Rekey
4294ec401d
- engine/indicators/regime.py: RegimeDetector(四法投票) + MultiTimeframeRegime(多周期并行) 四法: EMA200斜率 / 价格vsEMA200 / ATH回撤 / 窄幅盘整(<3%振幅) 全部 O(1)/bar 增量计算,适用于回测和实时 - engine/example/regime_display.py: 多周期牛熊矩阵展示脚本 独立加载各周期数据 → 运行判定 → 日线对齐矩阵 + 详细拆解 + 统计 输出 engine/backtest/REGIME_MATRIX_BTCUSDT.md - engine/example/regime_mtf_strategy.py: 多周期共识策略 + 四策略对比回测 MTF Consensus: 1w定方向 + 1d确认 + 4h EMA入场 vs Old Regime(单TF基线) vs Long/Short(无过滤) - engine/indicators/__init__.py: 导出 RegimeDetector, MultiTimeframeRegime
438 lines
16 KiB
Python
438 lines
16 KiB
Python
"""
|
||
多周期牛熊矩阵 — 方案一：独立多周期判定矩阵
|
||
|
||
在统一时间轴（1d bar 边界）上对齐全周期牛熊判定，
|
||
输出矩阵表格、统计信息和详细拆解。
|
||
|
||
用法:
|
||
source .venv/bin/activate && python example/regime_display.py
|
||
|
||
定制:
|
||
python example/regime_display.py --symbol ETHUSDT
|
||
python example/regime_display.py --matrix-rows 50 # 输出最近 50 个时间点
|
||
"""
|
||
|
||
import asyncio
|
||
import sys
|
||
from datetime import datetime, timezone
|
||
from pathlib import Path
|
||
|
||
_project_root = Path(__file__).resolve().parent.parent.parent
|
||
if str(_project_root) not in sys.path:
|
||
sys.path.insert(0, str(_project_root))
|
||
|
||
from engine.common.config import config
|
||
from engine.data.service import DataService, INTERVAL_MS, INTERVAL_TO_TABLE
|
||
from engine.indicators.regime import RegimeDetector
|
||
|
||
# ── 默认配置 ──
|
||
|
||
SYMBOL = "BTCUSDT"
|
||
TIMEFRAMES = ["1h", "4h", "1d", "1w"]
|
||
MATRIX_ROWS = 30
|
||
OUTPUT_FILE = None # 运行时设置
|
||
|
||
|
||
def parse_args():
|
||
"""解析命令行参数"""
|
||
args = sys.argv[1:]
|
||
kwargs = {"symbol": SYMBOL, "matrix_rows": MATRIX_ROWS}
|
||
i = 0
|
||
while i < len(args):
|
||
if args[i] == "--symbol" and i + 1 < len(args):
|
||
kwargs["symbol"] = args[i + 1].upper()
|
||
i += 2
|
||
elif args[i] == "--matrix-rows" and i + 1 < len(args):
|
||
kwargs["matrix_rows"] = int(args[i + 1])
|
||
i += 2
|
||
else:
|
||
i += 1
|
||
return kwargs
|
||
|
||
|
||
# ════════════════════════════════════════════════════════
|
||
# 1. 加载各周期数据并运行独立判定
|
||
# ════════════════════════════════════════════════════════
|
||
|
||
|
||
async def load_and_detect(
|
||
ds: DataService, symbol: str, timeframe: str
|
||
) -> tuple[list[float], list[dict], datetime, datetime]:
|
||
"""加载指定周期全量数据，逐根运行 RegimeDetector，返回 (times, regimes, 数据起始, 数据结束)
|
||
|
||
regimes[i] = {"time_ms": float, "close": float, "regime": str, "detail": dict}
|
||
仅从第 220 根 bar 开始有效判定，之前为 unknown。
|
||
"""
|
||
start, end = await ds.fetch_symbol_date_range(symbol, timeframe)
|
||
all_klines = await ds.fetch_klines(
|
||
symbol=symbol,
|
||
interval=timeframe,
|
||
start_time=start,
|
||
end_time=end,
|
||
limit=10_000_000,
|
||
)
|
||
|
||
det = RegimeDetector()
|
||
regimes = []
|
||
times = []
|
||
for k in all_klines:
|
||
times.append(k.open_time)
|
||
det.update(k.close)
|
||
if det.ready:
|
||
regime = det.detect(k.close, len(det._e200) - 1)
|
||
detail = det.detect_detail(k.close, len(det._e200) - 1)
|
||
else:
|
||
regime = "unknown"
|
||
detail = {"final": "unknown"}
|
||
regimes.append({
|
||
"time_ms": k.open_time,
|
||
"time_dt": datetime.fromtimestamp(k.open_time / 1000, tz=timezone.utc),
|
||
"close": k.close,
|
||
"regime": regime,
|
||
"detail": detail,
|
||
"ema200": det.ema200,
|
||
"ath": det.ath,
|
||
})
|
||
|
||
return times, regimes, start, end
|
||
|
||
|
||
# ════════════════════════════════════════════════════════
|
||
# 2. 时间对齐 — 以日线为基准，查找各周期的即时判定
|
||
# ════════════════════════════════════════════════════════
|
||
|
||
|
||
def find_regime_at(
|
||
regimes: list[dict], target_ms: float, interval_ms: float
|
||
) -> dict:
|
||
"""在指定周期的 regime 序列中，找到 target_ms 时刻的判定。
|
||
|
||
规则：取开盘时间最接近且不晚于 target_ms 的那根 bar 的判定。
|
||
若 target_ms 在该 bar 的 [open_time, open_time + interval_ms) 范围内，返回该 bar 的判定。
|
||
"""
|
||
best = None
|
||
for r in regimes:
|
||
bar_time = r["time_ms"]
|
||
if bar_time <= target_ms < bar_time + interval_ms:
|
||
return r
|
||
if bar_time > target_ms:
|
||
break
|
||
best = r
|
||
return best
|
||
|
||
|
||
def build_matrix(
|
||
times_by_tf: dict[str, list[float]],
|
||
regimes_by_tf: dict[str, list[dict]],
|
||
reference_tf: str = "1d",
|
||
num_rows: int = MATRIX_ROWS,
|
||
) -> list[dict]:
|
||
"""以 reference_tf 的 bar 时间点为基准，构建多周期判定矩阵。
|
||
|
||
Returns:
|
||
[{time_dt, time_str, 1h: regime, 4h: regime, 1d: regime, 1w: regime}, ...]
|
||
按时间倒序，只包含各周期均有有效数据的时间点。
|
||
"""
|
||
ref_regimes = regimes_by_tf[reference_tf]
|
||
# 只取有效区间（所有周期都过了 220 根 bar 之后）
|
||
ref_valid = [
|
||
r for r in ref_regimes
|
||
if r["regime"] != "unknown"
|
||
]
|
||
|
||
matrix = []
|
||
for ref_r in ref_valid[-num_rows:]:
|
||
row = {
|
||
"time_dt": ref_r["time_dt"],
|
||
"time_str": ref_r["time_dt"].strftime("%Y-%m-%d %H:%M"),
|
||
"time_ms": ref_r["time_ms"],
|
||
f"{reference_tf}_close": ref_r["close"],
|
||
}
|
||
for tf in TIMEFRAMES:
|
||
if tf == reference_tf:
|
||
row[tf] = ref_r["regime"]
|
||
row[f"{tf}_detail"] = ref_r["detail"]
|
||
row[f"{tf}_ema200"] = ref_r.get("ema200", 0)
|
||
row[f"{tf}_ath"] = ref_r.get("ath", 0)
|
||
else:
|
||
tf_regime = find_regime_at(
|
||
regimes_by_tf[tf],
|
||
ref_r["time_ms"],
|
||
INTERVAL_MS[tf],
|
||
)
|
||
if tf_regime:
|
||
row[tf] = tf_regime["regime"]
|
||
row[f"{tf}_detail"] = tf_regime["detail"]
|
||
row[f"{tf}_ema200"] = tf_regime.get("ema200", 0)
|
||
row[f"{tf}_ath"] = tf_regime.get("ath", 0)
|
||
else:
|
||
row[tf] = "—"
|
||
matrix.append(row)
|
||
|
||
# 倒序（最新在前）
|
||
return list(reversed(matrix))
|
||
|
||
|
||
# ════════════════════════════════════════════════════════
|
||
# 3. 统计分析
|
||
# ════════════════════════════════════════════════════════
|
||
|
||
|
||
def regime_icon(r: str) -> str:
|
||
return {"bull": "🐂", "bear": "🐻", "sideways": "⚪", "unknown": "⏳", "—": "—"}.get(r, r)
|
||
|
||
|
||
def calc_alignment_stats(matrix: list[dict]) -> dict:
|
||
"""计算各周期判定对齐统计"""
|
||
total = len(matrix)
|
||
if total == 0:
|
||
return {}
|
||
|
||
stats = {}
|
||
for tf in TIMEFRAMES:
|
||
cnt_bull = sum(1 for r in matrix if r.get(tf) == "bull")
|
||
cnt_bear = sum(1 for r in matrix if r.get(tf) == "bear")
|
||
cnt_side = sum(1 for r in matrix if r.get(tf) == "sideways")
|
||
stats[tf] = {
|
||
"bull": cnt_bull,
|
||
"bull_pct": cnt_bull / total * 100,
|
||
"bear": cnt_bear,
|
||
"bear_pct": cnt_bear / total * 100,
|
||
"sideways": cnt_side,
|
||
"sideways_pct": cnt_side / total * 100,
|
||
}
|
||
|
||
# 全周期一致（所有非 "—" 的周期判定相同）
|
||
all_agree = 0
|
||
for r in matrix:
|
||
regimes = [r[tf] for tf in TIMEFRAMES if r.get(tf) not in ("—", "unknown")]
|
||
if regimes and len(set(regimes)) == 1:
|
||
all_agree += 1
|
||
stats["all_agree"] = all_agree
|
||
stats["all_agree_pct"] = all_agree / total * 100
|
||
|
||
# 大小周期背离（1h 与 1w 相反）
|
||
diverge = 0
|
||
for r in matrix:
|
||
r1h = r.get("1h")
|
||
r1w = r.get("1w")
|
||
if r1h and r1w and r1h != "—" and r1w != "—":
|
||
if (r1h == "bull" and r1w == "bear") or (r1h == "bear" and r1w == "bull"):
|
||
diverge += 1
|
||
stats["diverge_1h_1w"] = diverge
|
||
stats["diverge_1h_1w_pct"] = diverge / total * 100
|
||
|
||
return stats
|
||
|
||
|
||
# ════════════════════════════════════════════════════════
|
||
# 4. 输出格式化
|
||
# ════════════════════════════════════════════════════════
|
||
|
||
|
||
def render_matrix_table(matrix: list[dict]) -> str:
|
||
"""渲染矩阵表格为 Markdown 格式"""
|
||
header = "| 时间 |"
|
||
sep = "|------|"
|
||
for tf in TIMEFRAMES:
|
||
header += f" {tf} |"
|
||
sep += "----|"
|
||
header += " BTC Close |"
|
||
sep += "-----------|"
|
||
|
||
lines = [header, sep]
|
||
for row in matrix:
|
||
line = f"| {row['time_str']} |"
|
||
for tf in TIMEFRAMES:
|
||
r = row.get(tf, "—")
|
||
line += f" {regime_icon(r)} {r} |"
|
||
close_str = f"{row.get('1d_close', 0):.2f}"
|
||
line += f" {close_str} |"
|
||
lines.append(line)
|
||
|
||
return "\n".join(lines)
|
||
|
||
|
||
def render_detail_table(matrix: list[dict], tf: str, n: int = 5) -> str:
|
||
"""渲染单个周期的详细判定表（最近 n 条）"""
|
||
lines = [
|
||
f"### {tf} 详细判定（最近 {n} 个时间点）",
|
||
"",
|
||
"| 时间 | 最终 | EMA200斜率 | 价格vsEMA200 | ATH回撤 | 窄幅盘整 | 振幅% | 牛票 | 熊票 | 震票 | EMA200 | ATH |",
|
||
"|------|------|-----------|-------------|---------|---------|------|------|------|------|--------|-----|",
|
||
]
|
||
|
||
detail_key = f"{tf}_detail"
|
||
for row in matrix[:n]:
|
||
d = row.get(detail_key, {})
|
||
f_icon = regime_icon(d.get("final", "—"))
|
||
ema200_val = row.get(f"{tf}_ema200", 0)
|
||
ath_val = row.get(f"{tf}_ath", 0)
|
||
range_pct = d.get("range_pct", 0) * 100
|
||
lines.append(
|
||
f"| {row['time_str']} | {f_icon} {d.get('final', '—')} | "
|
||
f"{d.get('ema200_slope', '—')} | {d.get('price_vs_ema200', '—')} | "
|
||
f"{d.get('ath_drawdown', '—')} | {d.get('price_range', '—')} | "
|
||
f"{range_pct:.1f}% | {d.get('bull_votes', '—')} | "
|
||
f"{d.get('bear_votes', '—')} | {d.get('sideways_votes', '—')} | "
|
||
f"{ema200_val:.2f} | {ath_val:.2f} |"
|
||
)
|
||
|
||
return "\n".join(lines)
|
||
|
||
|
||
def render_current_status(regimes_by_tf: dict[str, list[dict]]) -> str:
|
||
"""渲染当前最新状态"""
|
||
lines = [
|
||
"## 当前各周期牛熊状态",
|
||
"",
|
||
"| 周期 | 判定 | EMA200斜率 | 价格vsEMA200 | ATH回撤 | 窄幅盘整 | 振幅% | 牛票 | 熊票 | 震票 | EMA200 | Close |",
|
||
"|------|------|-----------|-------------|---------|---------|------|------|------|------|--------|-------|",
|
||
]
|
||
|
||
for tf in TIMEFRAMES:
|
||
regs = regimes_by_tf[tf]
|
||
if not regs or regs[-1]["regime"] == "unknown":
|
||
lines.append(f"| {tf} | ⏳ 数据不足 | — | — | — | — | — | — | — | — | — | — |")
|
||
continue
|
||
r = regs[-1]
|
||
d = r["detail"]
|
||
icon = regime_icon(r["regime"])
|
||
range_pct = d.get("range_pct", 0) * 100
|
||
lines.append(
|
||
f"| {tf} | {icon} **{r['regime']}** | "
|
||
f"{d.get('ema200_slope', '—')} | {d.get('price_vs_ema200', '—')} | "
|
||
f"{d.get('ath_drawdown', '—')} | {d.get('price_range', '—')} | "
|
||
f"{range_pct:.1f}% | {d.get('bull_votes', '—')} | "
|
||
f"{d.get('bear_votes', '—')} | {d.get('sideways_votes', '—')} | "
|
||
f"{r.get('ema200', 0):.2f} | {r['close']:.2f} |"
|
||
)
|
||
|
||
return "\n".join(lines)
|
||
|
||
|
||
def render_stats(stats: dict) -> str:
|
||
"""渲染统计信息"""
|
||
lines = [
|
||
"## 周期判定分布统计",
|
||
"",
|
||
"| 周期 | 🐂 牛市 | 占比 | 🐻 熊市 | 占比 | ⚪ 震荡 | 占比 |",
|
||
"|------|--------|------|--------|------|--------|------|",
|
||
]
|
||
for tf in TIMEFRAMES:
|
||
s = stats.get(tf, {})
|
||
lines.append(
|
||
f"| {tf} | {s.get('bull', 0)} | {s.get('bull_pct', 0):.1f}% | "
|
||
f"{s.get('bear', 0)} | {s.get('bear_pct', 0):.1f}% | "
|
||
f"{s.get('sideways', 0)} | {s.get('sideways_pct', 0):.1f}% |"
|
||
)
|
||
|
||
lines.append("")
|
||
lines.append("## 周期一致性统计")
|
||
lines.append("")
|
||
lines.append(f"- **全周期一致**（所有周期判定相同）：{stats.get('all_agree', 0)} 天 ({stats.get('all_agree_pct', 0):.1f}%)")
|
||
lines.append(f"- **1h ↔ 1w 背离**（1h 与 1w 方向相反）：{stats.get('diverge_1h_1w', 0)} 天 ({stats.get('diverge_1h_1w_pct', 0):.1f}%)")
|
||
|
||
return "\n".join(lines)
|
||
|
||
|
||
# ════════════════════════════════════════════════════════
|
||
# 主流程
|
||
# ════════════════════════════════════════════════════════
|
||
|
||
|
||
async def main():
|
||
kw = parse_args()
|
||
symbol = kw["symbol"]
|
||
matrix_rows = kw["matrix_rows"]
|
||
global OUTPUT_FILE
|
||
OUTPUT_FILE = (
|
||
Path(__file__).resolve().parent.parent
|
||
/ "backtest"
|
||
/ f"REGIME_MATRIX_{symbol}.md"
|
||
)
|
||
|
||
out: list[str] = []
|
||
|
||
def w(line: str = ""):
|
||
out.append(line)
|
||
print(line)
|
||
|
||
w(f"# 多周期牛熊判定矩阵 — {symbol}")
|
||
w()
|
||
w(f"> 生成时间：{datetime.now().strftime('%Y-%m-%d %H:%M:%S')} UTC")
|
||
w(f"> 判定方法：EMA200斜率 + 价格vsEMA200 + ATH回撤 + 窄幅盘整，四选二投票")
|
||
w(f"> 时间对齐基准：日线(1d) bar 边界")
|
||
w()
|
||
|
||
# ── 加载数据并运行判定 ──
|
||
ds = DataService(config.db)
|
||
await ds.connect()
|
||
|
||
try:
|
||
times_by_tf: dict[str, list[float]] = {}
|
||
regimes_by_tf: dict[str, list[dict]] = {}
|
||
date_ranges: dict[str, tuple[datetime, datetime]] = {}
|
||
|
||
for tf in TIMEFRAMES:
|
||
print(f" 加载 {tf} 数据...", end=" ", flush=True)
|
||
times, regimes, data_start, data_end = await load_and_detect(ds, symbol, tf)
|
||
times_by_tf[tf] = times
|
||
regimes_by_tf[tf] = regimes
|
||
date_ranges[tf] = (data_start, data_end)
|
||
valid_count = sum(1 for r in regimes if r["regime"] != "unknown")
|
||
print(f"{len(regimes)} 根 bar，{valid_count} 根有效（>220热身）")
|
||
|
||
# ── 数据范围 ──
|
||
w()
|
||
w("## 数据覆盖范围")
|
||
w()
|
||
w("| 周期 | 数据条数 | 有效条数 | 起始时间 | 结束时间 |")
|
||
w("|------|---------|---------|---------|---------|")
|
||
for tf in TIMEFRAMES:
|
||
regs = regimes_by_tf[tf]
|
||
valid = sum(1 for r in regs if r["regime"] != "unknown")
|
||
s, e = date_ranges[tf]
|
||
w(f"| {tf} | {len(regs)} | {valid} | {s.strftime('%Y-%m-%d')} | {e.strftime('%Y-%m-%d')} |")
|
||
w()
|
||
|
||
# ── 当前状态 ──
|
||
w(render_current_status(regimes_by_tf))
|
||
w()
|
||
|
||
# ── 历史矩阵 ──
|
||
matrix = build_matrix(times_by_tf, regimes_by_tf, "1d", num_rows=matrix_rows)
|
||
|
||
w(f"## 历史牛熊矩阵（最近 {len(matrix)} 个日线时间点）")
|
||
w()
|
||
w("> 以日线 bar 的 open_time 为基准，查找各周期在该时刻的即时判定。")
|
||
w()
|
||
w(render_matrix_table(matrix))
|
||
w()
|
||
|
||
# ── 各周期详细拆解 ──
|
||
w("## 各周期详细拆解")
|
||
w()
|
||
for tf in TIMEFRAMES:
|
||
w(render_detail_table(matrix, tf, n=min(5, len(matrix))))
|
||
w()
|
||
|
||
# ── 统计 ──
|
||
stats = calc_alignment_stats(matrix)
|
||
w(render_stats(stats))
|
||
w()
|
||
|
||
finally:
|
||
await ds.close()
|
||
|
||
# ── 写出文件 ──
|
||
OUTPUT_FILE.parent.mkdir(parents=True, exist_ok=True)
|
||
with open(OUTPUT_FILE, "w", encoding="utf-8") as f:
|
||
f.write("\n".join(out) + "\n")
|
||
|
||
print(f"\n✓ 结果已保存到: {OUTPUT_FILE}")
|
||
|
||
|
||
if __name__ == "__main__":
|
||
asyncio.run(main())
|