Future Contests - your suggestions

[Phantom]

How about tasks requiring advanced algorithms rather than backtracking? IMHO it's quite more interesting to develop a good solution for a difficult problem than implement heuristics...

[stuart]

Phantom - can you give a couple of examples? just so im certain what you mean.

Cheers

[Phantom]

Here is a big problemset: http://acm.uva.es/problemset/. Some of these are really very easy, and some of them really tough. What I would really like would be problems with big numbers of solutions with different time complexity for example: obvious exponential solution, quite tough O(N^2) solution and very tough linear time solution. There are dozens of problems of this kind. What about judging: you can have for example 10 testing sets each one with different size of input...
(It's easy to distinguish between solutions with different complexity. for example for N=1000: O(N) will count about 1000 times faster than O(N^2).)
And for each test accomplished in timelimit (with correct result of course), solution will receive one point. There is problem what to do if some solution received equal number of points. You can order them by for example submission time or total execution time. Maybe not bad idea is to give for example two or more tasks at one contest. It will cause bigger scores spreading.
But of course these tasks could be also judged in the same way you judged previous..., but this solution I mentioned favorizes good solutions over hackers' solutions.

I would be glad if you take my oppinion into consideration... If you are interested, I can give you a piece of advice about it...

[wjensen]

The problem with doing contests with some sort of polynomial time complexity is that chances are there are already very good solutions to the problem readily available on the internet.

I also find it a little humorous that you talk about "advanced algorithms" as opposed to heuristic ("hacker") algorithms. I suppose you think simulated annealing and genetic algorithms are simple?

My point is that sometimes it takes a lot more cleverness to come up with an acceptable solution to an impossible problem than it is to use a cookie cutter solution in a new way.

Now don't get me wrong, I'm not saying that problems with polynomial time solutions are a bad idea for these contests. I think they would be fun also. I just don't think you should lump all heuristic algorithms together into the "hacker" category.

Uhh my poor english level might cause some misunderstandings. I haven't called heuristic algorithms as "hacker" algorithms... I meant that solutions based on poor algorithm, but with lots of tricks signifficantly decreasing constant hidden under the 'O' sign. I had such a funny situation in one of contests, when for extremally simple task requiring just sorting (it was really easy to write linear time solution) one person wrote buble sort, but using trick with register variables in C and he received all possible points ;-), another example for what-I-mean hacker solution was using loop unrolling at code level.
I really appreciate good heuristic solutions, and I think these are very good and they very often require a lot of development effort. I would never call them hackers' solution. I know that many important problems nowadays are in NP-* or EXP classes. So these type solutions are the only available.
Maybe some words of defense for the solutions availibilty on the web: I took part in many algorithmical online contests and usually it is impossible, or very tough to find an apriopriate solution in the internet... It just depends on the problem because you can also find heursistic solutions over internet for many problems...

[Phantom]

Uhh, sorry - I forgot to log in.

[stuart]

Lots of good points, the truth is everyone likes different challenges!! I do intent to vary the contests and 'advanced algorithms' will certainly come up.
Phantom - send me an email if you have any good contest ideas
I wouild also like to say that many solutions to the contests may be found on the internet - I see that as a good thing, it lets people who do a little research see different angles before they write their own. And even if they do copy them it doesn't go to say it the best solution (and its not as much fun either).
Any ideas on contests - let me know.

cheers

[cagrET]

In newest contest, link to judges is missing

http://www.php-editors.com/judges.php

btw.
why noone told me that there is a Winner's Cup image ?



its great ! im gonna put it on my website B)

[cagrET]

yeah
it looks gosu on my site
http://cagret.prv.pl :P

[PHPdev]

I must agree with many, that puzzles get old. Personally I liked seeing the different solutions to the puzzles for a while, but personally I dont see any real world use. I mean can not most puzzles be solved using AI search techniques.

Please do not take this as a complaint, I LOVE the contensts (although I have never entered), and think they give some of us challenges we need.

If you wanted to give a real world problem, and make it fair you could judge the contest based on when the project was submitted, vs how well it works. Meaning if the project was submitted on the frist day of the contest, and it worked flawlessly, then obviously the contest would be over. Too allowing only 1 submission per user, would curb the users who want to keep submitting a project every day until the contest is over. I can see how this too could be dificult to implement...

Thanks

PHPdev