@GameTheory both your comments are in violation of our Etiquette rules 1 through 7 and our Moderation rule 3. I believe that's a record. To be explicit, your comments are:

  • Low quality.
  • Disrespectful.
  • Insulting.
  • Conflictual.
  • Strongly political.
  • Intolerant.
  • Calling for violence.

As a consequence I will ban you for a week. Please change your behaviour when you are back.

@Based unless the AI overlords decide to put smart humans in charge of human affairs, in which case your Metaculus score is a strong signal of your superior smartness.

Sorry for the double-resolution, it turns out 1.8 is not the same as 1817502.

Hello everybody. To make this more confusing than it needed to be, I thought this was the perfect time to start using my real (first) name. No more @(tenthkrige), I'm @Sylvain now. Yes I'm french.

Now that that's taken care of, back to platitudes: I'm genuinely happy to be able to help around here. I always welcome criticisms of what I do or don't do (praise too!), and hope my small part in this weird endeavour will turn out positive for all involved. See you around or in the comments here.

Some context from a mildly informed Frenchman: Of the last 7 [French Presidents](https://en.wikipedia.org/wiki/List_of_Presidents_of_France), 3 where re-elected for a second term. (1 died in office before he could run a second time). The last 2 were not re-elected. Macron is the first Centrist President in quite a while. He profited from the breaking down of the 2.5-party system, and his success broke it even further. This was possible because there were no strong candidates in the traditional parties. Now, one is even less likely to emerge. Interesti...
Thank you for all your feedback! I genuinely think Metaculus has the most thoughtful user-base on the internet. The original plan had two flaws: the update would impact closed but unresolved contest questions (not to be confused with tournaments), and would make the points-per-question metric difficult to compare before and after the update. To address these concerns here is the new and improved plan: 1. Still remove the 50% end bonus. 2. Multiply all points by 1.5. 1. This will lower the discrepancy in the points-per-question metric, since early...
Well this was tedious. [link to sheet](https://docs.google.com/spreadsheets/d/1wtkwEh_fS6iaChiTL_8aX92XHbJhdE83EPnV9_uHiEo/edit?usp=sharing). Result: **538 has a score of 0.417, to PredictIt's 0.485, 538 wins!** Comments: + 538 had a forecast for American Samoa, but PredictIt didn't, and the question did not include it, so it was ignored (sorry America Samoans). + PredictIt's markets are delisted from the site once closed, but they are still [duckalbe](https://duckduckgo.com/?q=predictit+Vermont+Democratic+primary&t=lm&ia=web). + PredictIt's UI d...
@JonathanRay points are averaged over the question's lifetime. If your points were negative for most of the time, then your average is negative too. @haukurth when a question resolves early, points are truncated proportionally to the fraction of its lifetime the question stayed open (thanks @BrunoParga for pointing out the relevant message.). This is necessary to preserve the properness of the scoring. You can get a more detailed explanation in our FAQ section aptly titled [Why did I only get a few points when I was right?](https://www.metaculus.com/hel...

@DanielFilan speaking for myself, because major powers seem (on average) a lot less belligerent now than they were over that period. I would for example be extremely surprised if EU countries declared war on each other in the next 30 years.

An important meta point I haven't seen made (and which I would have made sooner had real life not intervened): If the 2020-2016 difference had been in the other direction, I think it is very unlikely anybody would have looked into it as deeply, and we would have resolved negative without much discussion. If this counterfactual is correct, then we are setting a norm of only overturning close calls *that cost us points*, and we are eroding the value of Metaculus' endeavour. Tbc I'm not saying that looking more deeply into surprising events is bad, but t...

It seems to me that:

  1. This should resolve negative right now. Absence of evidence is actually evidence of absence.

  2. The resolution criteria clearly state "before 21 January 2021, Trump uses the pardon power", which does not require that we know about it, so if we at any point learn that he did, this will be re-resolved positive.

Sorry, I resolved to 45 rather than 45000000. Small difference, really. Fixed now.

To people who genuinely have a high credence: I find your position extremely optimistic, but ok.

To people who predicted high because in most scenarios that should resolve this negative, this doesn't resolve at all: Please consider what will be more important then. A few more MIPs might not matter all that much, but having helped correctly inform humanity on the dangers of the transition is its own reward.

I think some predictors missed that this is conditional on the event happening before 2200 (otherwise I don't see why it's 4x as likely to happen in the 2190s as in the 2180s).


All the legally cast votes have been counted, and we are happy to report that @beala and @casens have won this election!

Congratulations to all the candidates! I can say that personally I would have been happy with any two of you!

By the way:

We now have
markdown tables and
I think we
can all agree
that is pretty
cool !

Markdown syntax:

We | now | have
markdown | tables | and
I | think | we
can | all | agree
that | is | pretty
cool | ! | 
Insomnia idea: a checklist displayed during question writing/editing and during the pending and upcoming phases. Checklists are good at preventing avoidable oversights ([source](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3489073/)), also known as dumb mistakes. **Please keep the discussion below focused on the merits of such a checklist, without mentioning any particular item you want on it.** If the idea is judged good, I'll make a Discussion page to crowd-source the contents. <small>Of course no pressure on the dev team for implementation, but it'l...
@(galen) we can reformulate my three question types above as: 1. Questions are clustered with heavy overlap. 2. Questions are clustered with minimum overlap. 3. Questions are unclustered. The particular meaning I was gesturing at: for 2., not only do you want minimum overlap ("carving"), but you want the borders between questions to make sense ("at the joints"). Even more than that: we want them to make "semantic" sense (i.e. we care about it in the world), not just "syntactic" sense (i.e. the resolution criteria are simple). Success example: the [...
@(EJ) they seem to count the number of articles you read on their site. Purging their cookies should work. [ETA:] Long story short: NYT have an internal slack channel with 2k people in it. Some non-journalists (devs and assorted tech people apparently) politely said doxxing Scott was not ok. Some journalists politely replied it's not doxxing because "doxxing" implies the author intends for the subject to be harassed. Week-sauce defence obviously, but it shows Scott should not (imo) have used the word doxx, because not everybody is embarassed to motte a...