We've just rolled out a set of improvements to numerical questions, including:

- Slightly different slider and percentile behavior (moving the side sliders no longer moves the center one.)

- Vertical rather than horizontal listing of percentile values for open questions.

- The curve is now a [logistic distribution](https://en.wikipedia.org/wiki/Logistic_distribution) rather than a gaussian.  This has slightly less punishing tail behavior.

and most importantly:  The edges of range can now be "open" so that you can choose your distribution to assign probability to "outside the range." Thus for example in the [supersonic transport question](https://www.metaculus.com/questions/969/when-will-commercial-supersonic-flight-return/), any resolution after 2030 gives the same points, and can be considered "never." A few notes on this:

- You can't manually set the "out of range" percentage: it's just integrated from how much of the distribution is out of range.  A future update will very likely allow multi-modal distributions as a level-unlockable feature, though, which will give pretty much full control over this.

- The logistic distribution makes a nice exponential when the center of the distribution is out-of-bounds, which was the other reason for its adoption.

- There's still no "points along the way" for date questions, but our hope is that the next system will make it easier to make questions shorter-term (since this new system in some sense combines the date with the "will happen by" systems), and we might consider adopting a policy of only awarding points at question end-date, rather than at early resolution.

- We anticipate an open upper end being commonly used in date questions, and very rarely an open lower end.  Open ends will be less common for other numerical questions, though they may become more useful with multimodal distributions.

Thoughts?  Most of the changes are tweakable before being made permanent.