Tavern on the Meta - 2017-05-25

« first day (2486 days earlier) ← previous day next day → last day (2534 days later) »

12:04 AM

@JourneymanGeek I feel that way about SO.

@JourneymanGeek In theory, it would've made it better, since we can add an unlimited number of topics, each with their own link. In practice, some of those topics previously had sub-sections with anchors, so at least initially it got harder to link to things like "Avoid overt self-promotion".

(I eventually added an entirely separate topic for that particular section, primarily so I could link to it. We got a lot of questions in team@ from spammers)

Plus SO's busted Higher Score -> More Likely To Appear On Related List -> Gets More Votes uncontrolled positive feedback loop basically guarantees that no user will ever get close to the top rep spots at any point in the future, since it's just constantly streaming 200/day to a handful of people, so that kills that part of the game, too.

leaderboards

That's what leaderboards were created for. Bit of a shame they're on a separate site though.

I'm such a big fan of the decaying rep idea, lol. I know nobody else likes it.

So I keep it to myself.

There are certain behaviors that idea encourages that are... even worse than FGITW.

12:07 AM

That and a per post absolute rep cap with a full recalc.

Yeah I know, it's got 1000 downsides.

per-post cap ain't all bad, really. We could've made that work. Hell of a thing to retrofit though.

The shake-up would be fun.

I think Kevin M modeled it at one point; it wouldn't do much to knock the top users down.

So, if that's your goal, you'd have to set the cap really, really low.

...which brings us back to making FGITW literally the only way to earn rep

That's OK, I mean, if they stay active, they should stay on top.

At the moment nobody can ever overtake them, even if they disappear. It's guaranteed not possible. At least with a cap some day in the far future it'd be possible.

Well unless you're getting crazy accepts and bounties, I suppose.

Jeff wrote something about this a few years back, too lazy to search for it right now but the gist was... You really, really don't want to make "post lots of stuff all the time constantly forever" your primary ranking metric.

12:10 AM

You know it might be fun if league standing was displayed on user cards instead of rep.

The sort of person who joins a form and writes constantly, all-day every-day, is not the sort of person you generally want as the face of your site.

<---

Well, if they're contributing good content, no harm in that.

Where's the motivation to contribute good content?

You got two people, one posts 100 answers/day the other posts 20 but they're good answers not just whatever the top google result was. First person ranks 5x higher than the second.

Same as now. It doesn't change anything. You still gain rep relative to other people in the same way.

Yeah but that's how it works now, too.

It's totally possible to "grind for rep" now on SO, but it is not the most effective way to get rep.

Folks talk shit about "necro rep", but it balances out the grinders by a lot

12:14 AM

Then it's also not the most effec... well, meh. Whatever. It's all academic and moot.

Yeah the My Money For Jam query does have surprising results, I'll give it that.

@JasonC Here it is:

133

A: Revisiting the rep cap (yes, again)

I ran some numbers because I love the science. For the proposed rules of: You can only gain 100 rep per post You only gain rep for votes occurring within 6 months of the post's creation I queried the amount of reputation gained under these rules, and what percent that is of the user's reputa...

> For the proposed rules of:

- You can only gain 100 rep per post
- You only gain rep for votes occurring within 6 months of the post's creation

I queried the amount of reputation gained under these rules, and what percent that is of the user's reputation earned under the current system from post up and down votes.

Yeah, I've seen that. I found it lacking because it didn't cover the most important stat (I thought) which was the effect on rep increase rate. It only focused on the difference at a given point in time.

I don't think having huge gaps in the top spots is bad at all, I'm more bummed by the guaranteed bleak future for everybody else.

Basically the top spots are essentially permanently hard-coded to a given set of users at this point, which sort of defeats the game aspect. It's the same kind of mild demotivation that a high score list dominated by impossible hacked scores has in a game.

Not quite as demotivating since they are totally legit. Sort of. Depending on how you view SE's various big feedback loops.

Anyways right now I'm thinking about the help center. With that in mind, before I write this MSE post about how help center search sucks, is there already a post about how help center search sucks? I just skimmed through 10 pages of search results and didn't find much except a few old specific status-completed items.

Journeyman Geek

@JasonC or you walk into a pub and see an insanely high score on frogger

Pac-Man is the one I always lost sleep (and quarters) over. Grrrr.

Journeyman Geek

I mean, I can beat John T, who hasn't been on the site in years.

and I've overtaken a bunch of people

12:22 AM

Me too. But you'll never have the glory of being in the top 5. Nobody will, except the permanent top 5.

Journeyman Geek

(also, I kinda hope gravvity overtakes harrymc. Fair and square of course)

@JasonC ... on SU I'm number 6 ;)

Lol I mean on SO. Nice, btw.

Journeyman Geek

OH WAIT NO! I AM NUMBER 5!

user315433

Keep the rep income rules as they are, but introduce progressive tax. On April 15 of every year, users lose X% of their income from last calendar year.

I decided I want my goal to be get 20k on as many sites as possible. Spread myself out.

Journeyman Geek

12:23 AM

Cause I overtook wil.

@Gerry Can bounties be tax deductable?

user315433

And make them fill out more complicated forms if they earned over 100,000 that year.

Lol

And take their actual stuff if they don't.

It can be redistributed as swag.

user315433

@JasonC Only if they are on other people's questions. Then it's charity. Otherwise an investment.

Oh man

Journeyman Geek

12:24 AM

Or having a top 5 over 5 years rather than showing the all time ranking (which people can find anyway)

@Shog9 Before I try to come up with constructive ideas for help center search, how does it work? Is it just vanilla text searching? Do you guys use elasticsearch or anything?

Also is it entirely pointless to suggest some "Did you find this article helpful?"/"Is this what you were looking for?" feedback + search term tracking to improve results?

user315433

12:51 AM

Puzzling is a site with one of the "youngest" mod teams compared to site age: nobody is left from pro tem mods, the current moderators spent <1 year at the helm.

Journeyman Geek

@Shog9 Eventually though, the people who read through it the most, are probably the people already engaged enough on the site to link it to a new user when they need guidence. Its great there, just not really something I typically expect someone new to SE to read...

user315433

@a25bedc5-3d09-41b8-82fb-ea6c353d75ae Sorry to hear about your username situation. How did it happen?

Journeyman Geek

0_0

@Gerry I've seen uuid usernames before. It was a thing. Lemme see if I can dig it up.

Well there's meta.stackexchange.com/questions/279526/… but that's not what I remember reading, it was older than that. It's a fuzzy memory, now it's kind of bugging me.

meta.stackexchange.com/questions/177706/… happened, dunno what to make of it

meta.stackexchange.com/questions/290915/… is similar. But that's all I found. Oh well.

user315433

1:07 AM

> Good news: chat links on Stack Overflow are now https:// (no 301s yet; coming soon). Bad news: I broke most of the internal APIs. Oops. -- Nick Craver at 6:05 PM - 24 May 2017

user202362

so why does uber get so many negative news when airbnb escaped pretty unscathed?

user202362

in theory, they are in the same category right?

1:24 AM

@JasonC yes

@JasonC I don't think that's a terrible idea

user202362

1:59 AM

user image

5:56 AM

AI 2 - 0 Human

Google conquered the world again

user202362

in order to conquer the world, you need to be able to make physical difference

user202362

waiting for google to do gardening ...

delv-pls 20K deep something meta.stackexchange.com/questions/296395/… /cc @Bart @ShadowWizard

user202362

user image

user202362

per hour ... and it's cheaper nowadays because construction sector has cooled down

user202362

6:03 AM

I had to fork out USD$80 an hour when I hired tradies 2 years ago

7:20 AM

> Don't Panic

7:35 AM

delv-pls 20K Xterm meta.stackexchange.com/questions/296397/… /cc @Bart @sha

1 hour later…

8:36 AM

lol third Socratic badge...

@rene done, waiting for @Bart now...

@Derpy why?

@rene kaboom

\o/

@Telkitty not really

user202362

9:11 AM

@ShadowWizard yes, really

@ShadowWizard you forgot your towel again, didn't you?

9:32 AM

@Derpy panicking

1 hour later…

10:46 AM

→ 15 messages moved to Chimney

→ 3 messages moved to Chimney

2 hours later…

12:25 PM

@ShadowWizard man, there's lots of planning behind that 6y old issue.. — Paweł ♦ 23 secs ago

lol

user202362

1:07 PM

Your dog does behave like you, scientists prove - does that mean, if I see a particular breed of dog consistently acting dumb on youtube videos. I can assume owners of that particular breed of dog is the same?

Folks, have this come up in the past on the main meta.SE?

1

Q: I want to vote on questions in the Close Votes review queue

Sometimes I would like to vote on a question while reviewing close votes. I can already edit the question and add a comment from the same page. Is there a particular reason for the vote functionality being hidden in that view?

feature-request

user315433

@NickAlexeev It's essentially declined, although without a tag

user315433

I guess there is no status-declined tag because it was labelled "discussion". The message is clear, though.

user315433

Turns out, Computer Science Educators is in beta since Tuesday and I didn't even notice. That's how excited I am about this beta.

@Gerry didn't notice a new beta?!?!?! That's almost as bad as @Bart isn't on starboard......

user315433

1:20 PM

It's redundant when we have Nick Craver's Twitter account.

user315433

> Better curriculum names: "Computer Pseudoscience" "500 tips for compiler wars” “How to implement collections every framework already has" -- Nick Craver at 6:22 PM - 23 May 2017

@Gerry but he does not announce beta sites there

user315433

I mean the site is...

user315433

> Better curriculum uses of 4 years than computer science: “How to ask better questions” “How to Google stuff” “How to work with people" -- Nick Craver at 6:26 PM - 23 May 2017

user315433

It would make sense for @StackExchange account to announce new sites, but...

user315433

1:29 PM

@NickAlexeev A better reference is Why is voting removed from new review system?

user315433

There it is again, "new".

2:33 PM

@Gerry everything is new at some point. How you suggest we call something new when it's still new?

(and there is older version)

user image

Must admit, digg doing a good job with their titles. They are catchy.

user315433

@ShadowWizard SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED SPONSORED

user315433

I don't know, looks like direct-to-spam content to me.

2:50 PM

@Gerry heh, that's some bug either on digg side in their HTML format, or Google side for picking wrong parts when searching for "digg" (which I did to show them all together), in the actual view it's better:

user image

(the other messages below and above are more spammy)

user315433

The top answer on CSed so far says to teach Python 3 over 2. Okay... Top question: Analogy for teaching recursion

user315433

And the most pressing meta-issue is that Summer is coming and everyone will be on vacation when the private beta is evaluated.

user315433

3:28 PM

> Fun fact: when you analyze data to find "What is the most _______ programming language", the answer is usually "Haskell" pic.twitter.com/dbHPrtDGyw -- David Robinson at 7:59 AM - 25 May 2017

Heh

I wonder how close the correlation between "most popular" and "most activity in a tag" really is. To me it always reads as "most confusing".

user315433

Well yeah, exiting vim is a tremendously popular activity.

Or even "least popular", in that there aren't an exhaustive amount of resources about it already, hence the activity on SE.

user315433

Nobody's asking how to exit MS Word.

That is because that exit itself more then enough

3:32 PM

According to the common SE blog interpretation, that means that most people are very happy to stay in MS Word forever.

4:31 PM

@rene meta.stackoverflow.com/a/306141/616460 -- Lol. When reading that post I wondered what they were talking about because I didn't remember the banner. I also facepalmed a little. Then I felt compelled to clean my screen when I saw your screenshot.

Glad I could help

5:01 PM

Oh disorganization. Our annual upstate festival is starting today. I'm not so involved this year and not going. I just got a text "JASON. EMERGENCY. CALL ME. DO YOU HAVE OUR RADIOS???". Yeah, the day of the festival you guys just realized you somehow lost 30-something radios. $5000-something worth of equipment. This is why I'm taking the year off.

My response: "Nope. Have fun, suckaaaz!"

I hope they have all charged their mobile phones then. Assuming they have a mobile network there.

AT&T is up there, everything else is pretty spotty. It's in a valley and there's only one cell tower, sort of far away. That's what they'll probably do though. I've used Zello in the past, it's pretty good in a pinch. Also if the signal dies it will store comms and deliver them as soon as the signal comes back, which is cool.

Breaking News on CNN: Upstate festival ends in chaos, a search warrant is out for Jason C, stealing all radio's

Probably for the better, too, those radios are so illegal here. They overlap with a bunch of HAM frequencies. And they've definitely pissed operators off in the past.

@JasonC when something really bads happen everyone will click to that tower ...

I recall an incident at a festival at the beach at Hoek van Holland where police officers were under threat and could only phone in to the police station with their mobile phones because their radios didn't work properly on that remote location. That became nasty with shots fired I believe.

5:19 PM

Ugh

We keep a good, tight relationship with the local cops. We get our permits, we give them contact info, we mind the gate and keep them posted, and in return they generally respect our privacy. So far in the 6 years we've been doing this one we've been able to manage all problems internally. So far. Knock on wood.

5:34 PM

→ 16 messages moved to Chimney

[ SmokeDetector | MS ] Link at end of answer: SSL certificate for website - no legal company name? by Harsh Tiwari on serverfault.com

Wait what

@hichris123 Where the heck did you find 16 more messages

@JasonC From about three days ago...

Sigh... you always make me feel so inadequate.

Well, I can't just move ~3 messages, that looks kinda sad...

5:37 PM

Lol

5:51 PM

Darn it.. sore throat and coughs prevent me from sleeping...

nyquil

humidifier

drink water

soup

Thanks, I'll do them together

[ SmokeDetector | MS ] Offensive answer detected: How do I stop my cat from destroying everything? by user9554 on pets.SE

user315433

> It's a Maine Coon cat but he's Satan.

lol

user315433

5:58 PM

Is there a way to get "remaining flags" count from the API?

Like how many you have remaining to cast for the day?

user315433

Yes, that. Apparently not, although I don't see any FR to add it, either.

I don't think so. It's neither in the users object nor the flag_option object.

@Gerry So there is a related FR: meta.stackexchange.com/questions/245416/…, because if you could even get your history of flags you could compute remaining flags yourself. Also meta.stackexchange.com/a/262143/230261 pretty clearly states that flag info isn't available in 2.2, which confirms that at least.

user315433

The request to make mod messages anonymous moved to status-planned

I saw that

6:05 PM

@SmokeDetector Those comments...

And then I remind myself why I'm annoyed over things that literally have no impact...

Kasra's been active recently

No repro on a Pixel running version 1.0.94 -- are you using version 1.0.94? I'm looking through the crash logs for this right now. — Kasra Rahjerdi ♦ 26 mins ago

Probably because the apps have been getting a ton of flak on MSE and MSO lately.

user315433

@hichris123 If they annoy you, that's some impact already. And there's a chance someone out there is annoyed too.

Seems to be trying to make nice by being active and also focusing a bit more on FRs and stuff.

user315433

Flags a bunch of posts and comments as annoying

6:12 PM

@Gerry Yeah, but don't I have better things to do than be annoyed over that?

Evidence suggests you do not, so just roll with it, lol.

Be one with the annoyance. Don't fight it. Embrace it.

reminded me of bjbot

[ SmokeDetector | MS ] Pattern-matching website in answer: How does bodybuilding or weightlifting affect singer's voice? by Isaiah Feken on music.SE

user315433

NAA IMO

user315433

TIL wolframalpha.com parses LaTeX: I can directly copy-paste stuff like \frac{1}{32e^{2\pi}} \tan^2 \left(\frac{1}{256}\right) from a document without any edits.

user315433

6:18 PM

That looks like TMT... which stands for Too Many TLAs.

7:07 PM

Due to the recent changes with the http/https switch some internal routes are broken which could explain why you see this problem, despite providing the correct url. For obvious reasons I'm not going to test account deletion myself. — rene 3 mins ago

@rene Come on, don't be a wuuuusssss!

:D

user image

7:34 PM

@JasonC looks like we have a happy customer: meta.stackexchange.com/questions/296418/… sort of ...

See, it's not all negative experiences!

On the next mod election we should pull a stat for how many users you helped to get their account deleted. Highest score gets a diamond.

Lol. files away for 2017 winter bash hat idea

8:07 PM

@JasonC not an issue, btw. (cc @rene) Or at least, no repro here. :)

@AdamLear yeah, we saw the user managed it in the end. i wonder though what their first attempts were...

Most commonly, it's someone trying to paste in the se.com network profile

OK, my edit of the answer on how to delete your account should help a bit then

There's also possibly a bug in the regex where it fails if you don't include the display name and stop at the user id. I'm trying to remember if I fixed that or not.

so http://meta.stackexchange.com/users/155160/adam-lear works, but http://meta.stackexchange.com/users/155160/ doesn't

Ah

8:11 PM

I think I was going to tweak that, but realized that it opens up a potential hole - enforcing the full URL avoids having to validate the user id... which we don't always have since you don't have to be logged in to request deletion... and then I went to work on more important things. :P

lol

(if the trailing / was omitted, that is)

Let me add that in the answer as well ....

I WILL NEVER LET YOU DO THAT

already did: meta.stackexchange.com/a/7979/158100

Puh!

8:15 PM

I'm so... powerless...

8:25 PM

!!/reboot

!!/alive

@Mithrandir You doubt me?

1 hour later…

9:54 PM

Crap

@Shog9 Bad JavaScript skills and an infinite loop got me IP banned from api.stackexchange.com... will it pass eventually?

probably

I get myself blocked pretty regularly

Oh it's back. But now I have a backoff time of 12 hours. :(

Oh well, lol.

Now You Know

Sigh

Go play with js.cytoscape.org - it's pretty great.

9:56 PM

JSFiddle infinite loops bring chrome to a grinding halt, there's no way out. I think I made 9000 requests in about 10 seconds.

Whoa neat

It kind of reminds me of old praystation.net

I'm feeding data from SEDE into it to graph connections between users. Such fun!

Oh that does sound fun

Oh hey that reminds me. There's some algorithm, I think it's a basic statistics one, for finding correlated items in groups. Um, like, if as input each sample was a question and each value was a tag, it could spit out a big matrix with info about what tags were strongly associated with other ones. Do you know what that's called?

(I have no idea)

Yeah, I'm always afraid to ask questions on unfamiliar sites. Confidence, Jason, confidence. You can do this.

just search for "related tags"

I'd be shocked if this hasn't already been answered

user315433

10:33 PM

@JasonC clustering

user315433

Not an algorithm, but the problem for which a bunch of algorithms have been invented. And a tag on Stats site.

Hmm

So from my cursory read about clustering algorithms it seems that given e.g. a set of questions and their tags, it could tell me which questions were similar to eachother. Is it also appropriate for giving me quantified relationships between tags?

That is, I think, geometrically, I'd be looking for correlation between dimensions, in, say, a 50000-dimensional space, where each question is a point.

Oh wait

Can I not use PCA for this somehow

user315433

Doesn't look like a natural PCA problem to me. All features are 0-1, a question either has a tag or doesn't.

user315433

Would PCA work for boolean (binary) data types? by Alvin Nunez on stats.stackexchange.com

user315433

There is also graphoverflow.com/graphs/stackoverflow-tag-relations.html (which does not do any clustering, it's a straightforward visualization tool)

10:41 PM

Oh. Although that question there looks like a gold mine of possibly helpful info.

That's neat looking.

user315433

TechGraph: A landscape of a tag/tech in Stack Overflow by ccy on stackapps.com is what you were looking for. Web app at graphofknowledge.appspot.com

Huh, look at that. How cool. Now I just need to find out the algorithm there.

user315433

Apparently, the web app is not open source. :(

Yeah...

{"error_id":502,"error_message":"too many requests from this IP, more requests available in 42339 seconds","error_name":"throttle_violation"}

Getting there... one second at a time...

This CorEx algorithm looks promising.

11:26 PM

Oh, whoa

I have Matlab installed on here. I totally forgot about that.

Matlab 2011, lol. But hey, neat.

user315433

Matlab does Hierarchical Clustering but only with Statistics and Machine Learning Toolbox (additional $$$) and possibly not the 2011 version.

Somehow I do have the Statistics toolbox (no machine learning in 2011) and it does have hierarchical clustering. Sweet. Now I just need to relearn matlab.

I remember why I have it now, I did a job around that time for my old employer from pittsburgh, some matlab -> c++ port of something, and they gave me a copy.

Yeah the clustering stuff, it'll find question similarity, but...

I mean I suppose I could make it work if I treated each tag as a 14000000-dimensional point of which questions it was in.

Seems weird but maybe something interesting could be discovered there.

Also kind of unwieldy, that's a lot of dimensions.

I'm actually not using this for tags, I intend to use it to find packs of closely-associated close-voters.

« first day (2486 days earlier) ← previous day next day → last day (2534 days later) »

Transcript for

May '1725

Tavern on the Meta

MY GOD, IT'S FULL OF STARS AND BACKLOG! General friendly chit-...

bug crickets discussion excuses support waffles

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2024 Stack Exchange Inc; legal

mobile