SQL | some indicators of the game industry

Game data background “ Log in ” and “ Log out ” Two tables .
among ,“ role id” Field uniquely identifies the player .
Two days before opening the game （2022-08-13 to 2022-08-14） The login log and logout log of are as follows

Log in

Log out

1、 First day of service （2022-08-13）, Game DAU And the retention rate of the next day

select count(distinct t1. role id) as DAU, count(distinct t2. role id)/count(distinct t1. role id) as  Next day retention rate 
from (select * from  Log in  where  date ='2022-08-13') t1
left join
	(select * from  Log in  where  date ='2022-08-14') t2
on t1. role id=t2. role id

2、 On the first day of service opening （2022-08-13） Grade distribution , That is, the number of characters in each level

select  role id, Role level 
from (select *,rank() over(partition by  role id order by  Time  desc) rk
	  from (select  date , The login time  as  Time , role id, Role level 
			from  Log in 
			where  date ='2022-08-13'
			union all
			select  date , Logout time  as  Time , role id, Role level 
			from  Log out 
			where  date ='2022-08-13') t1
			) t2
where rk=1

3、 On the first day of service opening （2022-08-13） Level stagnation rate .
（ Level stagnation rate = The number of characters staying at this level / The total number of people who have reached this level ; If the player does not log out of the log , Then use the level information of the login log .）

WITH a as 
(select  Role level ,count(distinct  role id) as co
from (select  date , The login time  as  Time , role id, Role level 
	  from  Log in 
	  where  date ='2022-08-13'
	  union all
	  select  date , Logout time  as  Time , role id, Role level 
	  from  Log out 
	  where  date ='2022-08-13') t1
group by  Role level 
),
b as 
(select  Role level ,count(*) as co
from (select *,rank() over(partition by  role id order by  Time  desc) rk
	  from (select  date , The login time  as  Time , role id, Role level 
			from  Log in 
			where  date ='2022-08-13'
			union all
			select  date , Logout time  as  Time , role id, Role level 
			from  Log out 
			where  date ='2022-08-13') t1
	  ) t2
where rk=1
group by  Role level 
)
select a. Role level  as  role ,ifnull(b.co/a.co,0) as  Role stagnation rate 
from a left join b on a. Role level =b. Role level 
order by a. Role level 

4、 Please according to the time when players log in and out , Count the total online time of each player every day .
（ If there is no corresponding log out after the player logs in , You can use the same day 23:59:59 As the logout time , The calculation between times can consider using the timestamp function unix_timestamp.） Here I directly use timestampdiff, Choose according to your needs day\minute\second etc.

select t1. date ,t1. role id,sum(timestampdiff(second,t1. The login time ,ifnull(t2. Logout time ,concat(t1. date ,'23:59:59')))/60) as ' Duration /min'
from (select *,rank() over(partition by  role id order by  The login time ) rk
			from  Log in ) t1
left join 
			(select *,rank() over(partition by  role id order by  Logout time ) rk
			from  Log out ) t2
on t1. role id=t2. role id and t1.rk=t2.rk and t1. date =t2. date 
group by t1. date ,t1. role id
order by t1. date ,t1. role id

5、 Please according to the time when players log in and out , Count the distribution of online hours of each player on the first day of service opening .
（ If there is no corresponding log out after the player logs in , You can use the same day 23:59:59 As the logout time , The calculation between times can consider using the timestamp function unix_timestamp.【 Distinguish between online time periods ：0-30min,30min-1h,1-2h,>2h】

select  Duration distribution ,count(*) as  Number of players 
from 
	(select *,
				(case when min<=30 then '0-30min'
						 when min>30 and min<=60 then '30min-1h'
						 when min>60 and min<=120 then '1-2h'
						 else '>2h' end) as ' Duration distribution '
		from (select t1. date ,t1. role id,sum(timestampdiff(second,t1. The login time ,ifnull(t2. Logout time ,concat(t1. date ,'23:59:59')))/60) min
					from (select *,rank() over(partition by  role id order by  The login time ) rk
								from  Log in 
								) t1
					left join 
								(select *,rank() over(partition by  role id order by  Logout time ) rk
								from  Log out 
								) t2
					on t1. role id=t2. role id and t1.rk=t2.rk and t1. date =t2. date 
					group by t1. date ,t1. role id
					) t3 
	)t4
group by  Duration distribution