​
Overview

The AVG() window function calculates the average (arithmetic mean) of a set of numeric values within a window. This function allows you to compute averages over a set of rows that are related to the current row, such as rows within a partition of ordered set.

​
Syntax

The syntax for this function is as follows:

AVG(expression) OVER (
  [PARTITION BY partition_expression]
  ORDER BY sort_expression
  [ROWS | RANGE frame_specification]
)

​
Parameters

  • expression: column or expression that the function operates on (must be of numeric type)
  • ROWS or RANGE: (optional) frame specification to control which rows are included in the calculation relative to the current row

​
Example

For the needs of this section, we will use a simplified version of the film table from the Pagila database, containing only the title, length and rating columns. The complete schema for the film table can be found on the Pagila database website.

DROP TABLE IF EXISTS film;
CREATE TABLE film (
  title text NOT NULL,
  length int,
  rating int
);
INSERT INTO film(title, length, rating) VALUES
  ('ATTRACTION NEWTON', 83, 5),
  ('CHRISTMAS MOONSHINE', 150, 7),
  ('DANGEROUS UPTOWN', 121, 4),
  ('KILL BROTHERHOOD', 54, 3),
  ('HALLOWEEN NUTS', 47, 5),
  ('HOURS RAGE', 122, 7),
  ('PIANIST OUTFIELD', 136, 7),
  ('PICKUP DRIVING', 77, 3),
  ('INDEPENDENCE HOTEL', 157, 7),
  ('PRIVATE DROP', 106, 4),
  ('SAINTS BRIDE', 125, 3),
  ('FOREVER CANDIDATE', 131, 7),
  ('MILLION ACE', 142, 5),
  ('SLEEPY JAPANESE', 137, 4),
  ('WRATH MILE', 176, 7),
  ('YOUTH KICK', 179, 7),
  ('CLOCKWORK PARADISE', 143, 5);

​
Rolling Average by Rating

The query below uses the AVG() function to calculate the rolling average of length as rows are ordered by rating:

SELECT
    rating,
    length,
    AVG(length) OVER (ORDER BY rating) AS RollingAverageLength
FROM film
WHERE length IS NOT NULL
ORDER BY rating;

By executing the query above, we will get the following output:

 rating | length | rollingaveragelength 
--------+--------+----------------------
      3 |     77 |    85.33333333333333
      3 |    125 |    85.33333333333333
      3 |     54 |    85.33333333333333
      4 |    121 |   103.33333333333333
      4 |    106 |   103.33333333333333
      4 |    137 |   103.33333333333333
      5 |     83 |                103.5
      5 |    142 |                103.5
      5 |     47 |                103.5
      5 |    143 |                103.5
      7 |    157 |   122.70588235294117
      7 |    179 |   122.70588235294117
      7 |    176 |   122.70588235294117
      7 |    131 |   122.70588235294117
      7 |    136 |   122.70588235294117
      7 |    122 |   122.70588235294117
      7 |    150 |   122.70588235294117
(17 rows)

​
Time Series: Rolling Average Length over Last 3 Ratings

In this example, we will demonstrate a time series-style rolling average using a window frame of the current row and the two preceding rows, ordered by rating. This simulates a moving average over a sliding window of 3 rows:

SELECT
    rating,
    length,
    AVG(length) OVER (
      ORDER BY rating
      ROWS BETWEEN 2 PRECEDING AND CURRENT ROW
    ) AS rolling_avg_length_3
FROM film
WHERE length IS NOT NULL
ORDER BY rating;

The query above calculates the average length over the current rating and the two previous ratings (based on ordering by rating) smoothing the fluctuations by averaging over a fixed-size window:

 rating | length | rolling_avg_length_3 
--------+--------+----------------------
      3 |     77 |                 65.5
      3 |    125 |    85.33333333333333
      3 |     54 |                   54
      4 |    121 |   107.66666666666667
      4 |    106 |   117.33333333333333
      4 |    137 |   121.33333333333333
      5 |     83 |                   91
      5 |    142 |    90.66666666666667
      5 |     47 |                  109
      5 |    143 |   128.66666666666666
      7 |    157 |   127.33333333333333
      7 |    179 |   159.33333333333334
      7 |    176 |   170.66666666666666
      7 |    131 |                  162
      7 |    136 |   147.66666666666666
      7 |    122 |   129.66666666666666
      7 |    150 |                  136
(17 rows)