# Applying cumulative mean function to a grouped object

## Question:

I have a DataFrame `df` where each record represents a soccer game. Teams will appear more than once. I need to compute some sort of a rolling mean for each team scores(well, not exactly the rolling mean to the letter).

``````     date           home           away       score_h  score_a
166  2013-09-01     Fulham         Chelsea       0      0
167  2013-09-03     Arsenal        Everton       0      2
164  2013-09-05     Arsenal        Swansea       5      1
165  2013-09-06     Fulham         Norwich       0      1
163  2013-09-18     Arsenal        Swansea       0      0
``````

What I need to calculate, is the mean score for each team (home and away).

For brevity, let’s just do the home column:

``````grouped = df.groupby('home')
grouped = grouped.sort_index(by='date') # rows inside groups must be in asc order
``````

This results in:

``````    date    home    away    score_h     score_a
home
Arsenal     167     2013-09-03  Arsenal     Everton     0   2
164     2013-09-05  Arsenal     Swansea     5   1
163     2013-09-18  Arsenal     Swansea     0   0
Fulham      166     2013-09-01  Fulham      Chelsea     0   0
165     2013-09-06  Fulham      Norwich     0   1
``````

Question starts here

Now, I need to compute “rolling mean” for teams. Let’s do it by hand for the group named `Arsenal`. At the end of this we should wind up with 2 extra columns, let’s call them: `rmean_h` and `rmean_a`. First record in the group (`167`) has scores of `0` and `2`. The `rmean` of these is simply `0` and `2` respectively. For second record in the group (`164`), the rmeans will be `(0+5)/2 = 2.5` and `(2+1) / 2 = 1.5`, and for the third record, `(0+5+0)/3 = 1.66` and `(2+1+0)/3 = 1`.

Our DataFrame should now looks like this:

``````                    date       home         away    score_h score_a rmean_h rmean_a
home
Arsenal     167     2013-09-03  Arsenal     Everton     0  2    0       2
164     2013-09-05  Arsenal     Swansea     5  1    2.5     1.5
163     2013-09-18  Arsenal     Swansea     0  0    1.66    1
Fulham      166     2013-09-01  Fulham      Chelsea     0  0
165     2013-09-06  Fulham      Norwich     0  1
``````

I want to carry out these calculations for my data, do you have any suggestions please?

You can apply an `expanding_mean` (see docs) to each group:

``````grouped = df.sort(columns='date').groupby('home')
grouped['score_h'].apply(pd.expanding_mean)
``````

For pandas 1.4.3 you can use:

``````grouped = df.sort(columns='date').groupby('home')
grouped['score_h'].expanding().mean()
``````

Read more on the docs: expanding and expanding window.

Categories: questions Tags: , ,
Answers are sorted by their score. The answer accepted by the question owner as the best is marked with
at the top-right corner.