Hello,

Thank you very much for the kind words.

Regarding the target leakage, the fact that we get rid of it comes mostly from the fact that we do not include the current observation (the one we are trying to predict the target for) when we compute its encoded statistic. This is the ordering principle in which we compute the encoding statistic only using past data.

Regarding your second question, I am not sure to understand it. The encoded values will be different for every observation thanks to the ordering principle.