will give out a Leap prize for the person who made the greatest single contribution (percent improvement) on Saturday, as measured from 00:00 UTC to 23:59 UTC. Since Saturday midnight UTC is basically right now, this is acknowledging work that has already happened. But we will give you time to ponder the Big Sunday Push. This prize will be given to the person who makes the greatest cumulative contribution (based on the sum of percent improvements) to the contest. The period for this prize is midnight to midnight for the whole of Sunday, UTC. So this means it's starting right now.

Congratulations to Peter van der Walle with very nice code modification.
(I, personally, would save it for “Grand Finale”, however)

It is nice to see some unfamiliar names here.

I am especially grateful for returning us (at least temporarily) to the sanity of single-solver solution

"Sergey Y." <ivssnn@yahoo.com> wrote in message <ibrrji$c60$1@fred.mathworks.com>...
> Congratulations to Peter van der Walle with very nice code modification.
> (I, personally, would save it for “Grand Finale”, however)
>
> It is nice to see some unfamiliar names here.
>
> I am especially grateful for returning us (at least temporarily) to the sanity of single-solver solution

Thanks. I realized that the nickelfel (I think originally Bradley's) solver was very similar to mine. So I just plugged my tweak (slowing down near B) into this algorithm.

It is my first contest, so was too eager to try it. I didn't expect it to get the top spot, as I expected it to be slower then the leader. 'Unfortunately' the improvement in score was so large that it became first. As a result it was snatched by others before I could optimize it myself.

"Peter van der Walle" <walle@amolf.nl> wrote in message <ibrsoi$sf5$1@fred.mathworks.com>...
> "Sergey Y." <ivssnn@yahoo.com> wrote in message <ibrrji$c60$1@fred.mathworks.com>...
> It is my first contest, so was too eager to try it. I didn't expect it to get the top spot, as I expected it to be slower then the leader. 'Unfortunately' the improvement in score was so large that it became first. As a result it was snatched by others before I could optimize it myself.


Here are some “tricks of the trade”:
Trick 1 – if you have good idea, use it strategically. For Saturday Leap you want to create single “optimal” submission by trying it on available set and test set (see trick 2), for Sunday Push you may submit it immediately and then start franticly optimize it on the test set. For Grand Prize you have to create 5 subversion (slightly varying some parameters) and then submit them at the last minute. (And I mean The Last Minute. Even 3 minutes is enough for somebody to pick up your submission and make minor improvement)

Trick 2 – how to test you your submission on test set in “hidden mode”. Check Jan’s Sternenkatze 1 - Sternenkatze 7. He intentionally made code slower to keep it from the top. It was not that “hidden”, because he is a well known player and I checked his submission just in case even if they were not at the top, but it could work for you. (May be not anymore :-) )

I don't know how long it'll take to show up, but I just uploaded a file to the file exchange which should serve as a convenient visualization tool for this contest, if you haven't already created a better one for yourself. It can take multiple solver file handles and test boards and plots for each of them the following: the local winds for the board (as a single scaled arrow per location, rather than the runcontest visualization), points A and B, the sailing path (including the step off the edge of the map if that occurs), arrows indicating direction and magnitude of all instances of thruster usage along the path, and the score and total thruster usage for each solver in the legend. As usual for me, I haven't been able to participate much after twilight, but using this function has at least let me keep up with how improvements have affected the sailing paths and scores. If you're
'just watching,' this will let you take a much more detailed look at the results. I'm not sure what the best way to find it will be. I tagged it with 'contest', so maybe that'll be the best way to find it.

Cheers,
Mike Bindschadler

"Michael Bindschadler" <mikeREMOVETHISbind@gmail.com> wrote in message <ibs32v$6jr$1@fred.mathworks.com>...
> I don't know how long it'll take to show up, but I just uploaded a file to the file exchange which should serve as a convenient visualization tool for this contest, if you haven't already created a better one for yourself.

Thanks, Mike. I just approved your File Exchange submission.

Helen

I submitted commented version of the code as "Cat's comments"

"Sergey Y." <ivssnn@yahoo.com> wrote in message <ibs5rn$g01$1@fred.mathworks.com>...
> I submitted commented version of the code as "Cat's comments"

Thanks for doing this Sergey. I was about to do it as I traditionally do, but I'm glad you beat me to it!

"Sergey Y." <ivssnn@yahoo.com> wrote in message <ibs5rn$g01$1@fred.mathworks.com>...
> I submitted commented version of the code as "Cat's comments"

Thanks Sergey!

Helen

"Helen Chen" <helen.chen@mathworks.com> wrote in message <ibejc2$bf1$1@fred.mathworks.com>...
> This season's contest, Sailing Home, is officially underway! You are now a captain and your assignment is clear. You can find out more about the assignment in the contest rules at http://www.mathworks.com/matlabcentral/contest/contests/30/rules .
>
> Please use this thread for questions or discussion about the contest.
>
> Good luck to everyone!
>
> Helen and the Contest Team!

The Statistics page seems to be failing to update.

 
> > Does the rule: "no function handles" mean we can't use bsxfun?
> >
> > If we can use bsxfun what are the limitations? Would @times
> > be okay but not @(x,y)x.*y
> >
> > Thank you,
> > -Sean
>
> Hi Sean,
> I believe that the restriction on function handles means you can not use bsxfun.

The entry "more deap" by Yi Cao (the current leader as I write this) uses bsxfun.

When Late-Stage Twilight starts?

"Sergey Y." wrote
> When Late-Stage Twilight starts?

There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.

"Ned Gulley" <gulley@mathworks.com> wrote in message <ic0tbs$hk5$1@fred.mathworks.com>...
> "Sergey Y." wrote
> > When Late-Stage Twilight starts?
>
> There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.

Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.

"Helen Chen" <helen.chen@mathworks.com> wrote in message <ibejc2$bf1$1@fred.mathworks.com>...
> This season's contest, Sailing Home, is officially underway! You are now a captain and your assignment is clear. You can find out more about the assignment in the contest rules at http://www.mathworks.com/matlabcentral/contest/contests/30/rules .
>
> Please use this thread for questions or discussion about the contest.
>
> Good luck to everyone!
>
> Helen and the Contest Team!

It's all over except for watching the queue dwindle! Thanks to the Contest Team for the smoothest contest ever. The new contest machinery and rules have really improved the experience for me.

My final tweak, entered into the current leader and a few from the queue, was an improvement to the function "while_next_move". It turns out that the variable "next_move" is redundant, and one can directly access the "stop" variables instead. I also eliminated the interim variables xw,xy in the calculation of thrustCol and thrustRow. I don't think they are "worth their nodes".

I'm curious what other folks were trying to sneak in at the contest closing bell.

Good luck, and happy sailing, everyone!

the cyclist

"Ned Gulley" <gulley@mathworks.com> wrote in message <ic0tbs$hk5$1@fred.mathworks.com>...
> "Sergey Y." wrote
> > When Late-Stage Twilight starts?
>
> There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.

And speaking of the end, the contest queue is now closed from adding new entries. Once these entries have processed, we will announce the Grand Prize Winner!

Stay tuned and a big thanks to everyone for participating! It's been another fun contest!

Helen

> > There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.
>
> Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.

And we wont have to wait till evening for the queue to complete :)

- Amit

"Amitabh Verma" <amtukv@gmail.com> wrote in message <ic12d5$ioq$1@fred.mathworks.com>...
> > > There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.
> >
> > Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.
>
> And we wont have to wait till evening for the queue to complete :)
>
> - Amit

Although maybe we will, as the queue seems to be hanging right now. :-)

> Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.

And the very nervous wait to see whether large amount of never previously uploaded code behave the same way on the server as locally...

For the local testsuite, can anyone improve on a score of 2936 (slow running - would be a borderline timeout on the server...)? Would be interesting to see the best found paths on all the puzzles...

B

"Amitabh Verma" <amtukv@gmail.com> wrote in message <ic12d5$ioq$1@fred.mathworks.com>...
> > > There won't be a late-stage twilight for this contest. Something we will likely do in the future is make the queue dark, rather than have a late twilight. For this contest, we'll be using plain old daylight right up to the end.
> >
> > Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.
>
> And we wont have to wait till evening for the queue to complete :)
>
> - Amit

You mean evening of the next day,right ? :)

"the cyclist" <thecyclist@gmail.com> wrote in message <ic125k$3l8$1@fred.mathworks.com>...
> It's all over except for watching the queue dwindle! Thanks to the Contest Team for the smoothest contest ever. The new contest machinery and rules have really improved the experience for me.
>
> My final tweak, entered into the current leader and a few from the queue, was an improvement to the function "while_next_move". It turns out that the variable "next_move" is redundant, and one can directly access the "stop" variables instead. I also eliminated the interim variables xw,xy in the calculation of thrustCol and thrustRow. I don't think they are "worth their nodes".
>
> I'm curious what other folks were trying to sneak in at the contest closing bell.
>
> Good luck, and happy sailing, everyone!
>
> the cyclist


Thanks to Contest Team for nice problem and new reasonable rules. Now we have to wait 6 months until next sleepless week of “meaningless” code writing.

One of tricks I used was removing ~all(checked) conditions from while loop.


Best wishes to everybody.

Sergey

> > It's all over except for watching the queue dwindle! Thanks to the Contest Team for the smoothest contest ever. The new contest machinery and rules have really improved the experience for me.

Ditto - it's my first contest, but it's a great challenge. I really enjoyed the problem posed, which was fun to try and solve and unlike any coding I've had to do before.

> > I'm curious what other folks were trying to sneak in at the contest closing bell.

Mine is a largely new entry, based on the code I submitted just after the end of daylight. It searches in travelling-salesperson fashion from both A and B simultaneously, which gives significantly better results (~100 lower result for similar run times) on the testsuite than the original one way search that was retained in the mainline code. Unfortunately I forgot about cyclomatic complexity so the final score won't be so hot (+13 unnecessarily).

B

> > > Wow. No late-stage twilight and just 5 submissions per 15 minutes. It does really change the game.
> >
> > And we wont have to wait till evening for the queue to complete :)
> >
> > - Amit
>
> You mean evening of the next day,right ? :)

You need to consider the fact Autotweak is missing :)


> I'm curious what other folks were trying to sneak in at the contest closing bell.

Well I wanted to write a small fnc to check winds around B which was most probably why we needed to run the longer loop and thus skip run on algo 1 for that board to save some time but never got around to do that balancing work/contest. Though I may be wrong on this point.

Best wishes everyone..
- Amit

Once again an interesting contest! Many thanks to the contest team and all participants.

Regards
srach

Looks like it's congratulations to Andre Fioravanti. Well done!

B

Congrats.. Andre !

Well done contest team ! I did have a suggestion for the contest team. It might be a nice addition to include a check in the test-suite for the allowed functions. I know they are mentioned on the rules page but still it would definitely help. I guess I was looking forward to Bradley's Tradewind series... if not for the uncommented clear..

Regards,
- Amitabh

>they are mentioned on the rules page but still it would definitely help. I guess I was >looking forward to Bradley's Tradewind series... if not for the uncommented clear..

Me too! Would love to see how it would have done, but will have to wait till next time.

B

Thanks to the contest team for another fun contest! I unfortunately couldn't compete the past 36 hours since for some silly reason my wife thought it more important to focus on the birth of my son;) But based on the messages it looks like the finale was exciting as usual! See you all in the Spring!

Thanks again for a great contest. Like usual I only play darkness and twilight, but I keep an eye on what's going on. And it seems that the good challenge and the changes in the rules made it a great experience for everyone involved.

Congrats to all the winners.

Is it possible to have multiple nick names for a single email account? Current rule is each email account can only submit 5 entries within 15 minutes and no one allowed to have multiple email accounts. This is fine. However, with only one name, it will disadvantage some well known players because whatever files submitted by them will be quickly checked by others before it actually scored. For a fair play, we should hide players' identity from entry at least in the queue.

Yi

"Yi Cao" <y.cao@cranfield.ac.uk> wrote in message <ic2r66$kg9$1@fred.mathworks.com>...
> Is it possible to have multiple nick names for a single email account? Current rule is each email account can only submit 5 entries within 15 minutes and no one allowed to have multiple email accounts. This is fine. However, with only one name, it will disadvantage some well known players because whatever files submitted by them will be quickly checked by others before it actually scored. For a fair play, we should hide players' identity from entry at least in the queue.
>
> Yi

Late Twilight/Darkness would solve this problem I guess.

-- Andreas

"Alan Chalker" <alancNOSPAM@osc.edu> wrote in message <ic26ai$r3l$1@fred.mathworks.com>...
> Thanks to the contest team for another fun contest! I unfortunately couldn't compete the past 36 hours since for some silly reason my wife thought it more important to focus on the birth of my son;) But based on the messages it looks like the finale was exciting as usual! See you all in the Spring!


Congratulations to your new son, wife, and of course you!

"Yi Cao" <y.cao@cranfield.ac.uk> wrote in message <ic2r66$kg9$1@fred.mathworks.com>...
> Is it possible to have multiple nick names for a single email account? Current rule is each email account can only submit 5 entries within 15 minutes and no one allowed to have multiple email accounts. This is fine. However, with only one name, it will disadvantage some well known players because whatever files submitted by them will be quickly checked by others before it actually scored. For a fair play, we should hide players' identity from entry at least in the queue.
>
> Yi

Wouldn't it be easier if the source code of all entries in the queue remains hidden (even during the whole contest)? Once an entry is scored, the normal darkness/twilight/daylight rules would apply. The you would still see that someone has submitted something, but would not know, what he/she is up to.

By the way, I really like the 5 entries per 15 minutes rule.

Regards,
Stefan

Congratulations to Andre for winning this contest! He spot the flaw in my submission and fixed it within half a minute. In my final submissions, I applied the 'zoom' to the area around A, which gave some time improvement, and applied a few tweaks that were around in the last hour of the contest. Kudos also to those who submitted completely fresh code just before the final time, especially Werner and Bradley. And thanks to the Matlab team for this fun contest!

Sebastian

"Sebastian Ullmann" <sebastian.ullmann@web.de> wrote in message <ic2ukk$2fk$1@fred.mathworks.com>...
> Congratulations to Andre for winning this contest! He spot the flaw in my submission and fixed it within half a minute. In my final submissions, I applied the 'zoom' to the area around A, which gave some time improvement, and applied a few tweaks that were around in the last hour of the contest. Kudos also to those who submitted completely fresh code just before the final time, especially Werner and Bradley. And thanks to the Matlab team for this fun contest!
>

Congratulations to all for a very exciting finale, and to the matlab team for creating again a very challenging problem and a very smooth contest! (and my thumbs up to the new contest machinery and five-entry rule, that worked brilliantly; I am still wishing for the late-twilight and/or hidden-queue rules though...)

Regarding the ancestry of the winning entry I originally selected three interesting entries ('less significant' by Amitabh Verma, 'REP' by Gwendolyn Fischer, and 'continue' by Sebastian Ullmann, which can be traced back to the 'nickelfelpeter' contribution by Peter van der Walle and of course back to Nick Howe and Bradley Steel's own original contributions) that performed particularly well and/or complementary and put them together with my own 'fresh code' version to see what combination might work best. Mike Rusell quickly saw that my 'fresh code' was adding nothing to the mix :) (actually it got to break the best result with 'lasttry05' when combined with 'continue' by Sebastian Ullmann, but it took over 100 seconds in the attempt) and selected the 'REP' and 'continue' combination which survived until the winning entry through multiple improvements by Mike Russell, Magnus, Gwendolyn
Fischer, Tim Vaughan, Cameron Nowzari, David Jones, Sebastian Ullmann and Andre Fioravanti.

In any way, congratulations to all again for a great contest and to Alan for his even-better-than-matlab prize!

Alfonso

Thank you to the Matlab contest team, and congratulations to Andre!

This contest has inspired me to post on the file exchange some 2D vector field visualization functions that I have developed. I thought it was worth mentioning here in case anyone is interested -- these work well in visualizing the wind charts.