v3.0.0

• start using semantic versioning
• add new training set 'trainall', containing all training data, non-overlapping with test set
• modify commandline arguments to accept 'dir=' and 'sets=' arguments
  • dir= chooses target directory
  • sets= chooses which data set(s) to generate
• bugfix for a variablename that caused a crash sometimes

v2.1

• created v2 format
• more generic format, with a header
• previous format continues to be usable, and output files are non-overlapping, can generate both, if desired

v1.3

• creates a test set now, from 100 randomly drawn games, recorded in test_samples.py
• creates a training set, train10k from 10,000 randomly drawn games, non-overlapping with the test games

(note: indexing was broken in v1.2, off by one, sorry about that. Please ignore the v1.2 version :-( )

v1.1

• remove spam
• remove the sys.exit line :-P

v1.0

v1.0