MYSQL：30天移动平均股价

Question

Objective to select the average closing price of last 30 trading days for each stock, based on input date NOT latest date

Given that;

1) If the input date is 2015-09-18, to select average closing price of last 30 TRADING days, eg (2015-09-18, 2015-09-17, 2015-09-16,...30 days)

2) If the input date is 2015-09-01, to select average closing price of last 30 TRADING days, eg (2015-09-01, 2015-09-31, 2015-09-28,...30 days)

3) Table columns: ticker | code | date | open | high | low | close | volume

4) Table contains records for only TRADING days, weekends and holidays are excluded

5) For each stock, there are thousands of days of records, but we just want the 30 Day MA at any given point in time

Input date: 2015-09-01
Desired output

date     | ticker | code |  30 day average closing price 
2015-09-01  AAA    1122     1.33
2015-09-01  BBB    3344     0.79
... 
...
...and so on for each stock

Answer 1

EDIT This version should be flexible and I hope not too slow. Since no idea what means what in your table and how the table is called, I assume a table pricetable with code='AAA','BBB' and so and close meaning the price. Your given date should be '2015-09-12' here. Replace '2015-09-12' whith whatever you like.

30 should be the number of trading days. Replace by whatever you want in 'limit 30'.

ticker | code | date | open | high | low | close | volume

TRY THIS VERSION

select pt.code, sum(pt.close) / count(*) as average
from pricetable pt 
inner join 
  ( select distinct `date`
    from pricetable 
    where `date` <= '2015-09-12' # your given date
    order by `date` desc 
    limit 30 # the number of trading days
  ) as  d on d.`date` = pt.`date`
where code in ("AAA","BBB","CCC")
group by code
order by average desc;

Use some of the latter versions if the subselect get too slow.

You definitely should make a table with calendar days, and they get a column to mark trading days. It would be much easier if they also contain a reference to the 30-day-before and the 90-day-before.

 create table calendar_days 
 (
    id int(11) not null auto_increment primary key, 
    calday date not null, 
    is_trading tinyint, 
    trading_before_30 date, 
    trading_before_90 date
 );

Then it gets easy:

select pt.stockname, sum(pt.closing_price) / count(*) as avg_close
from calendar_days d
inner join pricetable pt on pt.day between d.calday and d.trading_before_30
where d.calday = '2015-08-15'
group by pt.stockname;

You can fill that calendar day table quite easy from your stock table.

EDIT: this way you can fill it at first requests of day, assuming that only trading days are in this table.

insert into calendar_days (calday, is_trading) 
select distinct day, 1
from pricetable pt 
where pt.day > (select max(calday from calendar_days);

update calendar_days d1 
inner join calendar_days d2 on d2.nr = d1.nr - 29
set d1.trading_before_30 = d2.calday 
where d1.trading_before_30 is null;

You could of course try to dynamically construct a list of the last 30 trading days out of your stockprice table, but that table has many many entries per day, so that would be a unneccessary slow down of your query - the day changes very seldom, just once every day.

You could even add a trigger to your closing day table so that on every insert it will check if calendar days are to be refilled.

A version without that table would look like:

drop temporary table if exists tmp_days;
create temporary table tmp_days (tradingday date not null primary key);
insert into tmp_days 
select distinct `day` 
from pricetable 
where day <= '2015-08-15'
order by day desc limit 30;

and then

select pt.stockname, sum(pt.price) / count(*) as average
from pricetable pt 
inner join tmp_days d on d.tradingday = pt.day;

The temp-table will be only visible in one connection.

Answer 2

As a starting point see https://dev.mysql.com/doc/refman/5.0/en/date-and-time-functions.html#function_date-add these functions. Should be like (not tested):

SELECT `date`, ticker, avg(`close`) from TRADES where
date <= '2015-09-01' and date >= DATE_SUB('2015-09-01', INTERVAL 30 DAY)
GROUP BY ticker