# Intro For the last few weeks I’ve been trying to learn how to do stuff in R. I decided to practice on this question, and plot the results of simple monte carlo simulations. I am neither a programmer, nor do I have a background in math, so assume that I’m partly [just winging it](https://xkcd.com/2048/). I’ve included links to the code used in the Code section. There was a lot of failure and struggle leading up to this, but now I think the results are good enough to be worth posting. The simulations that I’ve shown below could have easily been done...

@never_harsh Yes you are. There is an enormous confidence difference between, say, 90% and 99%. Projects fail sometimes in the real world, even by competent teams. A mere <=1% for all the ways a specific company could fail a big, novel project like landing a probe on mars is optimistic.

A reminder to new people: many of you have been flooding this question with demented forecasts.

The probability of SpaceX landing anything on Mars is currently 85%, but landing living humans is at 90%. That's incoherent math-wise. But also, humans (who easily perish from minor changes in their environment) should be seen as really hard to land in general, even without comparing to a robotic ship simply doing a touchdown.

— edited by Jotto

@isinlor wrote:

Could we stop using imperial units?

My car gets forty rods to the hogshead, and that's the way I likes it!

— edited by Jotto

1. Signalling prestige and whatever else [the elephant](https://www.goodreads.com/book/show/28820444-the-elephant-in-the-brain) wants. It's pleasing to have a signature skill I'm known for. 2. I can't tolerate the idea of being poorly calibrated. Especially on anything I had time to think about, and that isn't too chaotic or far out. And to find out that a dart-throwing chimp beat me across domains would be anguish for me. I can't accept that, and would do whatever it takes to beat the chimp. 3. I like reading up on a variety of topics. I'm als...
Three upward updates for me. #Incubation [SCMP 2020-Feb-19:](https://www.scmp.com/news/china/society/article/3051235/rare-coronavirus-can-can-have-long-incubation-period-warns) >“But there was only one out of all the 1,099 cases [with an incubation period of 24 days] and only 13 had incubation periods longer than 14 days.” >The World Health Organisation previously estimated that the incubation period of the coronavirus ranges between one to 12.5 days with five to six days as the median. 13 / 1,099 = 1.18% of cases last longer than the 14-day stand...

In tournament registration fields, the academic degree one forces me to pick an option, or "Other". I don't have any and never went to university. I know this hobby selects for an academic crowd but you should probably have a None field so blue-collar or non-academic people don't feel alienated.

Link rot is a major long-term problem on the internet. Gwern has a fine article on this topic. It could be helpful to save source backups on the Wayback Machine. Especially for links that are in question bodies, and which have a timeframe of 10+ years.

Data point: I have benefited from many of @Jgalt's posts, and I predict that Jgalt's future posts will bear many fruits.

There is testing for the live virus, and there's also testing for antibodies. The antibody test does not necessarily mean the person has the live virus at the time of the test. But it shows the person has had it. Antibody reactions linger long after the person had the live virus in their system.

So suppose if someone had a mild case that went undiagnosed, in April. Then suppose an antibody test is conducted on December 1st, that shows they must've had it at some point. That's positive resolution, right?

— edited by Jotto

In the hangout today, we brought up some questions where we differed a lot from the community median. I noticed that it was hard to search for this. Could we get a way to sort questions by how much our forecast disagrees with the community? Presumably in odds or relative percent, rather than percentage.

This would be useful. If some news was making people update away from you, then it will be prominent in such a list order. It also emphasizes questions that have disagreement (and update potential).

— edited by Jotto

I'm getting a (crudely-made) historical base rate roughly around one UK recession starting every ten years. [Sheet](https://docs.google.com/spreadsheets/d/16QsEyCSbUkHx_Z5gOoXQj7xEcpaeIs0K3LpIA3LHBr0/edit?usp=sharing), where my odds-format gave a 19.718% chance of a recession, or 4.071-to-1 odds in favor of no recession starting in the UK during the next 2 years. If I divide this by 2 as a fudge factor, and maybe because Brexit makes a recession more likely, I get 2.0357-to-1 odds against a recession in the timeframe, or a 32.941% chance of a recession...

I am hereby implementing the Jgalt gets props from me feature, for working hard on sharing information about a variety of topics and keeping the crowd more up-to-date, and in particular because bigger comment incentives aren't implemented, yet still the amount of effort is highly visible. Thank you, Jgalt!

EDIT: Grr, I discovered a calculation mistake I made when I first posted this. Base rate at the time of posting this was 31%, not 36%. I've made the [spreadsheet](https://docs.google.com/spreadsheets/d/1h1Lz8Z2TbfFuJTqrvGLkO5pwBqgoJhGhRS6SyY8O798/edit#gid=0) editable by anyone. Feel free to add another sheet and dabble with the data, which I got from the CBOE [website](http://www.cboe.com/products/vix-index-volatility/vix-options-and-futures/vix-index/vix-historical-data). The first sheet has data from 1992 to 2003. After that there was a change in ...
That guy who did the Bernie Madoff fraud investigation has put out a report saying that General Electric is conducting massive fraud. GE disagrees. [Reuters](https://www.reuters.com/article/us-ge-accounts/ge-shares-fall-on-madoff-whistleblower-calling-its-finances-a-fraud-idUSKCN1V519M) Seems like pretty scandalous and juicy material for a medium-term question! EDIT: I have written [a question](https://www.metaculus.com/questions/3109/will-general-electric-ge-file-for-chapter-7-or-chapter-11-bankruptcy-before-sep-18th-of-2024/) about this, which puts...
Our question-writing culture has a (non-urgent) problem of overly-influential styles. We have a lot of questions that put comment-level material in the question body. Here is an example in [this question](https://www.metaculus.com/questions/7539/if-simulation-when-created/). The first sentence says "Some researchers suggest the odds of our current reality being a simulation are very high." IMO this is very noisy. Even if it was quantified and accurate, do we assume it will be true years from now? Not necessarily. Intellectual fads come and go. So ...
Updating downward on China's heavy metal pollution.[[1]]( https://twitter.com/toad_spotted/status/1230118171698442240 ) This is a downward update to the extent that the rest of the world has less heavy metal poisoning. And to the extent that heavy metal exposure helps a pandemic spread. The Twitter source puts extra emphasis on cadmium, which seems to help influenza spread better.[[2]](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3588091/) A hedge is that I don't know how well that mechanism transfers to COVID-2019. But here are a couple other paper...

@Guba wrote:

Thank you for the lecture, mr. Professor. I know as perfectly as you do that...

I didn't see AngraMainyu as lecturing you or being condescending. Take it easy.

It would be nice to have some way to "save" distributions and reload different attempts. The Save Snapshot button on [Elicit](https://elicit.ought.org/) does let you do something like that. Note that some distortion is common, particularly with more complex distributions. It is common to tinker with a distribution, and then decide to return to the previous version. One might need to look at changes in the point-score feedback. Also, changes in how one component will affect the weight of the rest. It's too tedious, all this manual undoing and redoin...

@Tchetche Okay, that's enough. Please tweet your cheerleading to Elon, rather than posting it here.