Issue, briefly
Currently, dropped non-first games do not appear in FeatureExtractor.exclusions (which is later dumped as exclusions.csv). This should change.
Further, an exclusion due to duplicate ID should generate a warning.
More details
People's first game of the CFG is different than subsequent runs. This is why FeatureExtractor keeps only the earliest game for each player ID.
We want non-first games to be excluded by design, without asking the user. That said, a user running the pipeline might not be aware of duplicate IDs in the data.
This issue is about the pipeline being more "loud" about such exclusions, to make users aware of them.
Issue, briefly
Currently, dropped non-first games do not appear in
FeatureExtractor.exclusions(which is later dumped asexclusions.csv). This should change.Further, an exclusion due to duplicate ID should generate a warning.
More details
People's first game of the CFG is different than subsequent runs. This is why
FeatureExtractorkeeps only the earliest game for each player ID.We want non-first games to be excluded by design, without asking the user. That said, a user running the pipeline might not be aware of duplicate IDs in the data.
This issue is about the pipeline being more "loud" about such exclusions, to make users aware of them.