Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
×
Education

ACM Programming Contest Results Revised 68

Goonie writes: "After the controversy over the results of the ACM programming competition, it appears that the results are being revised. Check out the new standings. The winner hasn't changed, but some teams have moved up the rankings." Actually, from the look of the page at 4:15GMT this morning, "final rankings are under review." Let's hope the fairest decisions prevail, and that all involved are gracious.
This discussion has been archived. No new comments can be posted.

ACM Programming Contest Results Revised

Comments Filter:
  • by pb ( 1020 )
    What happened to them sounded pretty lame, I hate it when people can't just fess up and admit they made a mistake.

    The rules sounded somewhat weird, but it's probably no weirder than the programming contest [msoworld.com] I've been doing lately (its monthly, so hold your horses). (and MazeMan can admit he makes mistakes. :)
    ---
    pb Reply or e-mail; don't vaguely moderate [152.7.41.11].
  • I was very disappointed in this years contest. The automatic turnin program written in Java was not functioning at the Scranton site.

    All our submissions had to be hand submitted by the judges which greatly reduced the speed of which we could get back feedback (you could have submitted a program an hour into the contest and chances are you wouldn't have known if it worked or not).
  • by zpengo ( 99887 ) on Monday April 17, 2000 @06:38PM (#1126873) Homepage
    I looked at problem F, the controversial one.
    I don't see how anyone could have made enough sense of it for there to be a controversy.

    Damn good programmers, them.

  • Does anyone know which teams used what languages? I wonder if anyone hacked something together in perl....
  • how can one rate a programmer?

    i program in many a language
    i program so i can use my computer
    in different.ways

    programming is an art



  • I wonder if anyone hacked something together in perl....

    C, C++, Pascal, and Java are the only acceptable languages.
  • This article references another Slashdot article that was posted almost a month ago. Could someone remind me what the issue is, and why this proved to be controversial? I see something about a question F. Why was the question flawed?

    Thanks!
  • when this story first came out i was so happy for those guys. i went to school with both the Johns, and used to code a bit with them.
    Ive taken the year off uni (so im not eligable to join in the acm), and working as a coder. My last job was to write an LDAP enabled pop-server, over the weekend and during that time my girlfried (working as a waitress) earned more than me. god the IT industry is in a crap state. or maybe i just struck out with a bad job.

    heh. i dont know why im posting here. but on what turns out to be like $8 per hour, i cant afford professional help.

  • I agree, things at Lehigh weren't much better.

    I would like to add, however, that this contest is important to motivate people to take an interest in teaching group software development under a time constraint. What I have learned from being an ACM coder for the past few years greatly helped me on the job working with other coders, and people who had an idea but just didn't know how to code it up.

    Sharing a single computer was better influence over good programing practices (for speedier debuging if not having the occasional team-mate reading over your shoulder) than any software development course I've had.
  • Let's hope the fairest decisions prevail, and that all involved are gracious.

    No! I hope that the new results are terrible, and that everyone is pissed of. and that the resulting chaos destroys the ACM and everything it stands for!
  • simple (lower is better)

    - does it get the job done

    - does it get the job done every time

    - does it get the job done every time, and quickly

    - does it get the job done every time, quickly, and doesnt look like something larry wall would write

    so yeah, thats one way. they might have more stringent rules then me, but hey
  • I hate it when people can't admit they lost. Win and lose honorably.
  • I must have missed the controversy, or perhaps I'm looking at the wrong problem set, (gee, the pdf says 2000 ACM Programming Contest World Finals), but this doesn't seem to be at all vague. It is an application of graph theory. The classic question is 'what is the shortest route from node A to node B.' This just asks you to average all the answers to all the possible pairs of A and B.

    Now, this does not mean that it is easy to do, or that I remember the algorithm I once learned for finding the shortest path from a to b on a graph, but I don't see why anyone with a bit of computer science knowledge wouldn't be able to understand the problem.

    Of course, as I said, I missed the controversy, so that may not have been what the hubub was about.
    -Matt
  • I highly recomend storing the adjaceny list in a hash table for easy lookup, but past that there is no known fast (ie in polynomial time) alg. to find a shortest path - if someone out there knows of one: email me and we will go on tour!

    What is possibly the best (yes, yes, there is always that worst case where the path goes through all n nodes) solution uses a breadth first search using a queue with parent points to find your path back to your starting node.

    A depth first search is often not the best.
  • I was on the U of Washington team.
    I had prob F coded in 30min or so and got it wrong. We spent the rest of the time trying to figure out why!

    This *might* have us solving one more problem, but it prolly won't change the relative standings.
    Even if we would have gotten that one solved in 30min, I don't think we would have solved another one.

    Isn't this the first time the standing have ever been changed???
  • I don't recall... was there the ANSII standard imposed on that?

    (I seem to remember that ANSII does not allow for embeded Assembly code, which would have made for some slick bit shifting and fast output)
  • The teams recently received letters from the contest directors, who say that they have finally decided to "personally re-examine all submissions of Problem F from the archives" and that "the results of the re-examination will be posted in the standings by April 24th."

    So I'm not expecting to see any changes until April 24th, which is next Monday.

    If they make any changes, teams will be bumped up to a higher position, but no team will have its position lowered. Best of all, they have changed their official policy so that in the future, teams will have access to their programs after the contest is over, and also there will be a standard procedure for regrading after the contest is over, in case this happens again.

    I'm particularly happy since there's a reasonable chance that my team's score will improve, but I think everyone should be glad that they're making an effort to keep this a fair contest. I highly recommend this contest to all interested college students. If you haven't already, check out the problem set [baylor.edu] from this year.

  • by ChadN ( 21033 ) on Monday April 17, 2000 @07:13PM (#1126888)
    Check out this link about the Controversy over Problem F [slashdot.org], and look for Dominic_Mazzoni's post (+5 interesting). I'll quote from that article:

    Take a look at problem F. (Here are the contest problems [baylor.edu] in PDF if you're interested.) In a nutshell, you're given a complete directed graph, and you need to return the average length of all shortest paths between all pairs of nodes. The problem explicitly stated that you will only be given graphs in which there exists a path from every node to every other.

    This is not a hard problem to work out, but anyone who has had a formal course in computer science ought to recognize that the Floyd-Warshall all-pairs shortest path algorithm is designed to solve exactly this problem. Then all you have to do is add up all of the elements of the matrix and divide by n * (n-1).

    Except that the judges made a mistake, and tested our input using a graph that was not connected - in other words, there were nodes that could not reach other nodes via a directed path. This would not be a big deal, except that the problem explicitly stated that this would not occur. (Input validation is never a part of this contest.) Furthermore, without further explanation it is unclear how these nonexistent paths should affect the average. It turns out that the judges' solution was not counting these paths, and averaging only the paths that existed. Some teams did this by accident, and others (including Waterloo) figured it out only after submitting multiple runs and incurring large penalties. My team was one of the many that did not figure out the judges' mistake, so we did not get credit for the problem, even though our solution was certainly correct as the problem was worded. If we had received credit we would have had four problems correct, possibly putting us in the top ten. Of course, if we had received credit right away, we might not have wasted so much time figuring out what was wrong with our solution and we could have solved another problem in that time. Of course, many other teams were in a similar situation, so I have no idea what the final ranking would have been, but clearly it would have been different.


  • Points are awared for creating a working solution (since that's the point right?) and if there is need of seperating a tie then efficiency* is brought into play.

    * Sometimes they say that if your program runs in say >5 minutes it is considered non-functional, so a very naive approach that is technically valid is often not accepted.
  • by Anonymous Coward
    Um, there are several known algorithms for finding the shortest path between either one and any number of other nodes or between all nodes. I don't know what this problem says, but I have a hard time believing none of quite polynomial solutions would apply.

    Djikstra's - O(V lg V + E)
    Bellman-Ford - O(VE)

    IIRC, the problem with the problem (hehe) was that they said the graphs would be connected, and one of the test examples was not.
  • You would think that the hosts of the world's foremost programming competition would take the time to write its own database software with a CGI interface, or at least have the sense to use a more stable, open source package. But NO!! Here's what you get by going to the ACM finals page [baylor.edu] and clicking "Teams":


    Microsoft OLE DB Provider for ODBC Drivers error '80004005'

    [Microsoft][ODBC Driver Manager] Data source name not found and no default driver specified

    /past/icpc2000/finals/RosterPublicFull.asp, line 11

  • by Anonymous Coward
    So, the administration may have botched a little this year, but it was still a very interesting experience, being in the company of 240+ other intense geeks for four days. Plus you can't really beat a free trip to Florida. To repeat what some other guys have said, you should all go participate in programming contests; they're a lot of fun, and you can often win cool stuff! Oh, and props to IBM for the nicest promotional pens I have ever received. :)
  • by Dominic_Mazzoni ( 125164 ) on Monday April 17, 2000 @07:33PM (#1126893) Homepage

    Heh. You're thinking of the longest path problem, which is NP-complete. Finding the shortest path between a pair of nodes is easy - use a simple breadth-first search for an unweighted graph, or Dijkstra's algortihm for a weighted one.

    To find the shortest path between all pairs of nodes, you can use the Floyd-Warsha ll algorithm [brunel.ac.uk], which runs in O(n^3) time. It can be done faster, but in this particular case we were guaranteed n is no more than 100, so this was quite reasonable.

  • Yep. It was aweosme.
    Can't beat that IBM desktoy either:)
  • Okay, I know this is much lower scale than the championchip competition, but still. At our regional ACM contest, we had an painfully easy problem involving the angle between the minute and hour hand on an analog clock. The only difficult thing about this problem was the formating. In the ACM contest, it is very explicit that file output is the only thing judged, screen output is not to be even looked at. So when I jumped down at the keyboard to code this, I dumped some very wrong formating to the screen, but the file was fine. The solution got rejected. I commented out the code that printed to the screen, resubmitted, and bingo, answer accepted. Consider me (and my team) extreamly pissed. I was so tempted to throw the judges a really nasty communications form... Because of this screwup we were penalized and we would have gotten 8th place instead of 9th.

    I know, minor little rant, but it just goes to show that the ACM *never* is wrong. This still ticks me off.
  • by Anonymous Coward on Monday April 17, 2000 @07:55PM (#1126896)
    Does anyone know which teams used what languages?

    The team from MIT spoke only in 13th century latin. The team from Oklahoma State communicated solely in cattle brands.

    The team from NYU in Ebonics. The team from Dakota State University spoke only in a very slow drawl.

    The team from Georgia Tech kept referring to functions as crackers, and arrays as yankees. The team from UCLA constantly threw in random "likes" like, like this, like.

    The team from University of Winnipeg felt it necessary to say aye after each sentence. The team from University of Southern Florida wouldn't shut up about Elian, and often develoved into a Spanglish.

    The team from Brigham Young University kept referring to the moral crumbling of the ACM contest. The team from Texas Christian University always commented their code in the form: //John|3:16
  • by Anonymous Coward on Monday April 17, 2000 @07:57PM (#1126897)
    I was one of the University of Melbourne team (thats Melbourne, Australia).

    If it hasn't already been explained, problem F went as follows: find the average shortest path length between all pairs of nodes in a strongly connected directed graph. The judges appear to have put a non-strongly connected graph in the testing - one for which not all pairs of nodes have a connecting path. It was then pure luck whether you had chanced upon the same method for computing the average as the judges, so while some teams solved the problem in the first few minutes of the contest (notably St. Petersburg state U. ;), others struggled with it for the entire five hours, trying to find bugs in their correct code.

    Because it was by far the easiest of the problems, many teams began with this question. By the end of the five hours, many perfectly clever teams had not solved anything, due to the stuff-ups on the part of the judges.

    Now, making errors is one thing, and it certainly screwed up the contest for everyone, but in some sense errors are forgivable - what was really appalling was the way in which the situation was handled.

    A protest regarding the question was submitted (by my team) within a minute or so of the end of the contest - there was no official channel for appeals, so we were just told to write it down on a piece of paper and give it one of the judges. We never received a response of any sort.

    After the contest most of the teams were aware of the problem - one team had submitted a program designed to go into an infinite loop if the input graph was not strongly connected, thus working out what error the judges had made. In any case, we took our printouts to verify our solution, and later presented our arguments to the tournament director, Bill Poucher.

    I have never met someone so unprofessional, with no concept of accountability or responsibility, and little even of courtesy. I am not sure what his background is, but if Mr. Poucher were the director any public company, he would have been on the street - or in court - in seconds. He refused to listen to any of the requests from the teams or coaches, telling them instead their memories of the contest perhaps weren't clear. Nothing was done and the prizes were awarded, with Bill Poucher announcing at the ceremony that "every problem was solved by at least one team" - so that there weren't any problems.

    After the contest, in the face of protests from regional directors and coaches from around the world, with proofs that their solutions had been correct, the contest organisers continued to deny there had been any problem. It took quite a while for them to agree to remark the solutions.

    Now, I have no especial gripe at the judges that an error was made, because people make mistakes - it's just an indication that they need better validation on the testing programs. I'm also not especially fussed that my team didn't come first - because in some sense, we were one of the LEAST affected teams. Just look at the results of the American universities. I mean, it's nice to have a laugh at America getting its ass kicked once in a while ;), but in this case (having spoken to several of the teams) I think most of those teams just came up against problem F and got stuck.

    What I do have a problem with is the contempt the organisation showed to the tens of thousands of man-hours of work that go into the contest on the part of the contestants. No channel for appeals, no official responses to protests, no means to elect new directors, no accountability, not even a report saying "we made a mistake; we plan to do this to fix it". There was a similar problem in our regional contest this year, and the regional director (Raewyn Boersen) fessed up, and sent around an email describing the problems, the changes she would put in place to the validation of the competition, and the transparency with which future contests would be conducted.

    My question for slashdot is, how do you get an organisation like this to to act like it's accountable to the people who enter?

    In any case, I'd still recommend entering the contest to any CS students - its an amazing experience, however it goes. I've never learnt as much CS in such a short time as in the week's training leading up to the finals. Ask yourself - if you had to, how fast could you code up a 3D voronoi tesselation? Or a fast constraint solver? Can you find a bug in someone else's uncommented code under time pressure, or look at a problem and say how long it will take to code, and what the best order algoritm is? These are the kind of things you learn.

    cheers all,

    John FitzGerald, University of Melbourne,
    Australia
  • I suppose, but you rate code based on efficiency? Different people learn different styles... And someone has the gaul to give a rating? It is a challenge, but for what merit?

  • Yeah, i suppose you are right. But does it really matter what the results are going to be now? Even if the champion is dethroned, the guys who solved those algorithms are probably guaranteed a job at IBM whenever they want it. As for the officials, let them extend their 15 minutes of fame. After all, because of the controversy, they got another Slashdot mention. 8)


  • I believe so... and from my experiences in contest programming, there is no room for that anyway. It is more of a contest of how fast you can code a correct solution, not so much how fast it is.
  • by Anonymous Coward
    as far as i understand, the difficulty arose because shortest paths between nodes which have _no_ path between them are not really well defined - should it be infinity, or not included or what? the question mentions strongly connected graphs, which imply that all nodes have a shortest path to each other node, but a data set used in the marking included some nodes which were not connected to some others. this is pretty clearly a problem.
  • > After the contest most of the teams were aware
    > of the problem - one team had submitted a
    > program designed to go into an infinite loop
    > if the input graph was not strongly connected,
    > thus working out what error the judges had
    > made. In any case, we took our printouts to

    But how would they know that it *had* gone into an infinite loop? Isn't that the essence of the halting problem? ;-)

    -Andy
  • Click on Teams in the Finals side bar

    Microsoft OLE DB Provider for ODBC Drivers error '80004005'

    [Microsoft][ODBC Driver Manager] Data source name not found and no default driver specified

    /past/icpc2000/finals/RosterPublicFull.asp, line 11
  • by zCyl ( 14362 )
    Why are these results posted in a different form on a different less-professional looking site? What happened to the original? Does anyone still have the old url on hand, and does it show these revised results as well?
  • It doesn't exactly keep running forever ;-) you get a 'runtime limit exceeded' error if it takes too long.. this is usually an indication that your simple linear solution perhaps wasn't the best option.

    Another error usable for signalling back data could be runtime error.. like division by zero, etc. I imagine this could be a bit faster :-)

    Regards,

    Talence
  • I suppose, but you rate code based on efficiency?

    Running time to produce a correct answer.

  • yeah yeah...

    I'm sure hosts of a world famous programming competition would want to waste time and resources to write its own database software and reinvent. Maybe it's a foreign concept to some of you guys that sometimes the most practical, most time-saving solutions is the best one, even if it's not "the right thing".

  • The team from Winnipeg would not have been saying 'aye'. They would have been saying 'eh?'.
    Get it right eh?
  • As someone who works with computer systems that have been coded in the near to remote past by other people, I can assure you that if your program displays some values, someone somewhere perhaps in the remote future will look at them and try to use them. Even if your specification gives the file output as the designated deliverable, in the real world outside of academia, I would consider any incorrectly formatted output a problem with the code.

    Anomalous: inconsistent with or deviating from what is usual, normal, or expected
  • I don't know Mr. Poucher, but he sounds like the stereotypical American career academic. I've had the misfortune of studying under a few.

    They rise to a certain level of recognition, probably higher then they feel (deep down) they deserve. They resolve this insecurity about their competency by believing that they are actually the 'right hand of God', and look down on the unwashed masses with utter contempt, from their Ivory Tower.

    'How DARE this student (spoken through clenched teeth) question MY process and MY organization of MY contest!? Without ME he is nothing.'

    Typical case of recto-cranial inversion, easily cured with a clue-stick beating and a few months at a REAL JOB.
  • But how would they know that it *had* gone into an infinite loop? Isn't that the essence of the halting problem? ;-)

    Simple. They knew because the first step of the loop took 1 second, and each subsequent step n+1 took half the time of n -- so in 2 seconds the program had completed and infinite loop.

  • I was a member of the UCLA team that won the contest in 1989.

    The following year we failed to complete a problem at the Regionals because the mechanics for submitting solutions (on floppy disk) were broken. We discovered the problem and corrected it at the last moment, submitting our solution just as the final bell rang. The judge refused to accept the solution, much to our amazement and that of the nearby teams.

    When we protested, he coldly told us "Forget it. You guys don't deserve a break." None of the other judges would overrule him, since they hadn't been present at the submission desk.

    Turned out he was the coach for one of the other teams, and had a grudge against us because we were graduate students. (Indeed, the competition was limited to undergraduates the next year.)

    That incident kept us out of the Finals. I only wish I remembered his name, so I could besmirch him here on Slashdot :-).

    -- Dr. Pain
  • Absolutly, however... We were explicitly told that this would not be a problem.
  • by adaking ( 158188 ) on Tuesday April 18, 2000 @04:05AM (#1126914)
    I was at the contest. It was the first year for anyone from my school to make it, so there was quite a bit of pressure to do well. Especially since we were one of those schools that everyone kept asking us where it was. I worked on Problem F, and I quickly realized it was very similar to a problem that we had practiced. So I had the solution coded up and submitted in about 30 minutes. After getting a wrong submission, I kept revising it and submitting it. Never got a correct solution. I kept telling myself that there must be some weird case that I'm missing, and if I could find it and show that my program broke on that case, then I could figure out why and fix it. Unfortunately, I never found that case. I am positive the algorithm was correct - I copied it almost exactly from the Cormen book (which I saw many copies of there).

    In the last 5 minutes of the contest, we printed out all of our code to all of the problems that we worked on, so we could have something to go from. We gave the code to our coach, who coded it up and likewise never found a test case that broke it. So he wrote an email message to Bill Poucher.
    A few days later, we got a response. Here's a copy-paste of what we received :
    > > I will issue a general statement about this. There has been a lot of

    > > speculation based on some incorrect assumptions about the data set and
    > > how F
    > > was judged. As an exception, I am going to share some information about
    > > the
    > > problems, but I still am working on getting confirmation from all
    > > parties.
    > >
    > > Basically, all of the test cases were strongly connected. This has been
    > >
    > > verified by at least three very reliable sources. The case that caused
    > > the
    > > vast majority of rejection was a directed circuit. To build your own
    > > test
    > > data, make four or five test cases followed by:
    > >
    > > 1 2
    > > 2 3
    > > 3 4
    > > ...
    > > 99 100
    > > 100 1
    > > 0 0
    > >
    > > Be sure to terminate your list of test cases with another 0 0. It is
    > > easy
    > > to mishandle a sentinel-terminated loop of sentinel-terminated loops.
    > > Problems that would cause rejection: not explicitly using long
    > > integers,
    > > improper type casting, coercion problems, loop bound mistakes, incorrect
    > >
    > > initialization of the shortest path matrix, improper traversal of the
    > > shortest path matrix, improper memory allocation, exceeding stack
    > > limits,
    > > failing to reinitialize between test cases, faulty parameter
    > > passage, getting the algorithm wrong, and so forth.
    > >
    > > Be sure that you test it with the command-line compiled version of your
    > > submissions using the exact command-line compiler flags used in the
    > > World
    > > Finals. Make sure to precede this test case by five or six others to
    > > test
    > > re-initialization and proper memory management.
    > >
    > > I don't know whether Delphi suffers from the same malady that Borland
    > > C++
    > > suffers from, but Borland C++ defaults to 16-bit integers at the
    > > command-line, but 32-bits at the IDE level.


    The reason that 16-bit integers would break any potential solution is that for the case given where we have a 100-node loop is that the sum of distances could easily go beyond the 32767 mark. I guess the teams that did get it either found another algorithm to use or they explicitly used long integers. Pardon me, but ever since 32-bit operating systems came around, I have pretty much assumed that int == 32-bits.

    Since then, I have yet to hear of a "general statement", except for this Slashdot story.

    In my opinion, many of the results of the competition are invalid because of this problem. Had a correct response come back to us in the first few submissions, we would have gotten at least one, if not two more, correct. We wouldn't have changed the top-most standings, but it's possible that there are other teams that would have.
  • I gave up on programming contests in grad school, after I discovered that ill-defined questions and untested submission software were the norm. Maybe if I were the sort of person who likes hacking out -- er, reverse engineering -- random code I would have been happier, but IM absolutely correct O real computer science and/or software engineering is about something other than the ability to guess what errors the judges might have made.
  • by Anonymous Coward
    ...ever since 32-bit operating systems came around, I have pretty much assumed that int == 32-bits.

    That is a foolish assumption to make. What happens when you work on a small embedded 8- or 16-bit processor? What happens when you work on a 64-bit system?

    The C language spec makes no guarantee for the exact range of an int. It is only specified relative to the size of short and long. Sorry, but if you're going to be pedantic, you should be pedantic eveywhere and not assume type sizes that aren't specified exactly.

  • I participated in the ACM contests many many years ago. I was somewhat unhappy with the way the international level was run. One year, AT&T sponsored the contest, so they forced us to use SysV Unix and vi. Well, gosh, I had been raised on (gasp) PC editors, where I could use *cursor keys* to move around, and I was plonked into vi. I learned jklh as well as 'i' for insert and 'x' for delete one character. That slowed me down tremendously. Oh well. That wasn't my major complaint though. The main problem was that the contests are in an artificial setting -- your judges are sitting high above, handing down yes/no verdicts, without explaining WHY. Okay, fine. That's just how the contest is run. But what bugs me is that at the ceremony, when they're describing the contest to representatives from industry and government, they say that the contest is very much like the REAL WORLD! HELLO! In the real world your customers can tell you what they don't like. They don't come back and say, "no". It's a completely different environment when you can communicate with your customer -- to get clarification of requirements, to find out what they didn't like, to modify the requirements, and so on. In the ACM contest (back then, at least), you just can't communicate with the judges. You're totally isolated. Then they mislead the press into believing that doing well in these situations (learning unfamiliar tools in an hour, not communicating with customers, using no libraries) are somehow correlated with success in the real world. Bah.

    I had a lot more fun at the IEEE contests, run by none other than Mr. Transmeta, Dave Taylor.

    - Amit
  • I competed once in an ACM programming contest (the regional one this year), and have to agree that it's a sketchy affair run by people who really don't know what they're doing.

    For example, it was fairly obvious that the problem writers had no real grasp of Java, even though its use was allowed. One problem in the contest involved taking the reciprocal of arbitrarily long integers that are a power of 2 or 5. My group was the only one to finish this problem: all the other groups thought it was one of the most difficult. Considering that Java has its java.lang.BigInteger class, solving the problem was trivial, and a no-brainer.

  • The original posted rankings can be found at http://plg.uwaterloo.ca/~gvcormac/acmresults/final 2000.html This page also has a pointer to the revised results that were posted April 16, but the "rankings" section was deleted later the same day, replaced by "Final Rankings are under review" The trouble with re-judging is that it is impossible to undo the effect of the bad judging. Some teams were not affected by the problem, because they chose a method that happened by chance to conform to the judges' expectations. Some spent a great deal of time and did not have an accepted submission. Some spent a great deal of time diagnosing the data and deducing an approach that would be accepted. Some moved on very quickly when their first submission was rejected - of these many actually had it correct but many had trivial errors that they *could* have fixed had there been correct judgements (it was easy to infer that something was screwy by the number of teams failing to receive credit for this easy question). I am glad that the organizers are prepared to be more open about the process. I have little complaint about the issuance of more medals, but a reranking in which people can only move up makes no sense to me. And I can't think of any teams that deserve to be moved down. All sports and competitions admit the possibility of "bad calls." In nearly all such competitions the referee's decision stands and the results are not changed post hoc (except for disqualification or penalties due to the fault of the competitors). Everything should be done to make judging as reliable as possible, but bad calls are part of the game, and the standing is determined, at least in part, by how competitors respond to the situation.
  • The controversy was caused because the judges had incorrect test data. Although the question stated that the graph was connected, one of the test sets tested an unconnected graph. This wouldn't have been too bad, except for the fact that teams were forced to realize on their own what was going wrong, and guess at what kind of output the judges wanted for such a graph.
  • Maybe it's a foreign concept to some of you guys that sometimes the most practical, most time-saving solutions is the best one, even if it's not "the right thing".
    Tell us, how does something that costs a lot of money and doesn't work qualify as the most practical, most time-saving solution when compared with free, open, and stable solutions?
  • by Anonymous Coward
    Hello -

    I was at the contest. Keep in mind that B.Poucher had a lot on his mind. You were upset that the problem had been misjudged and you had to go and wait for the awards banquet.

    Poucher had to ensure that all of the contest floor, including large areas that you did not see, were completely cleared in one hour. He had to make sure that the dinner, the awards, the party, the transportation, and etc. were ready to go.

    He was under a lot of stress. Remember that he was responsible for every contest-related thing that you did in Orlando. While you were relaxing in your room, he was planning, setting up, tearing down, etc. While you were writing code, he was writing checks. The time put into the ICPC by the contestants is pretty trivial compared to the time put in by the ACM people and the IBM/host volunteers.

    Keep in mind that Poucher did make sure that the problem was re-examined. He just couldn't do it immediately after the contest. It would be a mistake to suggest that only the contestants and coaches thought that Problem F should be re-judged. Remember that all of those mean, evil judges are volunteers. Getting a large group of volunteers to agree to do something as yucky as rejudging a bunch of submissions by hand is like herding cats.

    Remember to walk a mile in the other guy's shoes.

    Peter Doege

    (Hey I'm not anonymous!)

  • Go Golden Knights.

    Booyah :)
  • Thank you!!!!

    As a student at UT Austin in the early 90's, I participated as staff on SPC93 and NPC94. By that time, Dave was at Id, but he would still show up for the days of the actual contest. There's no comparison between ACM and IEEE as far as excitement and competition go. The programs at IEEE are judged based on their performance against the other programs. No judges with human emotions of pride or prejudice come into the picture. Teams could actually help each other out if need be. There was open communication of all problems between the staff and contestants. It was a two-day contest, so teams had plenty of time to code a rather massive work. I never heard from a single team, winner or loser, who was disappointed with the outcome of the contest, the organization, or the communication channel with the staff. I'm just sorry to say that after NPC94, there was not enough student interest or outside funding to keep the project going. NPC is greatly missed.

    The words "Ivory Tower" have been used in other posts, and I can't think of any that better describe the ACM contest. My experience has been that in the realm of Computer Science, there are those who care only about the abstract, theoretical, and don't care so much about whether it works as HOW it works. Those are computer scientists. Then you have the people who have the attitude of "Get the job done... THEN get it done better." Those are engineers. *sigh* I recall being amazed to hear in '95 that Dr. Djykstra owned no PC of his own, and in fact had not actually USED one in two to three years. I've always been very hands-on, and I feel if you don't continue to use something, you forget it. If as a computer scientist, you don't USE a computer, IMO, you're really just a specialized mathematician.

    I'm not saying the ACM contest isn't worthwhile. In fact, it's great fun. I definitely wouldn't hold it up as a model of 'real-world' software development, well, maybe software development at Microsoft ;)
  • I went to the ACM site to check out the full list of teams (looking for my alma mater, Technion) and
    found that they need to revise their code as well (see below for the error message).

    a) Maybe _they_ should hire a medalist...
    b) Obviously, they should have used Linux.

    I got:

    Microsoft OLE DB Provider for ODBC Drivers error '80004005'

    [Microsoft][ODBC Driver Manager] Data source name not found and no default driver specified

    /past/icpc2000/Finals/RosterPublicFull.asp, line 11

    It's at http://acm.baylor.edu/past/icpc2000/Finals/
    and then click on teams.
  • That is a foolish assumption to make.

    No it isn't. If you're programming for portability - maybe. Knowing your target platform is absolutely imperative. The goal here wasn't to program for x different platforms - it was to show that you have what it takes to code up algorithms to solve some of the toughest theoretical CS problems, under the constraint of time.

    What happens when you work on a small embedded 8- or 16-bit processor?

    Especially on this type of work, you want to know exactly what the compiler is going to do - you don't want to throw in a bunch of junk code to make your code portable. Its called optimization . Anywayz, they weren't compiling on a small embedded processor, they were compiling on 32-bit x86 processors running NT where its assumed that the compiler being used by the contestants creates target code exactly the same as that used by the judges. If not - this should be mentioned.

    What happens when you work on a 64-bit system?

    Well, since you asked, in this case, the algorithm would still work. The fact remains that they weren't compiling on a 64-bit platform, tho - see above.

    The C language spec makes no guarantee for the exact range of an int.

    This isn't what's at issue here. Maybe in some other thread, your comment has relevance, however, nobody in this thread lacks knowledge of what is and isn't guaranteed by the C language spec. Nobody really cares what you know about the C language spec, either.

    I won't make the uninformed assumption that you've never been to a programming competition or that you don't know what its like to sit for hours trying to fix something that doesn't end up being a problem with your code ... but it would certainly seem that way. You're pounding issues that are highly irrelevant.

    as always - my $0.02

    Jobe

  • ...ever since 32-bit operating systems came around, I have pretty much assumed that int == 32-bits.

    That is a foolish assumption to make. What happens when you work on a small embedded 8- or 16-bit processor? What happens when you work on a 64-bit system?

    The C language spec makes no guarantee for the exact range of an int. It is only specified relative to the size of short and long. Sorry, but if you're going to be pedantic, you should be pedantic eveywhere and not assume type sizes that aren't specified exactly.

    That's a pretty harsh statement. I mean, speaking of foolish assumptions, aren't you assuming that they were using C? Most of what I've read implies they were using Delphi/Object Pascal.

  • How is creating their own database from scratch a "time-saving solution"?

    I don't know what webpage you're on, but when I accessed it, it looked fine to me. Spending X amount of time and effort to get a server that fails 1% of the time sure beats spending 10X amount of time and effort to get a server that fails 0% of the time. Especially since it's not mission-critical.

    Frac

  • I wasn't at the overall contest, and didn't see the problems given, but...

    I was webmaster and technical assistance for the 1999 Southeast Regional Contest [auburn.edu]. Bill Poucher came to the contest and was quite available. From my interaction with him (albeit on the side of the people administering the competition), he seemed quite reasonable and quite nice.

    This is not to say that the contest ran smoothly. A more detailed account [auburn.edu] of what the tech staff went through is available.

    There were some complaints, but they were handled politely, and by the end of the contest, everyone was tired but happy.

    Now, back to the international contest. I don't know what happened. I can guess that the judges got hit with a salvo of complaints and joined forces to repel the assualt :) Look at it from their point of view: Some of the teams got the problem the same way you did...this validates (in your mind) your solution. Other teams struggled but didn't. They then bring accusations that the problem was unfair. There are official channels for complaints as per the rules [baylor.edu] as listed on the official site.

    I said it before, and I'll repeat it. From my experience, Mr. Poucher is a decent and courteous individual. Managing a contest like this frays your nerves, and when people start bombarding you with challenges and accusations, it is easy to lose patience with them. There are also established procedures for dealing with appeals, and circumventing them by directly approaching the judges isn't likely to influence them positively.

    :)


    Who am I?
    Why am here?
    Where is the chocolate?
  • I don't know whether Delphi suffers from the same malady that Borland C++ suffers from, but Borland C++ defaults to 16-bit integers at the command-line, but 32-bits at the IDE level.

    That's a pretty clueless statement coming from the director of a programming competition. Borland C++ in fact ships with two command-line compilers:

    • bcc.exe - The DOS compiler, with 16-bit integers and pointers suitable for DOS' segmented memory model.
    • bcc32.exe - The win32 compiler, with 32-bit integers and pointers suitable for Win32's flat memory model.

    First rule of programming: Choose the right tool for the job. Apparently, this fellow skipped that class.

  • Dr. Poucher's a great guy. You just have to remember, in any contest, that you are catching the contest director in the absolute worst possible circumstances. I hope that you get to meet him some other time. Re: the ivory tower comment -- That may be a problem of some profs, but not Dr. Poucher -- he does a lot of "real world" work on the side.
    I have had two classes from Dr. Poucher (one theory and one software development) and I never knew him to be anything but fair to students.
    Ask yourself this, though, what would you have done in his situation? You knew that three people, whom you trusted, had checked the problem data. You knew that the problem had been solved on the data already. You've not encountered such a problem in ten years of running such a contest. You knew that every contestant at the contest was way keyed up and running on adrenaline.
    I agree about contests though. They are a lot of fun and you do learn a lot. This is true in general, IMHO. Texas has statewide academic contests in high schools -- participation in those takes away all the "test anxiety" many otherwise excellent students face. Programming contest is the same way. It gives you a great confidence, whether you win a contest or not, to just sit down and start slamming out code.
    Once again, I am sorry that you had to meet Dr. Poucher under such adverse circumstances!
    Zorn
  • That solution can't be right, can it?

    If "the matrix" is an adjacentcy matrix, just consider an edge between any two nodes. For a graph that would still be connected without the edge, the average shortest distance between nodes for the graph with the edge is bounded above by that for the graph without the edge, but, as the length of that edge goes to infinity, your sum of matrix elements diverges.

    Is there something more about graph, such as a space in which it can be embedded?

    I hope that this was not this guy's [other] issue with the competition.

  • "teams were forced to realize on their own what was going wrong, and guess at what kind of output the judges wanted for such a graph."

    And exactly how is that different from real world systems development? Sounds like a pretty realistic contest to me...

    sPh
  • by jesser ( 77961 )
    For high school students, there's a cool contest called usaco [usaco.org]. It's used to pick the US team for the international high-school level contests, but most of the participants are actually outside of the US. I think the problems are similar to (but easier than) the ACM problems.

    --

  • I sure is easy to tell who the flunkouts are on Slashdot...

    Ya. I is almust ase eezee to till hoo hus a moostirs tu... ;)
  • I am one of the University of Waterloo team members.

    It's nice that the Contest people are trying to make amends for the troubles they had in judging Problem F ("Page Hopping"), but the problem isn't so easily solved. The contest organizers can take into account whether or not a team had really solved Problem F or not, but they cannot take into account (nor did they say that they would) any of a dozen other factors that result from a problem like this (lost time, psychological factors, etc...). These factors make it nearly impossible to predict what the final standings would have been in a contest without mishap, and so I think that the effort on the part of the ACM contest judges may be misplaced.

    Anyways, the guarantee given about no team's ranking decreasing is much like the compression algorithm that never increases the size of the input...

    As it stands, I think that it is interesting that of the top six teams (those that got six or more questions), only Melbourne and Waterloo would have performed differently had the contest been run in an alternate universe without judging errors. A smuggled copy of the standings partway during the contest reveal that of these top six, only Melbourne and Waterloo did not get Problem F on the first try, although our team did eventually deduce the nature of the error with Problem F and submit a correct solution a little over halfway into the contest.

    I am certain that both teams suffered heavily due to the error in terms of lost time and added stress, but I feel that the Melbourne team (or perhaps just the one member) is unnecessarily bitter about it (perhaps due, in part, to a conversation the team had with Dr. Poucher afterwards, where he assured the team that, to the best of his knowledge, there were no problems with question F). For example, the Waterloo team (which was justifiable angry halfway during the contest when we discovered the nature of the error with Problem F) decided to get on with the business of doing the other problems. Afterwards, we decided to refrain from speculating on what we could have done with an extra hour or so of time during the contest, and we are still very happy about the entire contest and the things we learned on the way.

    I have less sympathy for people that gripe about spending all their time on problem F during the contest and getting nowhere, when there were 7 other problems that they could have tried their hand at. In particular, Problem E ("Internet Bandwidth") was a textbook network-flow problem, which didn't even require knowledge of any network flow theory. (Dynamic programming could have solved the problem within the time limits, although I didn't find anyone who did it this way.)

    Anyways, my point is that the problem of "fixing" the results is not as simple as rejudging F and giving an extra question to everyone who didn't get it during the contest, but did have a correct solution.

    Donny
  • Donny says:

    only Melbourne and Waterloo did not get Problem F on the first try,

    Tsinghua tried F and didn't get it, even in the rejudging. So they weren't directly affected in the same way as Waterloo and Melbourne. Nevertheless, they would have seen the scoreboard and may have made tactical decisions based on the perception that something was wrong with the data. For example, they may have had some trivial bug and 1/2 hour to spare at the end of the contest. Had a large fraction of the teams solved F, they would have had the savvy to spend some more time looking for a simple bug rather than perhaps working on a long shot.

    Bad judging affects everybody. Due to luck or skill, it affects some more than others. Tactical decisions are made in real time based on the judgements and there is no way to undo the effects. I encourage the ACM judging team to be more open about the process, to admit their mistakes, and to do a thorough and public post-mortem on this failure in their QA process.

    In my opinion this would be time and effort better spent than fiddling with the rankings in an attempt to please everyone.

    One thing that is done in the ECNA regionals is to write an "input checker" routine. Imagine that the contest question is recast as "print 'yes' if the input conforms to the given specification; otherwise print 'no'." I don't see how this year's F data could have passed such a test.

    Gordon Cormack
    Coach, Waterloo ACM Team

  • It is not quite accurate to say that they haven't had a judging problem in 10 years.

    They haven't had a judging problem that survived the end of the contest in 10 years.

    We have first-hand experience with two judging problems on previous finals. In 1995, we were awarded a balloon only to have it taken away much later. In 1997 we had a problem judged incorrect for some reason (probably due to ambiguity in the output). We moved on to another question and an hour or two later a "correct submission" message popped up for that question.

    Judging problems at the finals are rare but do occur. Judging problems at the regionals are more common, because they are run locally and those in charge may lack experience.

    In any event, judging problems do occur and dealing with them is part of contest preparation. However, it is important not to overestimate the chance of error - even in the worst-run regional, a judgement of "incorrect" is much more likely an error in the submission than an error in judging.

    Gordon Cormack
    Coach, Waterloo ACM team

No man is an island if he's on at least one mailing list.

Working...