This question is in relation to 'Anomalies'. Anomalies have 3 explanations:

1. Risk premium
2. Behavioural Biases
3. Data mining

I am quite unclear about 1. and 3.

For 1. (risk premiums), would that be the additional risk factors added to models that increase the explanatory factor (R-squared)? That is, if a risk factor increases the r-squared, it stands as a valid risk premium, and that an anomaly would be better addressed by risk premiums? For example, from Fama & French's 3 factor model to Carhart's 4 factor model which includes momentum anomaly, the 4 factor model had higher r-squared. Does this mean that the momentum anomaly is of a risk premium?