mirror of
https://ghfast.top/https://github.com/aeroxw/tick-stock-panel.git
synced 2026-09-12 17:54:15 +08:00
shy3130
21260f1053
- 盘中增量路径补算偏离列: compute_enriched_today 补 momentum_3d, live_agg 新增 _close_3d_ago 递推状态, 增量/全量回退两路径统一附着 今日偏离 (基准 = 历史帧昨收 × (1+指数实时涨跌) 外推, 排除盘中 写入的今日指数行), 修复盘中异动列表恒为空的问题 - 主板 ST 口径统一: 2026-07-06 起风险警示股票涨跌幅 10% 且异常波动 特别规定废止, 删除原 ±15%/10日+50%/30日+100% 从严表, ST 与普通 主板同标准 - 严重异常波动负向阈值对齐官方不对称口径: 10日 +100%(-50%), 30日 +200%(-70%), 跌方向更早触发; 前端规则表/窗口徽标同步双侧显示
437 lines
17 KiB
Python
437 lines
17 KiB
Python
"""异动边缘统计测试 — 偏离列附着 + 规则口径 + 快照接近度。"""
|
|
from __future__ import annotations
|
|
|
|
from datetime import date, timedelta
|
|
|
|
import polars as pl
|
|
|
|
from app.indicators.pipeline import (
|
|
attach_deviation_columns,
|
|
attach_deviation_columns_today,
|
|
benchmark_momentum_today,
|
|
load_benchmark_momentum,
|
|
)
|
|
from app.services.abnormal_moves import (
|
|
_hist_cache,
|
|
_hist_cache_lock,
|
|
board_of,
|
|
build_overview,
|
|
is_st_name,
|
|
rule_for,
|
|
)
|
|
|
|
|
|
def _write_index_daily(tmp_path, rows: list[tuple[str, date, float]]) -> None:
|
|
df = pl.DataFrame(
|
|
{
|
|
"symbol": [r[0] for r in rows],
|
|
"date": [r[1] for r in rows],
|
|
"close": [r[2] for r in rows],
|
|
}
|
|
)
|
|
for dt in sorted({r[1] for r in rows}):
|
|
target = tmp_path / "kline_index_daily" / f"date={dt.isoformat()}"
|
|
target.mkdir(parents=True, exist_ok=True)
|
|
df.filter(pl.col("date") == dt).write_parquet(target / "part.parquet")
|
|
|
|
|
|
def test_attach_deviation_columns_math(tmp_path) -> None:
|
|
# 上证指数 4 天等差 +1: 3日动量 = 13/10-1 = 0.30
|
|
# 个股 close 与指数同序列 → momentum_3d 缺失时按 close 就地补算, 偏离 = 0
|
|
days = [date(2026, 8, 13), date(2026, 8, 14), date(2026, 8, 15), date(2026, 8, 18)]
|
|
index_rows = [("000001.SH", d, 10.0 + i) for i, d in enumerate(days)]
|
|
_write_index_daily(tmp_path, index_rows)
|
|
|
|
stock = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH"] * len(days),
|
|
"date": days,
|
|
"close": [10.0 + i for i in range(len(days))],
|
|
# 10/30 日窗口已有动量列 → 直接使用
|
|
"momentum_10d": [None] * 4,
|
|
"momentum_30d": [None] * 4,
|
|
}
|
|
)
|
|
out = attach_deviation_columns(stock, tmp_path)
|
|
assert "deviate_3d" in out.columns
|
|
assert "momentum_3d" in out.columns # 就地补算
|
|
last = out.sort("date").row(-1, named=True)
|
|
assert abs(last["deviate_3d"] - 0.0) < 1e-9
|
|
|
|
|
|
def test_attach_deviation_columns_missing_benchmark(tmp_path) -> None:
|
|
# 无指数数据: 偏离列为 null, 不抛异常
|
|
stock = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH"],
|
|
"date": [date(2026, 8, 18)],
|
|
"momentum_3d": [0.2],
|
|
"momentum_10d": [0.5],
|
|
"momentum_30d": [1.0],
|
|
}
|
|
)
|
|
out = attach_deviation_columns(stock, tmp_path)
|
|
assert out["deviate_3d"][0] is None
|
|
|
|
|
|
# ── 盘中路径: 今日基准动量外推 + 单日帧偏离附着 ──────────────────
|
|
|
|
_BENCH_DAYS = [date(2026, 8, 11), date(2026, 8, 12), date(2026, 8, 13),
|
|
date(2026, 8, 14), date(2026, 8, 15), date(2026, 8, 18)]
|
|
|
|
|
|
def _write_sh_bench(tmp_path) -> None:
|
|
# 上证指数 6 日收盘 10..15, 末值 15 为昨收
|
|
_write_index_daily(tmp_path, [("000001.SH", d, 10.0 + i) for i, d in enumerate(_BENCH_DAYS)])
|
|
|
|
|
|
def test_benchmark_momentum_today_math(tmp_path) -> None:
|
|
_write_sh_bench(tmp_path)
|
|
quotes = pl.DataFrame({"symbol": ["000001.SH"], "change_pct": [0.10]})
|
|
|
|
out = benchmark_momentum_today(tmp_path, quotes)
|
|
row = out.row(0, named=True)
|
|
# 今收 = 15 x 1.10 = 16.5; 3 个交易日前的收盘 = 13 (与全量路径 shift(3) 同口径)
|
|
# mom3d = 16.5/13 - 1
|
|
assert abs(row["bench_mom3d"] - (16.5 / 13 - 1)) < 1e-9
|
|
# 10/30 日窗口收盘数不足 → null
|
|
assert row["bench_mom10d"] is None
|
|
assert row["bench_mom30d"] is None
|
|
|
|
# 无实时行情 → rt 按 0 处理: mom3d = 15/13 - 1
|
|
out0 = benchmark_momentum_today(tmp_path, None)
|
|
assert abs(out0.row(0, named=True)["bench_mom3d"] - (15.0 / 13 - 1)) < 1e-9
|
|
|
|
|
|
def test_benchmark_momentum_today_excludes_today_rows(tmp_path) -> None:
|
|
# 指数监控盘写入的今日行不能当昨收 (否则实时涨跌被重复叠加)
|
|
today = date.today()
|
|
rows = [("000001.SH", d, 10.0 + i) for i, d in enumerate(_BENCH_DAYS)]
|
|
rows.append(("000001.SH", today, 99.0)) # 今日脏行
|
|
_write_index_daily(tmp_path, rows)
|
|
|
|
out = benchmark_momentum_today(tmp_path, None)
|
|
assert abs(out.row(0, named=True)["bench_mom3d"] - (15.0 / 13 - 1)) < 1e-9
|
|
|
|
|
|
def test_attach_deviation_columns_today(tmp_path) -> None:
|
|
_write_sh_bench(tmp_path)
|
|
quotes = pl.DataFrame({"symbol": ["000001.SH"], "change_pct": [0.10]})
|
|
# 单日帧: 增量路径产出的 momentum 列 (无 date 历史, 无法 shift 补算)
|
|
today_df = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH", "000001.SZ"],
|
|
"momentum_3d": [0.5, 0.2],
|
|
"momentum_10d": [0.2, None],
|
|
"momentum_30d": [1.0, None],
|
|
}
|
|
)
|
|
out = attach_deviation_columns_today(today_df, tmp_path, quotes)
|
|
# SH: 0.5 - (16.5/13 - 1)
|
|
assert abs(out["deviate_3d"][0] - (0.5 - (16.5 / 13 - 1))) < 1e-9
|
|
# SZ 无深证基准 → 按选基设计回退上证基准 (rt=0): 0.2 - (15/13 - 1)
|
|
assert abs(out["deviate_3d"][1] - (0.2 - (15.0 / 13 - 1))) < 1e-9
|
|
assert "bench_close" not in out.columns
|
|
|
|
|
|
def test_attach_deviation_columns_today_missing_momentum(tmp_path) -> None:
|
|
# 全量回退路径可能缺 momentum_3d: 该窗口置 null, 其余窗口正常
|
|
days = [date(2026, 7, 1) + timedelta(days=i) for i in range(35)]
|
|
_write_index_daily(tmp_path, [("000001.SH", d, 10.0 + i) for i, d in enumerate(days)])
|
|
df = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH"],
|
|
"momentum_10d": [0.2],
|
|
"momentum_30d": [1.0],
|
|
}
|
|
)
|
|
out = attach_deviation_columns_today(df, tmp_path, None)
|
|
assert out["deviate_3d"][0] is None
|
|
assert out["deviate_10d"][0] is not None
|
|
assert out["deviate_30d"][0] is not None
|
|
|
|
|
|
def test_attach_deviation_columns_no_bench_close_leak(tmp_path) -> None:
|
|
# load_benchmark_momentum 新增 bench_close 列后, 冷路径输出不应泄漏该列
|
|
_write_sh_bench(tmp_path)
|
|
stock = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH"],
|
|
"date": [date(2026, 8, 18)],
|
|
"close": [15.0],
|
|
}
|
|
)
|
|
out = attach_deviation_columns(stock, tmp_path)
|
|
assert "bench_close" not in out.columns
|
|
frame = load_benchmark_momentum(tmp_path)
|
|
assert "bench_close" in frame.columns
|
|
|
|
|
|
def test_board_and_st_rules() -> None:
|
|
assert board_of("600000.SH") == "主板"
|
|
assert board_of("000001.SZ") == "主板"
|
|
assert board_of("301123.SZ") == "创业板"
|
|
assert board_of("688123.SH") == "科创板"
|
|
assert board_of("920001.BJ") == "北交所"
|
|
assert is_st_name("*ST 某某") is True
|
|
assert is_st_name("正常股") is False
|
|
|
|
main = rule_for("600000.SH", "正常股")
|
|
# 3日对称 ±20%; 严重异动负向更严: 10日+100%(-50%), 30日+200%(-70%)
|
|
assert main.thresholds == {3: (0.20, 0.20), 10: (1.00, 0.50), 30: (2.00, 0.70)}
|
|
# 2026-07-06 起主板风险警示股票与普通股票同标准 (原±15%特别规定已废止)
|
|
st = rule_for("600000.SH", "ST 某某")
|
|
assert st.thresholds == main.thresholds
|
|
assert st.st is True
|
|
gem = rule_for("301123.SZ", "正常股")
|
|
assert gem.thresholds[3] == (0.30, 0.30)
|
|
assert gem.thresholds[10] == (1.00, 0.50)
|
|
bse = rule_for("920001.BJ", "正常股")
|
|
assert bse.thresholds[3] == (0.40, 0.40)
|
|
|
|
|
|
class _FakeRepo:
|
|
"""最小 repo: get_enriched_latest 返回构造帧。"""
|
|
|
|
def __init__(self, df: pl.DataFrame) -> None:
|
|
self._df = df
|
|
|
|
def get_enriched_latest(self):
|
|
return self._df, date(2026, 8, 19)
|
|
|
|
|
|
class _FakeQuotes:
|
|
def get_index_quotes(self):
|
|
return pl.DataFrame(
|
|
{"symbol": ["000001.SH"], "close": [3300.0], "prev_close": [3270.0]}
|
|
)
|
|
|
|
|
|
def test_build_overview_closeness_and_status() -> None:
|
|
with _hist_cache_lock:
|
|
_hist_cache.clear()
|
|
df = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH", "300001.SZ", "000002.SZ"],
|
|
"name": ["股A", "股B", "股C"],
|
|
"close": [10.0, 20.0, 30.0],
|
|
"change_pct": [0.05, 0.02, 0.01],
|
|
"deviate_3d": [0.19, 0.35, 0.05],
|
|
"deviate_10d": [0.99, 0.40, 0.20],
|
|
"deviate_30d": [1.95, 2.10, 0.60],
|
|
}
|
|
)
|
|
result = build_overview(_FakeRepo(df), _FakeQuotes(), min_closeness=0.5, limit=10)
|
|
|
|
by_symbol = {r["symbol"]: r for r in result["rows"]}
|
|
# 主板: 3d阈值0.2 → 0.19/0.2=0.95 边缘; 指数实时 +30/3270≈0.00917 叠加后略增
|
|
a = by_symbol["600000.SH"]
|
|
assert a["status"] in ("edge", "triggered")
|
|
# 创业板: 30日 2.10/2.00 ≥ 1 → triggered
|
|
b = by_symbol["300001.SZ"]
|
|
assert b["status"] == "triggered"
|
|
# 000002: 3d 0.05/0.2=0.25, 10d 0.2/1=0.2, 30d 0.6/2=0.3 → 全部 < 0.5 被过滤
|
|
assert "000002.SZ" not in by_symbol
|
|
# 排序按接近度降序
|
|
closeness = [r["max_closeness"] for r in result["rows"]]
|
|
assert closeness == sorted(closeness, reverse=True)
|
|
assert result["counts"]["triggered"] >= 1
|
|
|
|
|
|
def test_build_overview_cache_date_today_no_double_count() -> None:
|
|
"""cache_date >= 今天时不再叠加实时涨跌 (避免重复计入)。"""
|
|
with _hist_cache_lock:
|
|
_hist_cache.clear()
|
|
|
|
class _TodayRepo(_FakeRepo):
|
|
def get_enriched_latest(self):
|
|
return self._df, date.today()
|
|
|
|
df = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH"],
|
|
"name": ["股A"],
|
|
"close": [10.0],
|
|
"change_pct": [0.05],
|
|
"deviate_3d": [0.19],
|
|
"deviate_10d": [None],
|
|
"deviate_30d": [None],
|
|
}
|
|
)
|
|
result = build_overview(_TodayRepo(df), _FakeQuotes(), min_closeness=0.5)
|
|
row = result["rows"][0]
|
|
assert abs(row["windows"]["3d"]["value"] - 0.19) < 1e-9
|
|
|
|
|
|
def test_build_overview_negative_side_stricter_threshold() -> None:
|
|
"""严重异动负向阈值更严 (10日-50%/30日-70%), 跌方向更早触发。"""
|
|
with _hist_cache_lock:
|
|
_hist_cache.clear()
|
|
|
|
class _TodayRepo(_FakeRepo):
|
|
def get_enriched_latest(self):
|
|
return self._df, date.today()
|
|
|
|
df = pl.DataFrame(
|
|
{
|
|
"symbol": ["600000.SH", "600001.SH"],
|
|
"name": ["跌一", "跌二"],
|
|
"close": [10.0, 20.0],
|
|
"change_pct": [-0.05, -0.05],
|
|
# -0.55: 旧对称口径 0.55/1.00=0.55 (观察); 新口径 0.55/0.50=1.1 (触发)
|
|
# -0.75: 30日 0.75/0.70≈1.07 (触发)
|
|
"deviate_3d": [None, None],
|
|
"deviate_10d": [-0.55, None],
|
|
"deviate_30d": [None, -0.75],
|
|
}
|
|
)
|
|
result = build_overview(_TodayRepo(df), None, min_closeness=0.5)
|
|
by_symbol = {r["symbol"]: r for r in result["rows"]}
|
|
a = by_symbol["600000.SH"]
|
|
assert a["windows"]["10d"]["threshold"] == 0.50
|
|
assert abs(a["windows"]["10d"]["closeness"] - 1.1) < 1e-9
|
|
assert a["status"] == "triggered"
|
|
b = by_symbol["600001.SH"]
|
|
assert b["windows"]["30d"]["threshold"] == 0.70
|
|
assert abs(b["windows"]["30d"]["closeness"] - round(0.75 / 0.7, 4)) < 1e-9
|
|
assert b["status"] == "triggered"
|
|
# 正向阈值不变: +100%/+200% (在正偏离用例中覆盖, 这里验证规则表)
|
|
main = rule_for("600000.SH", "正常股")
|
|
assert main.thresholds[10] == (1.00, 0.50)
|
|
assert main.thresholds[30] == (2.00, 0.70)
|
|
|
|
|
|
# ── 监控规则接入 (type=abnormal) ────────────────────────
|
|
|
|
import pytest
|
|
|
|
from app.strategy import monitor_rules
|
|
from app.strategy.monitor import MonitorRuleEngine
|
|
|
|
|
|
def _ab_rule(**overrides) -> dict:
|
|
rule = {
|
|
"id": "r_ab",
|
|
"name": "异动边缘",
|
|
"type": "abnormal",
|
|
"scope": "all",
|
|
"symbols": [],
|
|
"threshold_pct": 70,
|
|
"direction": "both",
|
|
"abnormal_window": "any",
|
|
"cooldown_seconds": 0,
|
|
"severity": "warn",
|
|
}
|
|
rule.update(overrides)
|
|
return rule
|
|
|
|
|
|
def _row(symbol: str, *wins: tuple[str, float], name: str = "股A",
|
|
board: str = "主板", rt_pct: float = 0.05) -> dict:
|
|
# wins: (窗口, 偏离值) — 阈值按交易所口径: 主板 3d=0.2, 10d=1.0, 30d=2.0
|
|
thresholds = {"3d": 0.2, "10d": 1.0, "30d": 2.0}
|
|
windows = {
|
|
key: {"value": value, "threshold": thresholds[key],
|
|
"closeness": round(abs(value) / thresholds[key], 4)}
|
|
for key, value in wins
|
|
}
|
|
return {"symbol": symbol, "name": name, "board": board, "st": False,
|
|
"close": 10.0, "rt_pct": rt_pct, "windows": windows}
|
|
|
|
|
|
def test_abnormal_rule_validation_and_defaults() -> None:
|
|
rule = monitor_rules.normalize({"id": "r1", "name": "n", "type": "abnormal"})
|
|
assert rule["direction"] == "both"
|
|
assert rule["threshold_pct"] == 70.0
|
|
assert rule["abnormal_window"] == "any"
|
|
monitor_rules.validate(rule)
|
|
|
|
monitor_rules.validate(_ab_rule(threshold_pct=100, direction="up", abnormal_window="3d"))
|
|
|
|
with pytest.raises(ValueError):
|
|
monitor_rules.validate(_ab_rule(abnormal_window="5d"))
|
|
with pytest.raises(ValueError):
|
|
monitor_rules.validate(_ab_rule(threshold_pct=0.5))
|
|
with pytest.raises(ValueError):
|
|
monitor_rules.validate(_ab_rule(asset_type="etf"))
|
|
with pytest.raises(ValueError):
|
|
monitor_rules.validate(_ab_rule(direction="entry"))
|
|
|
|
|
|
def test_engine_abnormal_edge_trigger_and_cooldown() -> None:
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule()])
|
|
assert engine.min_abnormal_closeness() == pytest.approx(0.7)
|
|
|
|
# 首轮观测不触发 (防新建规则刷屏); 0.10/0.2 = 50% 接近度, 低于阈值
|
|
assert engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.10))], now=1000.0) == []
|
|
# 上穿 70% → 触发 (0.16/0.2 = 80%)
|
|
events = engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.16))], now=1006.0)
|
|
assert len(events) == 1
|
|
ev = events[0]
|
|
assert ev["source"] == "abnormal"
|
|
assert ev["type"] == "abnormal_up"
|
|
assert ev["symbol"] == "600000.SH"
|
|
assert ev["abnormal_window"] == "3d"
|
|
assert ev["abnormal_closeness"] == pytest.approx(0.8)
|
|
assert "接近" in ev["message"] or "已达" in ev["message"]
|
|
# 持续高于阈值: 不重复触发 (边缘语义)
|
|
assert engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.18))], now=1012.0) == []
|
|
# 回落再上穿: cooldown=0 时再次触发
|
|
engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.10))], now=1018.0)
|
|
assert len(engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.17))], now=1024.0)) == 1
|
|
|
|
# cooldown 内的上穿被抑制
|
|
engine_cd = MonitorRuleEngine()
|
|
engine_cd.set_rules([_ab_rule(cooldown_seconds=3600)])
|
|
engine_cd.evaluate_abnormal([_row("600000.SH", ("3d", 0.10))], now=1000.0)
|
|
engine_cd.evaluate_abnormal([_row("600000.SH", ("3d", 0.16))], now=1006.0)
|
|
engine_cd.evaluate_abnormal([_row("600000.SH", ("3d", 0.10))], now=1012.0)
|
|
assert engine_cd.evaluate_abnormal([_row("600000.SH", ("3d", 0.16))], now=1018.0) == []
|
|
|
|
|
|
def test_engine_abnormal_stale_symbol_state_cleared() -> None:
|
|
"""标的跌出快照后状态应清回 False, 回升穿过阈值时可再次触发。"""
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule()])
|
|
engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.10))], now=1000.0) # 首轮 False
|
|
assert len(engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.18))], now=1006.0)) == 1
|
|
# 跌出预过滤区间 (快照中消失)
|
|
engine.evaluate_abnormal([], now=1012.0)
|
|
# 重新出现且超阈值 → 重新触发
|
|
assert len(engine.evaluate_abnormal([_row("600000.SH", ("3d", 0.18))], now=1018.0)) == 1
|
|
|
|
|
|
def test_engine_abnormal_direction_window_scope_filters() -> None:
|
|
# 方向: 只报上涨偏离
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule(direction="up")])
|
|
engine.evaluate_abnormal([_row("600000.SH", ("3d", -0.16))], now=1000.0)
|
|
assert engine.evaluate_abnormal([_row("600000.SH", ("3d", -0.19))], now=1006.0) == []
|
|
|
|
# 窗口: 只看 3d (10d/30d 的偏离不参与)
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule(abnormal_window="3d")])
|
|
engine.evaluate_abnormal([_row("600000.SH", ("10d", 0.98))], now=1000.0)
|
|
assert engine.evaluate_abnormal([_row("600000.SH", ("10d", 0.99))], now=1006.0) == []
|
|
|
|
# 作用域: 只监控指定标的
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule(scope="symbols", symbols=["600000.SH"])])
|
|
engine.evaluate_abnormal(
|
|
[_row("600000.SH", ("3d", 0.10)), _row("000001.SZ", ("3d", 0.10))], now=1000.0,
|
|
)
|
|
events = engine.evaluate_abnormal(
|
|
[_row("600000.SH", ("3d", 0.16)), _row("000001.SZ", ("3d", 0.19))], now=1006.0,
|
|
)
|
|
assert [ev["symbol"] for ev in events] == ["600000.SH"]
|
|
|
|
|
|
def test_engine_abnormal_down_direction_event_type() -> None:
|
|
engine = MonitorRuleEngine()
|
|
engine.set_rules([_ab_rule(direction="down")])
|
|
engine.evaluate_abnormal([_row("600000.SH", ("3d", -0.10))], now=1000.0)
|
|
events = engine.evaluate_abnormal([_row("600000.SH", ("3d", -0.16))], now=1006.0)
|
|
assert len(events) == 1
|
|
assert events[0]["type"] == "abnormal_down"
|