10 period moving average in MySql without using date

1
votes

I have a table of goalie data, snipet below

    year    gameid  player  sv% gamenum
2009    200165  John Smith  0.923   0165
2009    209754  John Smith  1.000   9754
2009    206938  John Smith  1.000   6938
2009    206155  John Smith  0.833   6155
2009    203021  John Smith  0.667   3021
2009    206472  John Smith  0.909   6472
2009    209524  John Smith  0.833   9524
2009    209351  John Smith  0.800   9351
2009    203056  John Smith  1.000   3056
2009    206761  John Smith  0.935   6761
2009    200466  John Smith  0.954   0466
2009    204171  John Smith  0.932   4171
2009    207876  John Smith  0.958   7876
2009    201581  John Smith  0.941   1581
2009    205286  John Smith  0.930   5286
2009    208991  John Smith  0.961   8991
2009    202696  John Smith  0.916   2696
2009    206401  John Smith  0.935   6401
2009    200106  John Smith  0.921   0106
2009    201381  John Smith  0.918   1381

I want to get the 10 game moving averages for each goalie, but I don't have dates or game numbers such as his first, second, third game, etc. The game ids are also assigned in the order they are played at the league level, so game 200106 could be his first game of season, and 200165 could be his 2nd, and so on.

My question is: How can I get the max(10 game moving average) and min(10 game moving average) grouped by each goalie for each year?

Also, is there a way to rank the game ids by goalie, year using MySql?

2
mysqlrankingmoving-average

2 Answers

1
votes

A 10 game moving average means that if you had less than 10 games, there is no meaningful average (not enough games). If you had 12 games, the average is taken between

1-10 (avg)
2-11 (avg)
3-12 (avg)
max / min across the 3 averages

The most efficient way to do this in MySQL would be to

select .. (involving 13 @variables to rownumber and rotate the last
           10 values into the variables, keeping track of
           @player, @year, @rownumber)
order by player, year, gameid

This will pass through the data only once, building the averages. An outer query will simply take min/max from this derived table. I'm not up for fleshing this out at the moment though.

0
votes

This is one idea (fair warning:not tested)

SELECT max(mavg) FROM 
(SELECT (SELECT avg(avgfield),min(gamenum) as gn FROM YourTable g WHERE g.gamenum>t.gamenum LIMIT 10),t.gamenum 
       FROM
       YourTable t
) d

or

SELECT max(mavg) FROM 
(SELECT t.gamenum FROM
       YourTable t INNER JOIN 
       (SELECT avg(avgfield),min(gamenum) as gn FROM YourTable g WHERE g.gamenum>t.gamenum LIMIT 10) q  ON q.gn = t.gamenum
) d