# Statistics on Sandbox Usage

Recently, I've noticed that sandbox usage has increased dramatically compared to what I remember. I've decided to gather some statistics on sandbox usage, which I am posting here in hopes that it is useful information.

I am going to offer some analysis of the data. For most of my analysis, Mk VI (the newest and currently open) is ignored, but I still provide data on it.

Sandbox | Open date | Duration | # answers | # posted | # deleted | # views
Mk I    | Mar 4'11  |    76    |     17    |     8    |     0     |   795
Mk II   | May 19'11 |   160    |     24    |    13    |     1     |  1297
Mk III  | Oct 26'11 |   223    |     20    |     8    |     5     |  1001
Mk IV   | Jun 15'12 |   346    |     24    |     5    |     2     |  1228
Mk V    | May 27'13 |   207    |     32    |     7    |     5     |   994
Mk VI   | Dec 20'13 |     6    |      9    |     0    |     1     |   191

Duration  = number of days open before being retired
# answers = total number of proposed questions in sandbox
# posted  = number of proposed questions posted to main site*
# deleted = number of proposed questions deleted by poster
# views   = number of times sandbox was viewed


*Note: I counted # posted by looking through the sandbox for links to the posted questions, which were often in the form of [posted]. There may be some errors, but I believe that it is reasonably accurate.

Sandbox Duration is probably a function of two things: the rate of new answers and the definition of "too crowded." A less active sandbox lasts for a longer period of time because it takes longer to fill up. It also appears, however, that the idea of "too crowded" has changed over time, as later sandboxes got rather big before being closed. Mk I had 17 answers before being closed, while Mk V had 32 answers.

It is interesting that # posted is almost the same in Mks I, III, and V. Mk II saw a big spike, while Mk IV was a little low.

The # deleted seems to be rather high in Mk III and Mk V.

From this raw data, we can compute some useful numbers. The first is % posted, which just so happens to be 41/126 = 32.54%. An more rigorous calculation gives the 95% confidence interval of (.2498, .4115). This shows that there is actually a decent amount of uncertainty as to the true rate of sandbox success.

Since only 41 out of 126 answers were posted to the main site, while only 14 were deleted, there are still 71 viable sandbox questions.

Sandbox Popularity Across Time:
Sandbox     |  Mk I  | Mk II  | Mk III | Mk IV  |  Mk V  | Mk VI
Answers/day |   .22  |  .15   |  .09   |  .07   |  .16   |  1.5
% posted   |   47%  |  54%   |  40%   |  21%   |  22%   |   0%
% deleted  |    0%  |   4%   |  25%   |   8%   |  16%   |  11%
Views/day   |  10.5  |  8.1   |  4.5   |  3.5   |  4.8   | 31.8
(days = duration of sandbox)


*Note: Views/day is heavily influenced by a number of factors. For example, a sandbox can continue to gather views after being retired, which inflates the number upwards. Also, posts tend to receive an initial burst of views for the first few days, which aren't representative of popularity over the lifespan of the post. Mk VI, for example, is almost brand new.

Here, we can see that the answers/day is somewhat U-shaped. The early and later sandboxes have similar numbers, while the middle two have very low numbers (.09 and .07). I suppose that this is somehow related to the popularity of the site as a whole.

The numbers for % posted generally decline over time. Some of this is may be because older sandboxes have had more time to incubate. That isn't the whole reason, though, because there was more than enough time to incubate while the older sandboxes were still open. I believe that the decline in success rate is real. This could indicate a decrease in the quality of submissions or an increase in expectations (raising the bar).

It is interesting that the first two sandboxes had very low % deleted numbers. Perhaps users are currently more likely to actually delete their propositions, rather than just abandoning them through inactivity.

The Views/day also follows a U-shape, with the first and last sandboxes having higher numbers. These are probably meaningless, as Mk I probably receives views for simply being the first sandbox, while Mk VI has received several views for simply being a new meta post. Mks III, IV, and V probably have realistic values.

It is also useful to talk about the utility value of sandboxes to the main site. It's hard to measure this, because I feel that sandboxes aren't really used to improve quantity. They are used more to improve quality. Quantity-wise, however, there are 1283 questions in the main site. Out of those, 41 or so went through sandboxes first, so about 3%. This doesn't mean much, however, without taking into account the quality of that 3%, which may be hard to measure (maybe using upvote counts).

While performing this research, I found that there was a great opportunity. Earlier, I said that there were still 71 proposed questions which were neither deleted nor posted. If we could revisit the first few sandboxes, polish up and publish their proposed questions, then we could probably add a few dozen good posts to the main site. This would take some effort though, and some organization.

I hope that this data is interesting to you. I thought it was interesting. I think that having data is important for any decision.

If you spot any errors, please tell me.

• Is the #answers including the deleted answers, or excluding them? In case of excluding, Sandbox VI has 8 answers and not 9. – ProgramFOX Dec 27 '13 at 9:01
• It is including them. – PhiNotPi Dec 27 '13 at 13:30