O4 Mirror: https://scp-wiki.wikidot.com/forum/t-16759617/8k-retrospective
Alright, so 8K just wrapped up and the contest marshals are all back from break.
This is not a thread to discuss voting irregularities and how to combat them, that is a seperate thread.
What worked
What didn't work
What should be changed for future Kcons
Speaking from a non contest marshall perspective and my thoughts on reading is that our articles are frankly getting far too long. The average article is up 30% since 7Kcon and reading every entry in the contest would take about 60 hours with a million words.
Speaking entirely from personal opinion I would like a ~10-15 thousand word limit for future Kcons to possibly cut this down. Though given the topic I am also against implementing this unless the same problem occurs in 9K.
Though I will say that there is zero correlation between article length and placing, data can be provided upon request.
One thing that I would actually like is a request in the contest rules to keep it under a count, this will not be a requirement but a request.
I said this somewhere else and I’m not sure where — I think that instead of any kind of word limit, we should have some sort of incentive to shorter articles. Rather than discouraging — or outright banning — one kind of article, we should encourage the other kind.
One idea for this: there is an upvote offset that is smaller the closer you get to some word limit, above which scores are calculated normally. There would need to be some mathematical analysis of word lengths and scores to really figure out how big this should be and what word count ranges it could apply to, but this might look like a +30 vote adjustment for an article under 500 words, a +10 vote adjustment for an article that’s 7,000 words, and no vote adjustment for an article above 10,000 words. (These numbers are all examples — I don’t know what appropriate figures look like.)
This would have the impact of encouraging short form articles for the kCon, but doing nothing to forbid long form. It would be a slight hurdle, but by no means insurmountable.
Alternatively to the above, we could have better promotion and marketing for articles under a certain word count. I’m not sure what else we could give to short form articles beyond upvote adjustments and promotion, but if anyone has other ideas for things I'd welcome suggestions. Anything that’s tangible and we can clearly promise to entrants as a reward for being under certain word count thresholds.
can I get some nice "no signatures on my forum"
sigma-9 css machine broke
understandable have a nice day
Here are some obvious operational issues we should resolve:
An x000 contest has multiple purposes, but one above all others.
The primary purpose of the contest is to decide what gets the x000 slot. Getting more people to find SCP, celebrating writing and giving writers something to be proud of, giving people stuff to read, letting authors claim slots early, and satisfying the fans with a big community events are all desirable secondary purposes, but they are not the primary purpose.
Others on O4 have disagreed with this primary purpose or that the primary purpose should be changed. I don't agree with those arguments and those stances so I won't be arguing for them but they're worth acknowledging.
The contest structure is meant to create some semblance of fairness, as opposed to, hypothetically, having a council of 5 admins just decide what gets the slot. The "voting should be determined on quality and not what you think should win" is more recent than I expected. Apparently, the oldest record I can find of it as an official rule is 6000, even if it was counted under the shenanigans clause in prior contests. I think "downvoting something during the duration of a contest and then later going back to upvote it after the contest ends" is hard to read as anything but shenanigans, to be frank.
In a naive model, "people", that being the voting audience, the readers, whoever, can only determine the most suitable piece to be SCP-x000 out of all the entries if they are able to read all the entries. We might assume that things that don't get read aren't "deserving to win" the x000 slot, but that's tautological even if it aligns with reality. The fundamental truth is that there is so much writing produced for every kcon that it is difficult for anyone with a full or part time job or schooling to read it all considering quality. We could outright say it's ok to downvote a work after 1000 words if it doesn't grab you, but that's enforcing/creating a cultural norm which is very different from making robust contest rules.
I brought up a model based on the concept of progressive taxation of upvotes in staff chat. To summarize, it was framed as a penalty of 1 upvote for the purpose of scoring for every 1,000 words above 10,000. Stormbreath's suggestion is akin to a progressive negative income tax, similar to what is implemented in the United States of America as the Earned Income Tax Credit. Under my model, a 20,000 word piece would be penalized by 10 upvotes, which has rarely been the make-or-break for the x000 slot. My proposal was fundamentally a tax — a tax on anyone who believes that their words deserve more attention from the readers than those of their peers. If so, then let them show it. This proved to be controversial, even though it is a midpoint compromise between the current status and an ironclad hard wordcount limit, which would be undeniably a clamp on creative freedom.
Again, however: the purpose of the X000 contest is to choose the next X000. With a 10,000 word limit, I do not think this proposal would have changed the results of the last few kcons. You see, I think the mere possibility that one might have to take the 8011 slot instead of the 8010 slot because of a rule accounting issue would be incentive enough encourage people to stick within the word limits. But this is also served by the supposedly common knowledge that longer pieces get less reads. However, how obvious is this information? Can we truly say that everyone entering the contest would be aware of it? I am aware of at least one 13,000 word work sitting at around +20 entered into a non-kcon by a newer author who was not properly informed of the risk-reward balance.
If that is the primary purpose and ultimate intended effect of vote bonuses or penalties for wordcount, then I think it may be sufficient to clarify the informational deficiencies:
Even if we implement an upvote modifier, I think we should put in H1 header text providing statistics about the previous kcon: total wordcount, wordcount of the winner, wordcount-weighted average rating(calculated as upvote per piece*words per piece/total words in contest), and rating-weighted average wordcount (calculated as upvote per piece*words per piece/total net votes in contest) very prominently above or as part of the rules of the new contest.
Allow me to walk through a sample calculation of the final statistic with a small sample of 4 works:
Winner: 8000 words, +446 votes
average work: 15,000 words, +100 votes
outlier longform: 25,000 words, +10 upvotes
outlier shortform: 500 words, +300 votes
only the winner is based on an actual work but these are all ballpark figures. So if we do the calculation we would get
wordcount-weighted average rating: (8000 * 446 + 15000 * 100 + 25000 * 10 + 500 * 300)/(8000 + 15000 + 25000 + 500) = 113 upvotes
upvote-weighted average wordcount: (8000 * 446 + 15000 * 100 + 25000 * 10 + 500 * 300)/(446 + 100 + 10 +300) = 6388 words
What this calculation tells us, and tells future prospective authors, is actually a matter for debate but one interpretation is that this would illustrate that once you adjust for how long the works are, the average work gets around +113. And once you adjust for individual weighting, the average work is around 6400 words long. So if we combine the two, we communicate that modestly-long pieces do moderately well, that the winner was not excessively long, and that .
Numbers are made up but you see the effect.
Additionally, in the interest of further transparency, even if we don't limit wordcounts we should require that all kcon pieces include a wordcount next to or below the rating module. Users should be aware of how long works are. AO3 has a wordcount and chapter count. If you pick up a physical book you can see how roughly long it is. I don't think this is an unreasonable requirement for a piece of writing entered into a contest.
Our users want more readable kcons. Our users deserve to continue to play a part in fair kcon selection. Everyone deserves the illusion of a fair shot at an X000 slot. Authors deserve creative freedom. If the situation is to be made more fair, then at bare minimum authors should be made aware of the environment they are writing in.
If, having all the information, authors still feel that they can perform well by writing 30k words, then they will be fully informed of the implications of their decisions.
What should be changed for future Kcons
I'd like to see an early posting period for less-experienced authors. Giving new authors' pieces a few days of not having to share the spotlight with big-name authors would allow them to get more views and votes, whether up or down. Broadly, I'm thinking of something like this:
I think this would be helpful regardless of any modifications that we make to the voting procedure itself.
Edit: The early posting could be limited to entries that have received crit (and therefore greenlights, for first-time authors.)
Perhaps the following stratification:
can I get some nice "no signatures on my forum"
sigma-9 css machine broke
understandable have a nice day
Preview component half-worked, but the timing of its usage being halfway through the contest made it less effective than it could've been. This should be remedied when it's taken more into account for the next contest.
As for word count limit, generally I think 8kon was an outlier and we shouldn't worry too much about it; but if we have to, I prefer giving bonus modifier for short articles as incentives rather than penalizing long articles or instating hard word count limit — The max modifier should be in the realm of 5-10 upvotes.
Kufat's idea fro multi-stage posting dates seems interesting.
My thesis here is pretty broad: Let people aside from the Contest Team help with the contest.
Now this is a bit difficult given the current way of doing things, but I think with a few minor changes it'll be for the best.
1. Announce the date of K-Cons in advance of them actually occurring.
I know the surprise is fun, believe me I know, but if we're talking about making changes to kcons to make them more manageable, for readers and writers and staff then the first thing that's gotta go is the surprise of the when.
If people know when kcon is, they can get things ready in a staff capacity. Discord Team can set up channels and roles and such, as can Crit, as can things like reading club and tagging and whatever else. We can get whatever newfangled slot selector thing up and running beforehand. I don't know, I can't think of everything off the top of my head, but just knowing a kcon is coming and when lets staff members get ready and set time aside in advance to be free to help in whatever way is needed.
I know that in terms of discoverability Discord Team, IO, Reading Club and the changes made to navigation seemed to be big contributors to helping with article discoverability, this despite many of these not being planned or in play at the very start of the contest. Giving advanced notice would allow for better discoverability in the future, which would I believe mitigate the issues that people are jumping straight to "change the contest" to solve, while also being minimally disruptive to the contest.
2. Consider allowing staff members that are not on the Contest Team be able to help marshalling the contest. Only if they aren't participating in the contest itself, of course. To my understanding currently if you join the Contest Team to help out you have to commit to years of not joining contests, since future contests are planned or discussed far in advance. Doesn't strictly have to be the case. After the contest is announced as coming you can recruit more people for this contest alone in some way, if that would be helpful
I think doing a word count limit or forcing there to be teams or anything else that would stifle creativity on such a large level is a frankly ridiculous thing to jump straight to. There's plenty of other small changes that can and should be implemented beforehand, and any big changes should IMO be testdriven in other, smaller contests first.
Wasn't initially going to write on this thread since I already communicated my opinion on this to Naepic in staffchat at some point, and I do trust the Contest team to make a good decision on this stuff ultimate. But I was told I should throw that in here anyway so hey.
I should note that I did sign up to be a marshal as someone who is not a member of contests, you are free to ask there.
For one I want to say that I'm pretty satisfied with the time slots this time. I think one week posting and then the rest for reading worked well and I think we can continue doing it like this.
With regards to length I'm also of the mind that this is an issue in large part caused by the Fantasy theme which invites more lengthy pieces to be written.
I'm not really too keen on a direct limit of word-count and I'm also a bit wary for an incentive system to write shorter works, though I could live with it. I think easier solutions are something like displaying articles in order of shortest work posted instead of, or additionally to, counter-chronological order on the contest hub.
