Experimenting with that formula on SEDE did give me this query that, by the looks of it, generates a list of hot questions.
/*
(log(Qviews)*4) + ((Qanswers * Qscore)/5) + sum(Ascores)
--------------------------------------------------------
((QageInHours+1) - ((QageInHours - Qupdated)/2)) ^ 1.5
*/
select top 1000
(
(log(qviews)*4) + ((qanswers * qscore)/5) + sumascores
)/(
power( (qageinhours+1) - ((qageinhours - qupdated)/2), 1.5)
) as hotness
, id as [Post Link]
from (
select viewcount as qviews
, answercount as qanswers
, score as qscore
, ISNULL((select sum(score)
from posts
where parentid = q.id),0) as sumascores
, datediff(hh, creationdate, getdate()) as qageinhours
, datediff(hh, lastactivitydate, getdate()) as qupdated
, id
from posts q
where q.posttypeid = 1 -- Q
) rawdata
order by 1 desc -- yeah I know order by ordinal is bad
That gives the following definitons of the parameters used:
- qviews
The number of views of a question
- qanswers
The number of answers on the question
- qscore
The score (upvotes-downvotes) on the question
- ascores
The score (upvotes - downvotes) of an answer
- qageinhours
The difference in hours between question creationdate and now
- qupdated
The difference in hours between question lastactivitydate and now
I'm sure a developer will correct me if I missed anything.
Do notice that the formula is from the early days (expect changes to be made) and is for the Hot questions tab.
For network hotness data not available in SEDE is used according to this answer from David Fullerton.