corridors (preliminary) results

ok, the results are in (thanks to zack!)

1 [DDY / Thierry LE FLOCH] => 670
2 [njq / Ngiam Jiquan] => 659
3 [free_prisoner / Bakr Alsharif] => 589
4 [KoszoJozsef / Kószó József] => 566
5 [ivo / Ivo Kovarik] => 514
6 [pietm / Piet Marcus van der Meulen] => 462
7 [ufosh / Piotr Maksymiuk] => 460
8 [Dios / Stephan Six] => 450
9 [SuperMario / Mario Sansone] => 410
10 [ShiVAs / Audrius Karabanovas] => 406
11 [Hoack / Aleksey Linetskiy] => 394
12 [adamg / Adam Gawarkiewicz] => 381
13 [malcomix / Marco Spizzichino] => 324
14 [bjorn / Björn Paetzel] => 185
15 [nicolasconnault / Nicolas Connault] => 118
16 [ckennelly / Chris Kennelly] => 57


maps:
http://zombie.codewalkers.com/cw/corridors/maps.zip

results:
http://zombie.codewalkers.com/cw/corridors/results.txt
(no text, but you will understand what is what ;))

judge-log:
http://zombie.codewalkers.com/cw/corridors/judge-log.zip
(2M text file, zipped to 200k)


you can (and should) download this stuff and check if all was correct.

if nobody objects, we will make this official in a couple of days.


thanks for being patient. have fun..

what, no comments yet?

i thought you were all eager to see the results..

three questions:

1.you said:
why didnt you do so?

2.all maps look quite alike (there are no maps with more than 20 rooms [i havent looked at all of them]).
why?

3.why havent you judged map 50?

good work zombie and matt.
Please don't make this Contest the last one, cause all the coders love being here with you in it.

Let's see some (ultra-unofficial ) summarized cost and time of building the corridors,
which is not shown in the result arrays (because who cares? ;) )

$32784 247.82s KoszoJozsef
$32874 904.70s ivo
$32910 21.68s DDY
$32960 17.41s njq
$33026 108.77s ufosh
$33142 20.78s free_prisoner
$33170 54.43s pietm
$33234 39.74s SuperMario
$33318 213.96s malcomix
$33370 49.37s Dios
$33390 38.02s ShiVAs
$37208 59.77s bjorn
$43738 84.27s nicolasconnault_CorridorsBase
* $27278 33.78s adamg
* $30662 41.76s Hoack
* $219906 18.06s ckennelly

* = some room not connected


So, think twice, before you want to build corridors, it's not so easy!



Thanks for judge this contest in 2004! HaPpY NeW YeAr + BUÉK!

well i just want to say thanks zombie and matt for everything you did for this contest, no matter how long it took. I personally have never sent in a script, but i always have a LOT of FUN time making a script and just playing around with the contest. this is i think the best (and maybe only, because i dont know of any others) php contest on the web and would be really disapointed if they stopped. So PLEASE matt, dont make them go away, i need something to kill the time sometimes and i find that a codewalkers contest usually hits the spot.

at least things are moving now =p

i can see alot of work put into creating the maps, but i have some objections towards them .. esp with regards to the number of rooms (as also highlighted by ivo) .. basically there are no maps which have > than 18 rooms ! (while we all thought there would be up to 50)

just for info ... here's a list of the maps and number of rooms they contain ... generated by quick script i had put together

i did do exactly that. in fact, that is one of the reasons why it took me so long to judge (at first, i couldn't find a way to do that).


this is how it was all done: zack made some maps with his generator. then, he sent them to me, and i modified them (3/4 of them) to cover the special cases.

the modification was to delete some cells (that where maybe part of rooms) and add 3-4 rooms in special position so that it could be solved better than simple MST.


1) first quoter (around 16 maps -- it was random) i didn't modify.


2) second quoter, i cleared one corner of the map, and added 3 rooms in such way that it could be connected with advanced algorithm.

for example, look at the map13.txt, lower left corner. and look at bjorn's solution as to how that should be connected with lowest cost (just that corner).


3) for the third quoter, i cleared some cells in the middle, and added 3 rooms to be connected.

for example, check out the map18.txt (center of the map), and check out DDY's solution of how that should be solved with lowest cost.


4) for the last quoter, again i cleared the middle of the map, but this time i added 4 maps to be connected.

you can see the example in map7.txt, but the only one who did solve this in the best possible way was Hoack (unfortunately, he didn't do so well with some other maps).




18 is less than 50, last time i checked..

but really, i don't understand what you are complaining about? did i trick anyone? did i make any false statement?

i don't think so..

nah you didnt trick anyone or make false statements .. just that in most programming competitions .. limits play a very important role in determining the best algorithm to deploy

and almost for all (except this one of course), at the very least, the limit test cases are part of the test data

btw, an easy way to make sure that the maps are always better than the standard mst, is to generate maps and just run them against a standard mst program and the rest of the submitted entries... and reject the map if the best result from those submitted = the mst result ... doesnt take very long to code the mst soln (or simply request for it, i'm sure quite a few of the participants here have already written one in the process)

imagine a script, that always finds the best solution in 30 seconds.
it would finish 3rd.
Array
(
[DDY] => 700
[njq] => 672
[best] => 645
[KoszoJozsef] => 599
[free_prisoner] => 599
[ivo] => 543
[pietm] => 472
[ufosh] => 472
[Dios] => 460
[SuperMario] => 418
[ShiVAs] => 413
[Hoack] => 405
[adamg] => 388
[malcomix] => 331
[bjorn] => 186
[nicolasconnault_CorridorsBase] => 118
[ckennelly] => 57
)

i think this is mainly because of the unchanged maps.

if you would take out the 10 maps on wich 'Dios' (whos script is standard MST as far as i can see) gets the best score, this would result in:
Array
(
[best] => 583
[DDY] => 559
[KoszoJozsef] => 520
[njq] => 510
[ivo] => 479
[free_prisoner] => 441
[ufosh] => 378
[pietm] => 351
[Hoack] => 349
[Dios] => 332
[SuperMario] => 323
[ShiVAs] => 303
[adamg] => 282
[malcomix] => 254
[bjorn] => 145
[nicolasconnault_CorridorsBase] => 96
[ckennelly] => 47
)

dont get me wrong:
i dont want to attack you in any way.
you are the judge, you have to decide how to judge.
but i think its ok to mention things like the above.
and once again: thanks for the contest and the effort.


btw on maps with more than 25 rooms the script of njq is (in my tests) hard to beat (while mine and jozsefs perform rather poor).

last post was mine

Yup, standard MST! ;)
Didn't had the time to come up with something better! ;(

Happy new year to all of you!

Hi, and Happy new year to all !

I really hope this contest was not the last.

I made a table to "see" the results:
http://cigosoft.org/pub/results.php

ivo: i don't know what to say. what if there was one script that solved every map with the best possible solution in 29 seconds, and another one that solved 50% of the maps with best solution, and the other half with the second best solution, but all that in just 1 second.

who should win then? (i don't know)

so, i think that this maps are fair for everyone. if you don't think so, make a suggestion of what we should do.

one solution might be to only count those 3/4 maps that i modified. if most of the contestants thik this is a good idea, i will agree..

there is no real 'fair' solution to this.
i personally dislike maps like #1, where almost every script gets the best result. this makes time overimportant.

so i would consider it fair to leave out the unchanged maps.

anyway, the difference is small, so the discussion is more or less academic.

its up to you, i wont complain ;-)