feat: add portfolio phase — discretize alpha weights into tradable positions

Adds a fourth pipeline phase modeling A-share microstructure: lot sizes,
the 2023-08-10 Main Board increment change, STAR 200-share minimum/odd-lot
rules, limit-up/down, suspensions, volume caps, costs, and slippage.

Two layers: research (continuous weights → return/Sharpe/turnover/Fitness,
no IC per repo convention) and execution (state-dependent lot rounding +
two-stage greedy exposure repair + next-open reference simulator).

Wires `portfolio build/simulate/eval` into the CLI and adds the
POSITION/FILL/PNL schema contracts. Covered by tests/test_portfolio.py.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Yuxuan Yan
2026-06-10 11:23:04 +08:00
parent 7faeb77c50
commit 94ab679a75
12 changed files with 1734 additions and 3 deletions
+40
View File
@@ -42,3 +42,43 @@ COMBO_COLUMNS: Final[list[str]] = [
"combo_name", # str: identifies which combo (e.g. 'equal_weight')
"weight", # float64: combined weight, signed
]
# Required columns for portfolio (position) parquet files.
# A position is a tradable integer holding derived from a target weight under
# A-share lot/board rules. Produced by the `portfolio build` phase.
POSITION_COLUMNS: Final[list[str]] = [
"symbol_id", # str
"date", # date
"portfolio_name", # str: identifies this construction run
"target_weight", # float64: w = alpha / sum(|alpha|); signed, sum(|w|)=1
"target_value", # float64: v_target = booksize * w (signed dollar exposure)
"target_shares", # float64: q_target = v_target / price (continuous, signed)
"position_shares", # int64: discretized + repaired integer shares (signed)
"position_value", # float64: position_shares * price (signed)
"price", # float64: construction price (close by default)
]
# Required columns for execution-simulator fill parquet files.
FILL_COLUMNS: Final[list[str]] = [
"symbol_id", # str
"date", # date: the EXECUTION date (open[t+1] of the target date)
"portfolio_name", # str
"prev_shares", # int64: realized position carried in
"target_shares", # int64: requested target for this execution
"traded_shares", # int64: signed delta actually executed
"realized_shares", # int64: resulting position (blocked trades revert to prev)
"blocked", # int: 1 if the trade was (fully or partially) blocked
"trade_cost", # float64: commission + slippage in yuan
]
# Required columns for execution-simulator per-day PnL parquet files.
PNL_COLUMNS: Final[list[str]] = [
"date", # date
"portfolio_name", # str
"gross_exposure", # float64: sum(|position_value|)
"net_exposure", # float64: sum(signed position_value)
"pnl", # float64: daily mark-to-market PnL (yuan), net of cost
"cost", # float64: total trade cost that day (yuan)
"turnover", # float64: sum(|traded_value|) / booksize
"n_positions", # int: count of nonzero holdings
]