JOIN
Get Time
forums   
Search | Watch Thread  |  My Post History  |  My Watches  |  User Settings
View: Flat  | Threaded  | Tree
Previous Thread  |  Next Thread
Challenge System | Reply
The Challenge System needs some _major_ work. Someone can recieve over 1000 points and not submit any programs. Less points need to be thrown around here.
Re: Challenge System (response to post by jimbeam) | Reply
Actually, why should the challenger recieve any points for breaking your code? I think most people just went around and entered the most random shit they could think of in the hopes it would break your code. The code you break should lose those points, while the breaker receives none.

Oh yeah - and could we have clearer directions while you're at it? This is reminding me of CS 2704.

Just my $0.02,

jimbeam can be picked up at your nearest ABC supercenter.
Re: Challenge System (response to post by jimbeam) | Reply
I think everyone agrees that the challenge system needs to be modified. Here's our latest potential fix: a sucessful challenge is worth HALF the points that the submission was worth, and an unsucessful challenge results in a penalty of a QUARTER of what the submission was worth. That will help keep the cost-benefit relationship in tact.
Re: Challenge System (response to post by ads09) | Reply
I don't think the awarded points should be based on the value of the submission at all. Why should you get _less_ points for invalidating a solution the person worked on _longer_. I the beta rounds I participated in, it was clear that the less time spent on a problem, the easier it was to challenge. I think the points you get should depend on how long it takes you to find an error.

Also, I think everyone should get to challenge any code they want, even if it has already been challenged. If the idea is to reward people for finding errors, then everyone should have a chance to find them, not just the fastest or luckiest clickers. I think we could live without chat or immediate notification of successful challenges.

Re: Challenge System (response to post by Soli) | Reply
soli, while i agree with you should get more points for a challenge that someone worked on longer (perhaps pointval - submissionval?), i don't really agree that it's a good idea to let multiple people challenge one problem, since the conservation of points thing would get really out of whack. i'm not saying points have to be totally conservative, but the goal here was to make submissions be most important, and that might disappear if everyone got to challenge everything. as for points being based on time taken to challenge, we've considered that, too, and what do you do if you have the situation of "public class Foo { }" as the program? pretty easy to challenge, and worth quite a bit?
Re: Challenge System (response to post by ads09) | Reply
If you want the points to be conservative, then why not allow challenges for free, and if successful just remove the score for the bad entry. The incentive is to lower the score of your opponents, thus making it more likely for you to win. Otherwise, with 8 ppl in a round, almost any scoring technique gives the challenge round many many more possible points then the coding round (as evidence, I submit my scores in the first two beta tests, and those were small rounds!) The new scoring makes the 1000 pt problem the only one worth challenging for the most part.
-Soli
Re: Challenge System (response to post by Soli) | Reply
The challenge round does not exist only to test correctness. That is the purpose of the system tester. The challenge round was intended to add an element of strategy in scoring and skill in debugging.
Re: Challenge System (response to post by chuck_tc) | Reply
The problem I have with the current system (even though it is probably 200% better then the previous one) is simple:
round 1) possible 1750 pts for anyone - very fair, imho.
round 2) 2000-4000 pts (assuming 7 other ppl in the room) split amount the fastest and luckiest clickers. In many cases it has been very obvious that a person did not understand the question with a single glance at the code. Whoever clicks view code on these people first gets the free points, at which point everyone else is _shut out_. Scoring in the new system, I estimate, will be something like: 20-40 pts for bad 250 answers, 75-125 pts for 500's, 400-600 pts for 1000's. I'm fine with testing skill in debugging, but I'd rather leave the "scoring strategy" (ie winning by clicking the right place first) out.
Re: Challenge System (response to post by Soli) | Reply
I posted in the strategy forum that we should have fixed rewards for the challenges (something like 100 for challenging a 250 point problem, 150 for the 500 problem, and 200 for the 1000 one). This way you still get a decent reward for a challenge, but it won't be the most significant part of the contest. IMO, the challenge round should be able to separate two coders who are close in score after the coding round, but it should not allow someone to get huge numbers of cheap points.
Re: Challenge System (response to post by Spam_vt) | Reply
100, 150, and 200 are still rather large numbers to be throwing around, imho, especially with 21 programs to challenge in an 8 person room. I expect most scores will be within 200 pts of each other after the coding round.
-Soli
RSS