PredictionBook is now read-only ( read more ).

A Level 7 (Chernobyl/2011 Japan level) nuclear accident will take place by end of 2020.

Created by ceterisparibus on 2011-11-23; known on 2021-01-01; judged wrong by PseudonymousUser on 2021-01-01.

  • ceterisparibus estimated 40% on 2011-11-23
  • gwern said “https://en.wikipedia.org/wiki/International_Nuclear_Event_Scale#Level_7 to be judged, of course, by the IAEA” on 2011-11-23
  • gwern estimated 30% and said “nuclear power plants operating since 1954 or the last 57 years; 2 level 7 accidents for 1 every 28.5 years. odds of next one within 10 years? poisson distribution, but won’t bother calculating it” on 2011-11-23
  • JoshuaZ estimated 25% and said “Risk goes up as plants age, but goes down as new accidents make people more careful. Not sure that Poisson dis is the right model. ” on 2011-11-23
  • gwern said “arguably as plants get older, the crews get more experienced and the procedures more tested, etc. the human element was key in fukushima, 3 mile island, and chernobyl. doesn’t the dispersion of accidents look poisson-like?” on 2011-11-23
  • PseudonymousUser estimated 45% on 2011-11-24
  • Anubhav said “Intuition would suggest that the probability of an accident can’t remain constant with time, so Poisson isn’t the right model. Empirical data overrules intuition, but…. three data points? ” on 2011-11-24
  • JoshuaZ said “Two data points. Three Mile Island was only level 7. But I think Gwern is correct in that empirically modeling accident frequencies in other contexts Poisson seems to work. So would be right to do. But 2 data points? Hard to make that reliable. ” on 2011-11-24
  • Anubhav said “http://en.wikipedia.org/wiki/List_of_civilian_nuclear_accidentsI’ll see if excel can fit Poisson to this list. ” on 2011-11-24
  • Anubhav said “Actually, I’m not sure which of the lists here I should use. http://en.wikipedia.org/wiki/Lists_of_nuclear_disasters_and_radioactive_incidents” on 2011-11-24
  • JoshuaZ said “Well, if both lists end up looking Poisson then it should be safe. If neither does then that’s not a good model. If only one does shrug? ” on 2011-11-24
  • Anubhav said “http://dl.dropbox.com/u/50825627/nuclear%20accidents.xlsxTried fitting civilian accidents/ accidents + incidents (see the worksheets). No match. Military accidents are obviously not Poisson, a single glance tells you that.” on 2011-11-24
  • gwern said “yeah, military nuclear accidents obviously goes down as time passes – new developments cease, old nukes are refined and miniaturized, etc. soviet union was pretty careless early on. expect a big drop post-cold war.” on 2011-11-24
  • Anubhav said “Yeah, there was a big drop in the 70s and a bigger one post cold-war. Anyhow, civilian accidents per decade tend to cluster around the 3-6 range, and the probability of there being 3,4,5 or 6 accidents seems to be roughly equal. ” on 2011-11-25
  • Anubhav said “If I include civilian ‘incidents’, it’s the same story, but number of accidents are now restricted to the 4-8 range. I notice that the upper bound is twice the lower bound in each case, but I doubt that’s anything but a fluke” on 2011-11-25
  • Anubhav said “http://bit.ly/u3H7sM See the BIN worksheet. Used that model with a binomial distribution assuming probability of accident being level 7 = 1/26. Given that we’ve already had one lv7 accident this decade, the probability of another one = 7%. ” on 2011-11-25
  • Anubhav said “Of course, the binomial assumption isn’t exactly a good one, seeing as we know it doesn’t follow Poisson. But I couldn’t think of any other way to solve this… ” on 2011-11-25
  • Jayson Virissimo said “This thread is a good example of why we really need a Less Wrong style comment system attached to the predictions.” on 2011-11-25
  • Anubhav said “^Seconded. Whoever gets rid of the bullets and the quotation marks will earn my undying gratitude. ” on 2011-11-25
  • Anubhav estimated 7% on 2011-11-26
  • chemotaxis101 estimated 40% on 2011-11-26
  • RobertLumley estimated 25% on 2011-11-27
  • gimpf estimated 25% on 2011-11-28
  • TrE estimated 7% on 2011-11-29
  • Grognor estimated 23% on 2011-12-05
  • Serge estimated 17% on 2011-12-06
  • ceterisparibus changed their prediction from “A Level 7(Chernobyl/2011 Japan level) nuclear accident will take place by end of 2020.” on 2011-12-10
  • saturn estimated 13% on 2011-12-11
  • JoshuaZ estimated 31% and said “Adjusting upwards due to this: http://www.thebulletin.org/web-edition/op-eds/second-life-the-questionable-safety-of-life-extensions-russian-nuclear-power-plan” on 2011-12-13
  • Tiresias estimated 10% on 2011-12-20
  • Serge estimated 20% on 2012-03-06
  • seifip estimated 20% on 2013-04-14
  • themusicgod1 estimated 38% on 2016-09-30
  • pranomostro estimated 20% on 2018-11-25
  • pranomostro estimated 10% on 2018-11-25
  • Baeboo estimated 3% on 2018-11-26
  • Baeboo estimated 5% on 2018-11-26
  • Bruno Parga said “Mark wrong?” on 2021-01-01
  • PseudonymousUser said “I don’t see one on https://en.wikipedia.org/wiki/List_of_nuclear_power_accidents_by_country” on 2021-01-01
  • PseudonymousUser   judged this prediction wrong on 2021-01-01.