Re^2: RFC: Hide Very Bad Answers From Visitors
by jdporter (Paladin) on Jul 24, 2018 at 21:44 UTC
|
Yeah, I was thinking vaguely of something along those lines as well. But I don't think using -$NORM is quite what we want, because that would mean that as the average node "quality" goes up, we become even more tolerant of even worse nodes. Instead, how about maybe something like -20 + $NORM. That would result in a threshold of -10.3 today. To put this in context: such a threshold would result in having no nodes posted today hidden, but all of this week's worst nodes of the week hidden (and then some).
I reckon we are the only monastery ever to have a dungeon stuffed with 16 ,000 zombies.
| [reply] [d/l] [select] |
|
| [reply] |
|
| [reply] |
|
If you're going to use a formula like that, I would lobby for clamping to negative: $NORM<20 ? -20 + $NORM : -1;. Otherwise, if our $NORM quality spikes suddenly to 25*, some poor schmuck who "only" got +4 for a mostly-unnoticed post for some "Re^9" gets his reply hidden.
*: okay, probably not likely with modern Best Nodes scores... but the idea is that I believe positive-voted nodes shouldn't be evaluated as "very bad", and so protection should be built in against that admittedly-unlikely event.
| [reply] [d/l] |
|
Definitely agree with the clamping suggestion (for any system with a dynamic threshold), although I'd lower the "high" end clamp a bit, probably to -3, so that middling-to-decent responses don't get hidden simply because one or two people are having a bad day (or just want to be assholes) and give it drive-by downvotes before it receives any upvotes.
| [reply] |
|
that would mean that as the average node "quality" goes up, we become even more tolerant of even worse nodes.
That only holds true if the actual mean quality increases. If $NORM goes up (or down) purely as a function of how many votes are cast in total (which it will), then using -$NORM seems perfectly rational.
Instead, how about maybe something like -20 + $NORM.
That also suffers form the same problem as just using -7: it puts an arbitrary fixed point into things which becomes problematic if $NORM changes greatly from where it is today.
| [reply] [d/l] [select] |
|
| [reply] [d/l] |
|
Re^2: RFC: Hide Very Bad Answers From Visitors
by Anonymous Monk on Jul 25, 2018 at 07:04 UTC
|
Perhaps we should consider the distribution of the votes and hide posts with rating in 10th percentile and below. The number 0.1 is subject to discussion, of course.
Pro: no need for complicated formulae based on $NORM. Contra: it still may not mean what we think it means.
| [reply] [d/l] [select] |
|
| [reply] |
|
What if quantiles are computed over last year over whole section / whole website and not just the current thread? That should give us a representative threshold of a "really bad node rating", I think.
| [reply] |