rolling calcs checks

[dpopadic]

Rolling functions double checks.

looking at Japanese league in separation we get the following factor data for goal superiority factor with k=3 (lagged):

However, when performing calculations across leagues we get the following:

[dpopadic]

Previous season games for the team kobe are taken into account whereas the 1st 3 get neutralised for new entrants into the league:

Hence, the calculations DO NOT seem correct. Note that this is not lagged here - so when comparing you need to look 1 date further above. For example, for the following dates it should look like:
2019-08-23: +3
2019-08-31: +8

[dpopadic]

The issue is when neutralising the scores for new entrants via neutralise_scores. The function performs its tasks as required. However, there seems to be an issue with the data.

kobe is in 2 different leagues (Japan J1 League, Japan J-League) in the same season (2019).

[dpopadic]

Leaving it for now since it's doing what it is supposed to do. Concluding that modelling and factor construction by league is preferred at this stage due to this and other issues.