v0.0.8

• open source
• IR increase to 2.143160 with TopkDropoutDynamicStrategy

v0.0.7

Merge pull request #5 from ChenglongChen/dev

move policy

v0.0.6

changes

• add more days_ahead and hold_thresh

performance

• with TOPK_LIST and N_DROP_LIST
  • TOPK_LIST = [1, 2, 4, 6, 8, 10]
  • N_DROP_LIST = [1, 2, 3, 4, 5]
• performance with cost when days_ahead=7, hold_thresh=2, topk=10 and n_drop=5
  • mean = 0.00100018148258493
  • std = 0.009446687200984873
  • annualized_return = 0.23804319285521336
  • information_ratio = 1.6333819538230134
  • max_drawdown = -0.1381429410306242

v0.0.5

changes

• adjust hold_threshold=2

performance

• with TOPK_LIST and N_DROP_LIST
  • TOPK_LIST = [1, 2, 4, 6, 8, 10, 15, 20, 25, 30, 35, 40, 45, 50]
  • N_DROP_LIST = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25]
• performance with cost when topk=30 and n_drop=9
  • mean = 0.00080406949555358
  • std = 0.006220215022820138
  • annualized_return = 0.19136853994175204
  • information_ratio = 1.9942365288964645
  • max_drawdown = -0.10651395704491373

v0.0.4

changes

• remove diff features with ChangeInstrument and DropInstrument

performance

• with TOPK_LIST and N_DROP_LIST
  • TOPK_LIST = [1, 2, 4, 6, 8, 10]
  • N_DROP_LIST = [1, 2, 3, 4, 5]
• performance with cost when topk=10 and n_drop=4
  • mean = 0.0008124964587127172
  • std = 0.009565672157545894
  • annualized_return = 0.19337415717362672
  • information_ratio = 1.3103715729984229
  • max_drawdown = -0.13605229867254498

v0.0.3

changes

• add diff features with ChangeInstrument and DropInstrument
• add days_ahead for label

performance

• performance with cost
  • annualized_return = 0.18234649262028707
  • information_ratio = 1.214480002412785
  • max_drawdown = -0.13976082839358217

v0.0.2

changes

• add ops for dataframe to support cross section ops such as rank, corr, cov etc
• add GTJA Alpha: https://guorn.com/static/upload/file/3/134065454575605.pdf
• fix turnover issues in TopkDropoutStrategy: microsoft/qlib#809

v0.0.1

changes

• add neutralize

performance

• performance with cost when
  • annualized_return = 0.214063
  • information_ratio = 1.486732
  • max_drawdown = -0.139090