Am I understanding this correctly that this calculation means the most-recent games are more heavily weighted in the resulting matrices? Team A can have an average offensive rating of 90 vs Team B over 6 games. If Team A then has an offensive rating of 110 against Team B in their 7th meeting, the matrix will now have the value (90+110) / 2 = 100 even though their actual season average vs Team B is ~93.
Is this intentional (or am I missing something)? I'd assume a true average would be more preferable than the above case. I'm not sure if there's a simple way to track the number of updates to each value to achieve this.
Regarding the following line in update_df:
new_value = (float(old_value) + float(value)) / 2
Am I understanding this correctly that this calculation means the most-recent games are more heavily weighted in the resulting matrices? Team A can have an average offensive rating of 90 vs Team B over 6 games. If Team A then has an offensive rating of 110 against Team B in their 7th meeting, the matrix will now have the value (90+110) / 2 = 100 even though their actual season average vs Team B is ~93.
Is this intentional (or am I missing something)? I'd assume a true average would be more preferable than the above case. I'm not sure if there's a simple way to track the number of updates to each value to achieve this.