7
\$\begingroup\$

Back when I first posted a challenge, it was closed as unclear, on the basis that – despite having an objective victory condition – the victory condition allowed ties. After searching Meta, I can't find a rule about this anywhere, so I think it's about time to start a discussion on what is and isn't acceptable in terms of victory condition tiebreaks.

So my question is: a challenge with an objective victory condition that allows you to compare two entries to see which is better, what rules for tiebreaks are acceptable? Is it acceptable to have no tiebreak (and allow "equally good" submissions to tie for the win)? If there is a tiebreak, are there any restrictions on the sorts of tiebreak that should be allowed? In each case, why?

\$\endgroup\$
  • 1
    \$\begingroup\$ Related. \$\endgroup\$ – Martin Ender Dec 10 '16 at 21:29
  • \$\begingroup\$ More context would be useful, because I don't think it's the same to have a tiny probability of a tie as to have all answers scoring identically. \$\endgroup\$ – Peter Taylor Dec 13 '16 at 8:53
3
\$\begingroup\$

In most cases, yes. The submissions don't have to be perfect on those conditions if there isn't much competition, though. Having an unbeatable answer also doesn't mean the challenge is closed for new submissions, if they use new languages or approaches.

This isn't ideal for a challenge. But in many cases the asker didn't have an idea about how easy or difficult their challenge is, not to say how pointless it is if there isn't a tiebreaker, and when there will be an answer with the best possible score. The point is not to be a victory condition, but to be objective.

We don't really even need the implications of winning a challenge, as it doesn't seem to be a feature we like to begin with.

The asker could make exceptions for some very special types of challenges, if there is a valid reason.

\$\endgroup\$
-5
\$\begingroup\$

The winning condition should always select exactly one winner, within reasonable limits

This is handled by the fact that, by default, the first submission to reach the tied score wins. Assuming that two submissions are not posted/edited at the exact same time to reach the exact same score (which is a reasonable assumption), there will always be exactly one winner under that rule.

If for any reason it was impossible to choose exactly one winner, the winning criterion would fail to be objective, meaning it isn't allowed. Our site's scope is for challenges with exactly 1 winner, tips, and special-circumstance challenges with no winners (the showcase and catalogs).

\$\endgroup\$
  • 6
    \$\begingroup\$ You're just describing the current situation. I was hoping for arguments for and against, rather than just descriptions. \$\endgroup\$ – user62131 Dec 11 '16 at 12:38
-6
\$\begingroup\$

Challenges and their victory conditions should, as far as possible, give scope to improve any entry.

When the objective victory condition rule has been discussed in the past, the general consensus (and one that I agree with) is that the victory condition exists to drive creativity. If a challenge's requirements are "do X", and not "do X, but Y is better", there's no scope to do anything else once you've just done X. On the other hand, if it's possible to improve your entry, that motivates you to actually improve it, increasing the quality of the site. So a victory condition that makes improvement possible is really the heart of the "objective victory condition" rule, and the reason for the rule is more important than the rule itself.

As such, I think we should modify our rule for victory criteria as follows: a challenge should be designed so that, no matter what you've done for the challenge so far, it's always possible to do more, and to know what is required of you to do that. In most challenges, this is equivalent to requiring an objective victory criterion, because without one, it's impossible to know how to improve.

This also means that the important factor isn't competing against other users; it's competing against yourself. The purpose of a victory condition isn't to tell whether user A's answer is better than user B's answer. It's to let user A improve themselves as much as possible, pushing themself as far as they want to. In other words, ties between users are completely harmless; but unimprovable answers are very dangerous, as they render the competition pointless from then on.

So I'd suggest that we change our list of off-topic reasons like this:

  • Instead of Questions without an objective primary winning criterion are off-topic, as they make it impossible to indisputably decide which entry should win.,
  • our rule (and off-topic reasoning) should be Questions which can be given an answer that cannot objectively be improved on are off-topic, as the posting of a "perfect answer" makes competition impossible..

Implications of this:

Victory conditions with ties are acceptable, and should be encouraged in situations where no obvious tiebreak exists.

Sometimes in a challenge, there's a way to make a "major improvement" to the goal which requires creativity, and "minor improvements" on the tiebreak which don't really. If there's no tiebreak, then improving an answer needs to be a major improvement, and if that's the heart of the challenge, we want to be focusing effort on it. (See this question for more details about situations in which tiebreaks actively hurt the question.)

"Earliest submission" is never viable as a tiebreak, and in particular, is inferior to having no tiebreak at all.

Once you've posted a submission, it's clearly physically impossible to post another submission earlier without a time machine. In other words, this is about as unimprovable as it gets for solutions that have already been posted, and as such discourages rather than encourages competition.

Before you've posted your answer, the tiebreak also has negative implications. Many of the challenges posted here are fairly easy (because the reputation system gives higher rewards for easier challenges, something that's impossible to do anything about without closing easy challenges on sight – something which would be a bad idea, as they're fun for new golfers to compete in – or getting diamond moderators to do a community wiki conversion or to delete the resulting reputation manually, which would be way too much work). As such, there's often an objectively right answer. A good example of a worthwhile tiebreak would be to encourage people to improve their posts via better descriptions, explanations, and the like (which is close to a guided ); we already have consensus that doing so is desirable. The FGITW tiebreak does the exact opposite of that, encouraging people to spend a minimum effort on their submission (and perhaps even posting it before it's properly tested), on the basis that spending any extra effort takes more time. And on easy challenges (note: I define "easy" as any challenge in which the shortest answer is to directly translate the words of the specification into Jelly :-P), the byte count is necessarily going to end up tied, so the tiebreak will control the actual victory. Note that the situation with no tiebreak would be better here, as it would remove any incentive to post a bad post simply to do better on the tiebreak.

\$\endgroup\$
  • \$\begingroup\$ The "earliest submission" rule is pretty unavoidable. Often, there are many possible answers of the same length, differing in unimportant details. If you remove the "earliest submission" rule, people will be obliged to post unimportant slightly modified answers. \$\endgroup\$ – anatolyg Dec 11 '16 at 19:31
  • 3
    \$\begingroup\$ There are a lot of things I can agree with in this answer, but overall my vote is down because I cannot agree with closing challenges for not making the optimal solution unattainable. I always strive to choose the parameters of my challenges to ensure an optimal solution will never be reached, to keep the competition open ended. However, it is rarely possible to tell for certain whether this has been achieved, so I think it is something we should encourage, rather than require. \$\endgroup\$ – trichoplax Dec 11 '16 at 22:53

You must log in to answer this question.