Sock puppet accounts have been removed, and the question resolves positively. Sorry for the confusion, all.

**New updates!** - For continuous questions, you can now hover over the prediction graph to see the relative probability that you have assigned to any given point. - Also for continuous questions, you now have the option of viewing your prediction as either a PDF (probability density function, the default) or a CDF (continuous distribution function). To unlock the CDF you must either be level 3 or unlock the corresponding power on your [profile page](/accounts/my-account/). - Bug fix: on the profile page, accounts with over 100 comments will now load ad...
# New feature time! Our new developer @M4 has been working hard on implementing some of the excellent suggestions that have been made in this thread. The biggest new feature is the new Metaculus *reminder system*. For a nominal tachyon fee, anyone can write themselves a reminder on any question and then get emailed their reminder at a later date. You can pick the exact date that you want to be reminded, or you can be reminded to come back to a question once it opens, or you can use reminders to alert you to a big change in the community prediction. We'...
Metaculus now allows nested replies in the comments. If you reply to someone (using the "reply" button) your comment will appear directly below theirs. This way we can have multiple discussions going on at once on a single question, and it makes it easier to reply to an old comment while keeping the proper context. Many thanks to @ghabs whose generously donated programming time made this possible. All compliments should go to him, all complaints to me. We've got some backend updates under way, but once those are complete we'll start making progress on ...

@Pshyeah wrote:

I'd like to see highlighting (or some other visual marker) for new comments since I last read a question. Could be a horizontal line, could be a slightly other background colour, doesn't matter, the point is, when I see there are x new comments on a question I'm interested in, then I either have to vaguely recall which comment was last, or count from the top.

This is now implemented! If you go to a question with unread comments, there will be a bright yellow line separating the last unread comment from the first read comment.

@(contact.sylvainchevalier) wrote: > I'd like a way to extract my track record data to a usable format, ideally CSV. Maybe as a level reward ? @(jzima) wrote: > And perhaps the overall metaculus track record as wel? Sure, I don't see why not. You can now view a json export of your track record at [/accounts/my-profile/track-record-export/](/accounts/my-profile/track-record-export/) and an export of the overall Metaculus track record at [/questions/track-record/export/](/questions/track-record/export/). Sorry for not providing it in CSV. We'll likely be...
**New updates!** With this update you can embed mini versions of Metaculus questions on your own (non-Metaculus) webpages. Just click the "embed" button above and to the right of the question title, copy the code, and put it wherever you'd like. In addition, we've made a few style tweaks and bug fixes: - When you haven't yet made a prediction on a question, the prediction sliders will now look a little different. They'll be gray (binary questions) or dotted (graph on continuous questions). - The community prediction will now be hidden until 10 differe...
I should add the *reason* why we can't give people full points for an early-resolving question. If we did, it'd be super easy to game the system to maximize points. Let's say you think there's only a 10% chance that Cohen got a jail sentence at this hearing, and a 1% chance that anyone would receive a sentence at any other time. If only your final prediction counted, then you should predict 99% the morning of the hearing. If he gets sentenced to jail then you would get a big pile of points, and if he doesn't then you'd just dial your prediction back to 1...

Two small new updates:

  • Continuous predictions now allow for much (3x) narrower distributions. If you're feeling bullish, now's the time to update your predictions.

  • There used to be some cases where the Metaculus prediction on the continuous prediction graph wouldn't match the text and, relatedly, cases where the community prediction on the track record didn't match what it showed on individual question pages. Both of these should now be fixed.

@(Jotto999) wrote: > Would it be possible for me to not only list all of the questions that I've made predictions on, but also rank them by how long ago it was that I changed my prediction or made a comment? I've started keeping a spreadsheet because it's too easy to neglect a forecast by merely forgetting. I want that Tetlockian habit of more frequent updating, with whatever information I've recently come across. This is now implemented! On the questions list page, go to "my predictions" and then order by "oldest prediction". Your prediction that has ...

@frxtz We recently made some big backend changes which may be the root cause of this bug. I wasn't able to reproduce it, so I didn't immediately fix it, but I'll now bump it towards the top of my to-do. Thanks!

Over the [50 resolved questions](/questions/?resolve_time__gt=2017-07-22T12:00:00Z&resolve_time__lt=2018-01-05T02:28:00Z&status=resolved&order_by=resolve_time) since this question's opening, 3 players both predicted at least 25 times and on average beat the Metaculus prediction's Brier scores: - @AndrewMcKnight averaged 0.120 vs 0.150 for Metaculus - @jzima averaged 0.162 vs 0.186 for Metaculus - @traviswfisher averaged 0.146 vs 0.148 for Metaculus Note that the average Metaculus score is different for different players, since different players predict...

@chris.j.cundy: Done! You should now see a little indicator that shows where the result lies along distribution.

@Jgalt In order to change the minimum width I also had to adjust the way the width of the distribution relates to the width of the sliders, so although your prior distributions shouldn't have changed, the placement of the slider knobs may look different.

@darkives Sorry about that! The button should now be invisible until either 10 people have predicted or two days have passed, whichever comes first.

@frxtz Thanks for catching that. Should be fixed for the next batch of reminders!

@Jgalt It looks like this was a problem that affected a few early predictors on each question, and you were an early predictor on each of them. It should be fixed now, and your proper points should get added to your score.

We're working on the root problem, so hopefully you won't see it again. If you do, let us know! Thanks!

@nextbigfuture The Metaculus points are awarded based on your average score over the course of the question with a large bonus (50%) given to your prediction at the end. Unfortunately, it looks like you had Joe as the underdog for most of the time, and your high probability at the end wasn't enough to wash it out. You can see the details in the "Your point history" plot above the comments. Better luck next time!

@Jgalt Evidently my algorithm for calculating the position of tick marks on log scales starts to act a little funny when the log scale is almost linear. If you make your window narrower you ought to see more labels appear.

@predictors would anyone be opposed if we just made it a linear scale? This would technically change people's predictions, but only by a small amount. The end points would stay the same, but a prediction at 6.2 would shift to 6.3.

@traviswfisher With about 220 questions the chance that a user wins due to luck drops to 10%. By 500 questions it drops to less than 1%. Of course, there's no guarantee that similar numbers would hold if we reran the experiment. The chance for a lucky player goes way up when there are more players making independent predictions, especially if a bunch of those players predict pretty close to the Metaculus prediction. If we improve the Metaculus prediction such that it is much more confident in the correct outcome then the chance for a lucky player should...