zhutoutoutousan
605faf5310
Add cluster audit pipeline, united EA updates, brochure generators, and publication hygiene (gitignore, MT5 path desensitization, pre-upload scan). Remove tracked reports, models, and binary artifacts from the repo. Co-authored-by: Cursor <cursoragent@cursor.com>
773 lines
31 KiB
Python
773 lines
31 KiB
Python
#!/usr/bin/env python3
|
||
"""Generate cluster backtest evaluation: charts + LaTeX PDF."""
|
||
from __future__ import annotations
|
||
|
||
import json
|
||
import re
|
||
import subprocess
|
||
import textwrap
|
||
from collections import defaultdict
|
||
from dataclasses import dataclass
|
||
from datetime import datetime
|
||
from html import unescape
|
||
from pathlib import Path
|
||
|
||
import matplotlib
|
||
matplotlib.use("Agg")
|
||
import matplotlib.dates as mdates
|
||
import matplotlib.pyplot as plt
|
||
import MetaTrader5 as mt5
|
||
import numpy as np
|
||
import pandas as pd
|
||
from matplotlib.patches import Rectangle
|
||
|
||
ROOT = Path(__file__).resolve().parent
|
||
|
||
|
||
def _find_tester_report() -> Path:
|
||
"""Locate MT5 Strategy Tester HTML (ReportTester-*.html) in cluster-latest."""
|
||
candidates = sorted(ROOT.parent.glob("ReportTester*.html"), key=lambda p: p.stat().st_mtime, reverse=True)
|
||
if not candidates:
|
||
raise FileNotFoundError(
|
||
"No ReportTester*.html found. Export from MT5 Strategy Tester into frontline/cluster-latest/"
|
||
)
|
||
return candidates[0]
|
||
|
||
|
||
REPORT_HTML = _find_tester_report()
|
||
FIG_DIR = ROOT / "figures"
|
||
DATA_DIR = ROOT / "data"
|
||
TEX_FILE = ROOT / "cluster_evaluation.tex"
|
||
|
||
plt.rcParams.update({
|
||
"figure.dpi": 150,
|
||
"savefig.dpi": 150,
|
||
"font.size": 9,
|
||
"axes.titlesize": 10,
|
||
"axes.labelsize": 9,
|
||
})
|
||
|
||
STRATEGY_META = {
|
||
"RM_EMA_Cross": {"name": "RM MidPoint EMA Cross", "group": "gold_core", "type": "trend"},
|
||
"RM_RSI_Follow": {"name": "RM MidPoint RSI Follow", "group": "gold_core", "type": "trend"},
|
||
"RM_RSI_Reverse": {"name": "RM MidPoint RSI Reverse", "group": "gold_core", "type": "mean_rev"},
|
||
"RS_XAUUSD": {"name": "RSI Scalping XAUUSD", "group": "gold_core", "type": "scalp"},
|
||
"RCO_RSIConsolidation": {"name": "RSI Consolidation", "group": "gold_core", "type": "range"},
|
||
"RS_BTCUSD": {"name": "RSI Scalping BTCUSD", "group": "crypto", "type": "scalp"},
|
||
"RS_TSLA": {"name": "RSI Scalping TSLA", "group": "equity", "type": "scalp"},
|
||
"RS_NVDA": {"name": "RSI Scalping NVDA", "group": "equity", "type": "scalp"},
|
||
"RS_APPL": {"name": "RSI Scalping AAPL", "group": "equity", "type": "scalp"},
|
||
"RC_RSICrossOver": {"name": "RSI CrossOver Reversal", "group": "gold_core", "type": "reversal"},
|
||
"SE_SuperEMA": {"name": "SuperEMA", "group": "gold_core", "type": "trend"},
|
||
"DB_DarvasBox": {"name": "Darvas Box", "group": "gold_core", "type": "breakout"},
|
||
"ES_EMASlope": {"name": "EMA Slope Distance", "group": "gold_core", "type": "trend"},
|
||
"ST_SimpleTrendline": {"name": "SimpleTrendline", "group": "multi", "type": "hedge"},
|
||
"RRA_EURUSD": {"name": "RSI Asian EURUSD", "group": "fx", "type": "session"},
|
||
"RRA_AUDUSD": {"name": "RSI Asian AUDUSD", "group": "fx", "type": "session"},
|
||
"UB_USDJPY": {"name": "USDJPY Buster", "group": "fx", "type": "breakout"},
|
||
}
|
||
|
||
|
||
def classify_comment(c: str, symbol: str = "") -> str | None:
|
||
c = c.strip()
|
||
if not c:
|
||
return None
|
||
cl = c.lower()
|
||
if "darvas" in cl:
|
||
return "DB_DarvasBox"
|
||
if "ema cross distance" in cl:
|
||
return "ES_EMASlope"
|
||
if "ema crossover trade" in cl:
|
||
return "RM_EMA_Cross"
|
||
if c in ("Buy Order", "Sell Order"):
|
||
return "RC_RSICrossOver"
|
||
if "united superema" in cl:
|
||
return "SE_SuperEMA"
|
||
if "rsiconsolidation" in cl.replace(" ", ""):
|
||
return "RCO_RSIConsolidation"
|
||
if "simpletrendline" in cl.replace(" ", ""):
|
||
return "ST_SimpleTrendline"
|
||
if "rsi overbought crossover" in cl or "rsi oversold crossover" in cl:
|
||
return "RRA_EURUSD" if "EUR" in symbol else "RRA_AUDUSD" if "AUD" in symbol else "RRA_Asian"
|
||
if c.startswith("UB range"):
|
||
return "UB_USDJPY"
|
||
if "rsi follow" in cl:
|
||
return "RM_RSI_Follow"
|
||
if "rsi reverse" in cl:
|
||
return "RM_RSI_Reverse"
|
||
if "rsi scalping" in cl:
|
||
sym_map = {
|
||
"AAPL.NAS": "RS_APPL", "ADBE.NAS": "RS_ADBE", "BTCUSD": "RS_BTCUSD",
|
||
"NVDA.NAS": "RS_NVDA", "TSLA.NAS": "RS_TSLA", "XAUUSD": "RS_XAUUSD", "MU.NAS": "RS_MU",
|
||
}
|
||
return sym_map.get(symbol, f"RS_{symbol}")
|
||
if c.startswith("sl ") or c.startswith("tp ") or c == "end of test":
|
||
return None
|
||
return None
|
||
|
||
|
||
def st_side(comment: str) -> str | None:
|
||
cl = comment.lower().replace(" ", "")
|
||
if "simpletrendlinebuy" in cl:
|
||
return "long"
|
||
if "simpletrendlinesell" in cl:
|
||
return "short"
|
||
return None
|
||
|
||
|
||
def parse_num(s: str) -> float:
|
||
s = s.strip().replace(" ", "").replace(",", "")
|
||
if not s:
|
||
return 0.0
|
||
if s.endswith("K"):
|
||
return float(s[:-1]) * 1000
|
||
return float(s)
|
||
|
||
|
||
def parse_deals(html: str) -> list[dict]:
|
||
marker = '<div style="font: 10pt Tahoma"><b>成交</b></div>'
|
||
start = html.find(marker)
|
||
sub = html[start:] if start >= 0 else html
|
||
row_re = re.compile(
|
||
r"align=right><td>([^<]*)</td><td>(\d+)</td><td>([^<]*)</td><td>([^<]*)</td>"
|
||
r"<td>([^<]*)</td><td>([^<]*)</td><td>([^<]*)</td><td>(\d+)</td><td>([^<]*)</td>"
|
||
r"<td>([^<]*)</td><td>([^<]*)</td><td>([^<]*)</td><td>([^<]*)</td></tr>",
|
||
re.I,
|
||
)
|
||
open_map: dict[str, str] = {}
|
||
open_side: dict[str, str] = {}
|
||
open_stack: dict[str, list[tuple[str, str, str | None]]] = defaultdict(list)
|
||
deals = []
|
||
|
||
for row in row_re.findall(sub):
|
||
time_s, _did, symbol, typ, direction, _vol, _price, order, comm, swap, profit, balance, comment = row
|
||
if typ.strip().lower() == "balance":
|
||
continue
|
||
net = parse_num(profit) + parse_num(comm) + parse_num(swap)
|
||
comment = unescape(comment.strip())
|
||
symbol = symbol.strip()
|
||
direction = direction.strip().lower()
|
||
strategy = classify_comment(comment, symbol)
|
||
side = st_side(comment) if strategy == "ST_SimpleTrendline" else None
|
||
|
||
if direction == "in":
|
||
if strategy:
|
||
open_map[order] = strategy
|
||
if side:
|
||
open_side[order] = side
|
||
open_stack[symbol].append((order, strategy, side))
|
||
continue
|
||
if direction != "out":
|
||
continue
|
||
|
||
if strategy is None:
|
||
if open_stack[symbol]:
|
||
_o, strategy, side = open_stack[symbol].pop(0)
|
||
elif order in open_map:
|
||
strategy = open_map[order]
|
||
side = open_side.get(order)
|
||
else:
|
||
strategy = "UNATTRIBUTED"
|
||
elif open_stack[symbol]:
|
||
# explicit close comment: still consume FIFO slot for side/symbol match
|
||
for i, (o, s, sd) in enumerate(open_stack[symbol]):
|
||
if s == strategy:
|
||
_o, _, side = open_stack[symbol].pop(i)
|
||
break
|
||
|
||
open_map.pop(order, None)
|
||
open_side.pop(order, None)
|
||
open_stack[symbol] = [(o, s, sd) for o, s, sd in open_stack[symbol] if o != order]
|
||
|
||
try:
|
||
ts = pd.to_datetime(time_s, format="%Y.%m.%d %H:%M:%S")
|
||
except Exception:
|
||
ts = pd.to_datetime(time_s.replace(".", "-", 2))
|
||
|
||
deals.append({
|
||
"time": ts,
|
||
"symbol": symbol,
|
||
"strategy": strategy,
|
||
"side": side,
|
||
"net": net,
|
||
"balance": parse_num(balance),
|
||
"comment": comment,
|
||
})
|
||
return deals
|
||
|
||
|
||
def load_xau_h1(start: datetime, end: datetime) -> pd.DataFrame:
|
||
if not mt5.initialize():
|
||
raise RuntimeError("MT5 init failed")
|
||
sym = "XAUUSD"
|
||
if not mt5.symbol_select(sym, True):
|
||
for alt in ("GOLD", "XAUUSDm", "XAUUSD."):
|
||
if mt5.symbol_select(alt, True):
|
||
sym = alt
|
||
break
|
||
rates = mt5.copy_rates_range(sym, mt5.TIMEFRAME_H1, start, end)
|
||
mt5.shutdown()
|
||
if rates is None or len(rates) < 100:
|
||
raise RuntimeError("No XAUUSD H1 data from MT5")
|
||
df = pd.DataFrame(rates)
|
||
df["time"] = pd.to_datetime(df["time"], unit="s")
|
||
df.set_index("time", inplace=True)
|
||
return df
|
||
|
||
|
||
def market_features(df: pd.DataFrame) -> pd.DataFrame:
|
||
out = df.copy()
|
||
out["ret"] = out["close"].pct_change()
|
||
out["ema20"] = out["close"].ewm(span=20).mean()
|
||
out["ema50"] = out["close"].ewm(span=50).mean()
|
||
hl = out["high"] - out["low"]
|
||
hc = (out["high"] - out["close"].shift()).abs()
|
||
lc = (out["low"] - out["close"].shift()).abs()
|
||
tr = pd.concat([hl, hc, lc], axis=1).max(axis=1)
|
||
out["atr14"] = tr.rolling(14).mean()
|
||
out["atr_pct"] = out["atr14"] / out["close"] * 100
|
||
up = out["high"].diff()
|
||
down = -out["low"].diff()
|
||
plus_dm = np.where((up > down) & (up > 0), up, 0.0)
|
||
minus_dm = np.where((down > up) & (down > 0), down, 0.0)
|
||
atr = out["atr14"]
|
||
plus_di = 100 * pd.Series(plus_dm, index=out.index).rolling(14).mean() / atr
|
||
minus_di = 100 * pd.Series(minus_dm, index=out.index).rolling(14).mean() / atr
|
||
dx = (plus_di - minus_di).abs() / (plus_di + minus_di) * 100
|
||
out["adx14"] = dx.rolling(14).mean()
|
||
out["trend_slope"] = (out["ema20"] - out["ema50"]) / out["close"] * 100
|
||
out["chop"] = out["ret"].rolling(24).std() * np.sqrt(24) * 100
|
||
return out
|
||
|
||
|
||
def find_drawdowns(equity: pd.Series, top_n: int = 5) -> list[dict]:
|
||
peak = equity.cummax()
|
||
dd = equity - peak
|
||
dd_pct = dd / peak.replace(0, np.nan) * 100
|
||
episodes = []
|
||
in_dd = False
|
||
start = None
|
||
trough_t = None
|
||
trough_v = 0.0
|
||
peak_v = 0.0
|
||
for t, v in dd.items():
|
||
if v < -1 and not in_dd:
|
||
in_dd = True
|
||
start = t
|
||
trough_t = t
|
||
trough_v = v
|
||
peak_v = peak.loc[t]
|
||
elif in_dd:
|
||
if v < trough_v:
|
||
trough_v = v
|
||
trough_t = t
|
||
if v >= -0.5:
|
||
episodes.append({
|
||
"start": start, "trough": trough_t, "end": t,
|
||
"depth_usd": abs(trough_v), "depth_pct": abs(trough_v / peak_v * 100) if peak_v else 0,
|
||
})
|
||
in_dd = False
|
||
if in_dd:
|
||
episodes.append({
|
||
"start": start, "trough": trough_t, "end": equity.index[-1],
|
||
"depth_usd": abs(trough_v), "depth_pct": abs(trough_v / peak_v * 100) if peak_v else 0,
|
||
})
|
||
episodes.sort(key=lambda x: x["depth_usd"], reverse=True)
|
||
return episodes[:top_n]
|
||
|
||
|
||
def plot_candles(ax, df: pd.DataFrame, title: str):
|
||
sub = df.tail(min(120, len(df)))
|
||
xs = np.arange(len(sub))
|
||
for i, (t, row) in enumerate(sub.iterrows()):
|
||
o, h, l, c = row["open"], row["high"], row["low"], row["close"]
|
||
color = "#26a69a" if c >= o else "#ef5350"
|
||
ax.plot([i, i], [l, h], color=color, linewidth=0.8)
|
||
ax.add_patch(Rectangle((i - 0.35, min(o, c)), 0.7, abs(c - o) or 0.5, facecolor=color, edgecolor=color))
|
||
ax.plot(xs, sub["ema20"].values, color="#1565c0", linewidth=1, label="EMA20")
|
||
ax.plot(xs, sub["ema50"].values, color="#ff8f00", linewidth=1, label="EMA50")
|
||
ax.set_title(title)
|
||
ax.set_xlim(-1, len(sub))
|
||
ax.legend(loc="upper left", fontsize=7)
|
||
step = max(1, len(sub) // 6)
|
||
ax.set_xticks(xs[::step])
|
||
ax.set_xticklabels([sub.index[j].strftime("%m-%d") for j in range(0, len(sub), step)], rotation=30, ha="right")
|
||
|
||
|
||
def monthly_issue_text(strat: str, monthly: pd.Series, meta: dict) -> list[str]:
|
||
issues = []
|
||
neg = monthly[monthly < 0]
|
||
if len(neg) > 0:
|
||
worst_m = neg.idxmin()
|
||
issues.append(f"{worst_m} 单月亏损 {neg.min():,.0f} USD")
|
||
pos = monthly[monthly > 0]
|
||
if len(pos) > 0 and len(neg) > len(pos):
|
||
issues.append(f"亏损月份 ({len(neg)}) 多于盈利月份 ({len(pos)})")
|
||
if strat == "ST_SimpleTrendline":
|
||
issues.append("对冲型策略：多空均有信号，需分方向评估而非整体关停")
|
||
if meta.get("type") == "scalp" and monthly.std() > abs(monthly.mean()) * 2:
|
||
issues.append("月度波动大，手数复利放大后尾部风险显著")
|
||
if monthly.tail(3).sum() < 0 and monthly.sum() > 0:
|
||
issues.append("近 3 个月转弱，存在 regime change 迹象")
|
||
return issues[:4]
|
||
|
||
|
||
def latex_escape(s: str) -> str:
|
||
return s.replace("\\", "\\textbackslash{}").replace("_", "\\_").replace("&", "\\&").replace("%", "\\%")
|
||
|
||
|
||
def build_latex(summary: dict) -> str:
|
||
s = summary
|
||
robots_tex = []
|
||
for strat in s["strategy_order"]:
|
||
meta = STRATEGY_META.get(strat, {"name": strat, "type": "other"})
|
||
m = s["monthly"].get(strat, {})
|
||
issues = s["issues"].get(strat, [])
|
||
total = s["totals"].get(strat, 0)
|
||
robots_tex.append(
|
||
f"\\subsubsection{{{latex_escape(meta['name'])}}}\n"
|
||
f"累计净利润 \\textbf{{{total:,.0f}}} USD。"
|
||
+ ("\\par\\noindent\\textbf{逐月问题：}\\begin{itemize}\\setlength\\itemsep{2pt}\n"
|
||
+ "".join(f"\\item {latex_escape(x)}" for x in issues)
|
||
+ "\\end{itemize}" if issues else "")
|
||
)
|
||
|
||
dd_rows = []
|
||
for i, ep in enumerate(s["drawdowns"][:5], 1):
|
||
dd_rows.append(
|
||
f"{i} & {ep['start'].strftime('%Y-%m-%d')} & {ep['trough'].strftime('%Y-%m-%d')} & "
|
||
f"{ep['end'].strftime('%Y-%m-%d')} & {ep['depth_usd']:,.0f} & {ep['depth_pct']:.1f}\\% \\\\"
|
||
)
|
||
|
||
st_total = s["totals"].get("ST_SimpleTrendline", 0)
|
||
body = f"""
|
||
\\section{{执行摘要}}
|
||
本报告基于 MT5 策略测试器 HTML 报告（初始资金 \\$3,000，杠杆 1:1000，余额复利缩放）对 UnitedEA 集群进行逐机器人、逐月、回撤期市场结构分析。
|
||
|
||
\\begin{{table}}[H]
|
||
\\centering
|
||
\\caption{{组合层面关键指标}}
|
||
\\begin{{tabular}}{{lr}}
|
||
\\toprule
|
||
指标 & 数值 \\\\
|
||
\\midrule
|
||
总净盈利 & {s['portfolio']['net_profit']:,.0f} USD \\\\
|
||
盈利因子 & {s['portfolio']['pf']:.2f} \\\\
|
||
夏普比率 & {s['portfolio']['sharpe']:.2f} \\\\
|
||
最大净值回撤 & {s['portfolio']['max_dd_pct']:.2f}\\% \\\\
|
||
总交易笔数 & {s['portfolio']['trades']:,} \\\\
|
||
胜率 & {s['portfolio']['win_rate']:.2f}\\% \\\\
|
||
\\bottomrule
|
||
\\end{{tabular}}
|
||
\\end{{table}}
|
||
|
||
\\textbf{{核心结论：}}
|
||
\\begin{{itemize}}
|
||
\\item 利润高度集中于 XAUUSD 上的 RM EMA Cross、RSI Scalping XAU、RSI Consolidation；名义多品种分散，实际为黄金 beta 集群。
|
||
\\item 2025 末至 2026 初复利手数放大后，收益与回撤同步膨胀；评估 edge 需配合固定手数对照。
|
||
\\item SimpleTrendline 为\\textbf{{对冲型}}（多空双向），整体虽亏但分方向与品种后可能仍有价值，不宜简单一刀切关闭。
|
||
\\item 最大回撤期 XAUUSD 呈现更高 ATR、更低 ADX（震荡/假突破增多），趋势型与突破型策略易共振亏损。
|
||
\\end{{itemize}}
|
||
|
||
\\section{{组合曲线与回撤}}
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.95\\textwidth]{{figures/equity_drawdown.pdf}}
|
||
\\caption{{净值曲线与回撤百分比}}
|
||
\\end{{figure}}
|
||
|
||
\\begin{{table}}[H]
|
||
\\centering
|
||
\\caption{{主要回撤 episode（Top 5）}}
|
||
\\begin{{tabular}}{{clllrr}}
|
||
\\toprule
|
||
\\# & 起始 & 谷底 & 恢复 & 深度(USD) & 深度(\\%) \\\\
|
||
\\midrule
|
||
{chr(10).join(dd_rows)}
|
||
\\bottomrule
|
||
\\end{{tabular}}
|
||
\\end{{table}}
|
||
|
||
\\section{{逐机器人月度热力与累计贡献}}
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.98\\textwidth]{{figures/monthly_heatmap.pdf}}
|
||
\\caption{{各机器人逐月净利润热力图（USD）}}
|
||
\\end{{figure}}
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.85\\textwidth]{{figures/strategy_contribution.pdf}}
|
||
\\caption{{各机器人累计净利润贡献}}
|
||
\\end{{figure}}
|
||
|
||
\\section{{SimpleTrendline：对冲型专项分析}}
|
||
SimpleTrendline 基于高周期 MA 交叉趋势线，\\textbf{{多空双向}}触发，设计目标是对冲而非单边趋势押注。
|
||
回测合计 {st_total:,.0f} USD，但 2026 年在黄金高位剧烈震荡中大幅回撤。
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.95\\textwidth]{{figures/st_long_short_monthly.pdf}}
|
||
\\caption{{SimpleTrendline 多空方向逐月 P/L 分解}}
|
||
\\end{{figure}}
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.95\\textwidth]{{figures/st_by_symbol_monthly.pdf}}
|
||
\\caption{{SimpleTrendline 分品种（XAU/GER/BTC）逐月 P/L}}
|
||
\\end{{figure}}
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.98\\textwidth]{{figures/top6_monthly.pdf}}
|
||
\\caption{{核心六机器人逐月 P/L 明细}}
|
||
\\end{{figure}}
|
||
|
||
\\textbf{{建议：}}保留策略框架，但 (1) 在 ADX>{s['regime']['dd_adx']:.0f} 且 ATR 分位>{s['regime']['dd_atr_pct']:.0f}\\% 的强趋势月减少逆势侧仓位；
|
||
(2) 与 RM/RCO 等同向 exposure 设上限；(3) 2026 类高位宽幅震荡月单独降 LOT。
|
||
|
||
\\section{{回撤期 vs 平稳期：XAUUSD H1 市场结构}}
|
||
从 MT5 拉取 XAUUSD H1。回撤窗口取 Top3 drawdown episode（{', '.join(s['regime'].get('dd_months', []))}），对照期为 2024-03 至 2024-09 平稳盈利段。
|
||
|
||
\\begin{{table}}[H]
|
||
\\centering
|
||
\\caption{{市场特征对比（亏损月 vs 对照月）}}
|
||
\\begin{{tabular}}{{lrr}}
|
||
\\toprule
|
||
特征 & 亏损月均值 & 对照月均值 \\\\
|
||
\\midrule
|
||
ADX(14) & {s['regime']['dd_adx']:.1f} & {s['regime']['calm_adx']:.1f} \\\\
|
||
ATR\\% & {s['regime']['dd_atr_pct']:.2f} & {s['regime']['calm_atr_pct']:.2f} \\\\
|
||
日收益波动(chop) & {s['regime']['dd_chop']:.2f} & {s['regime']['calm_chop']:.2f} \\\\
|
||
EMA20-50 斜率\\% & {s['regime']['dd_slope']:.3f} & {s['regime']['calm_slope']:.3f} \\\\
|
||
\\bottomrule
|
||
\\end{{tabular}}
|
||
\\end{{table}}
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\begin{{subfigure}}{{0.48\\textwidth}}
|
||
\\includegraphics[width=\\textwidth]{{figures/xau_dd_candles.pdf}}
|
||
\\caption{{回撤谷底附近 K 线 + EMA}}
|
||
\\end{{subfigure}}
|
||
\\hfill
|
||
\\begin{{subfigure}}{{0.48\\textwidth}}
|
||
\\includegraphics[width=\\textwidth]{{figures/xau_calm_candles.pdf}}
|
||
\\caption{{盈利平稳期 K 线 + EMA}}
|
||
\\end{{subfigure}}
|
||
\\caption{{XAUUSD H1 形态对比}}
|
||
\\end{{figure}}
|
||
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.95\\textwidth]{{figures/xau_regime_monthly.pdf}}
|
||
\\caption{{XAUUSD 逐月 ADX / ATR\\% 与组合月度 P/L 对照}}
|
||
\\end{{figure}}
|
||
|
||
\\section{{各机器人逐月问题诊断}}
|
||
{chr(10).join(robots_tex)}
|
||
|
||
\\section{{分品种月度 P/L}}
|
||
\\begin{{figure}}[H]
|
||
\\centering
|
||
\\includegraphics[width=0.95\\textwidth]{{figures/symbol_monthly.pdf}}
|
||
\\caption{{主要交易品种逐月净利润}}
|
||
\\end{{figure}}
|
||
|
||
\\section{{改进路线图}}
|
||
\\begin{{enumerate}}
|
||
\\item \\textbf{{P0 风控：}}设置 ORCH\\_MaxBalanceScale 上限、每机器人 max lot cap；XAUUSD 总 exposure 上限。
|
||
\\item \\textbf{{P1 精简：}}关闭 RRA EURUSD、RS AAPL；SuperEMA 降权；RS NVDA 观察。
|
||
\\item \\textbf{{P2 SimpleTrendline：}}分方向/分品种调参，震荡月（低 ADX + 高 ATR）减半 LOT，勿整体删除。
|
||
\\item \\textbf{{P3 验证：}}固定手数复测 2023--2026；Walk-forward 2026 Q1 作为 OOS。
|
||
\\item \\textbf{{P4 监控：}}月度 dashboard 跟踪 RM EMA Cross 与 ST 多空比、ADX 过滤命中率。
|
||
\\end{{enumerate}}
|
||
|
||
\\appendix
|
||
\\section{{数据来源}}
|
||
报告 HTML: ReportTester-*.html（MT5 导出）；K 线: MT5 XAUUSD H1；成交归因基于 order comment FIFO。
|
||
"""
|
||
|
||
header = textwrap.dedent(r"""
|
||
\documentclass[11pt,a4paper]{ctexart}
|
||
\usepackage{graphicx}
|
||
\usepackage{booktabs}
|
||
\usepackage{geometry}
|
||
\usepackage{float}
|
||
\usepackage{caption}
|
||
\usepackage{subcaption}
|
||
\usepackage{hyperref}
|
||
\usepackage{xcolor}
|
||
\geometry{margin=2.2cm}
|
||
\title{UnitedEA 集群回测综合评估报告\\ \large 2023.07 -- 2026.06 · 逐月诊断 · 回撤期 K 线对比}
|
||
\author{自动生成 · cluster-latest/main.mq5}
|
||
\date{\today}
|
||
|
||
\begin{document}
|
||
\maketitle
|
||
\tableofcontents
|
||
\newpage
|
||
""")
|
||
|
||
return header + body + "\n\\end{document}\n"
|
||
|
||
|
||
def main():
|
||
FIG_DIR.mkdir(parents=True, exist_ok=True)
|
||
DATA_DIR.mkdir(parents=True, exist_ok=True)
|
||
|
||
html = REPORT_HTML.read_text(encoding="utf-16")
|
||
deals = parse_deals(html)
|
||
df = pd.DataFrame(deals).sort_values("time")
|
||
df["month"] = df["time"].dt.to_period("M").astype(str)
|
||
|
||
# Portfolio summary from HTML header
|
||
m_net = re.search(r"总净盈利:</td>\s*<td[^>]*><b>([^<]+)</b>", html)
|
||
m_pf = re.search(r"盈利因子:</td>\s*<td[^>]*><b>([^<]+)</b>", html)
|
||
m_sh = re.search(r"夏普比率:</td>\s*<td[^>]*><b>([^<]+)</b>", html)
|
||
m_dd = re.search(r"相对净值亏损:</td>\s*<td[^>]*><b>([0-9.]+)%", html)
|
||
m_tr = re.search(r"交易总计:</td>\s*<td[^>]*><b>([^<]+)</b>", html)
|
||
m_wr = re.search(r"盈利交易 \(% 全部\):</td>\s*<td[^>]*><b>[0-9]+ \(([0-9.]+)%\)", html)
|
||
|
||
portfolio = {
|
||
"net_profit": parse_num(m_net.group(1)) if m_net else df["net"].sum(),
|
||
"pf": float(m_pf.group(1)) if m_pf else 0,
|
||
"sharpe": float(m_sh.group(1)) if m_sh else 0,
|
||
"max_dd_pct": float(m_dd.group(1)) if m_dd else 0,
|
||
"trades": int(m_tr.group(1).replace(" ", "")) if m_tr else len(df),
|
||
"win_rate": float(m_wr.group(1)) if m_wr else 0,
|
||
}
|
||
|
||
# Equity curve
|
||
eq = df.groupby("time")["balance"].last().sort_index()
|
||
eq_daily = eq.resample("D").last().ffill()
|
||
peak = eq_daily.cummax()
|
||
dd_pct = (eq_daily - peak) / peak * 100
|
||
|
||
fig, axes = plt.subplots(2, 1, figsize=(11, 5), sharex=True, gridspec_kw={"height_ratios": [3, 1]})
|
||
axes[0].semilogy(eq_daily.index, eq_daily.values, color="#1565c0", linewidth=1.2)
|
||
axes[0].set_ylabel("Balance (USD, log)")
|
||
axes[0].set_title("Portfolio Equity")
|
||
axes[0].grid(True, alpha=0.3)
|
||
axes[1].fill_between(dd_pct.index, dd_pct.values, 0, color="#ef5350", alpha=0.5)
|
||
axes[1].set_ylabel("Drawdown %")
|
||
axes[1].set_xlabel("Date")
|
||
axes[1].grid(True, alpha=0.3)
|
||
fig.autofmt_xdate()
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "equity_drawdown.pdf")
|
||
plt.close()
|
||
|
||
drawdowns = find_drawdowns(eq_daily, 5)
|
||
|
||
# Monthly by strategy
|
||
monthly_strat = df.pivot_table(index="month", columns="strategy", values="net", aggfunc="sum", fill_value=0)
|
||
totals = df.groupby("strategy")["net"].sum().sort_values(ascending=False)
|
||
strategy_order = [s for s in totals.index if s != "UNATTRIBUTED"]
|
||
|
||
# Heatmap - top strategies
|
||
top_strats = strategy_order[:14]
|
||
heat = monthly_strat[top_strats].T
|
||
fig, ax = plt.subplots(figsize=(14, 7))
|
||
vmax = np.percentile(np.abs(heat.values), 95)
|
||
im = ax.imshow(heat.values, aspect="auto", cmap="RdYlGn", vmin=-vmax, vmax=vmax)
|
||
ax.set_yticks(range(len(top_strats)))
|
||
ax.set_yticklabels([STRATEGY_META.get(s, {}).get("name", s) for s in top_strats], fontsize=7)
|
||
ax.set_xticks(range(len(heat.columns)))
|
||
ax.set_xticklabels(heat.columns, rotation=45, ha="right", fontsize=7)
|
||
ax.set_title("Monthly Net P/L by Robot (USD)")
|
||
plt.colorbar(im, ax=ax, shrink=0.6)
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "monthly_heatmap.pdf")
|
||
plt.close()
|
||
|
||
# Contribution bar
|
||
fig, ax = plt.subplots(figsize=(9, 5))
|
||
colors = ["#2e7d32" if v >= 0 else "#c62828" for v in totals[strategy_order]]
|
||
ax.barh([STRATEGY_META.get(s, {}).get("name", s)[:28] for s in strategy_order], totals[strategy_order].values, color=colors)
|
||
ax.axvline(0, color="black", linewidth=0.8)
|
||
ax.set_xlabel("Net Profit (USD)")
|
||
ax.set_title("Cumulative Contribution by Robot")
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "strategy_contribution.pdf")
|
||
plt.close()
|
||
|
||
# ST long/short monthly
|
||
st_df = df[df["strategy"] == "ST_SimpleTrendline"].copy()
|
||
st_monthly = {"long": {}, "short": {}}
|
||
for side in ("long", "short"):
|
||
sub = st_df[st_df["side"] == side]
|
||
if len(sub):
|
||
st_monthly[side] = sub.groupby("month")["net"].sum().to_dict()
|
||
all_months = sorted(monthly_strat.index)
|
||
long_s = [st_monthly["long"].get(m, 0) for m in all_months]
|
||
short_s = [st_monthly["short"].get(m, 0) for m in all_months]
|
||
x = np.arange(len(all_months))
|
||
fig, ax = plt.subplots(figsize=(12, 4))
|
||
ax.bar(x - 0.2, long_s, 0.4, label="Long", color="#26a69a")
|
||
ax.bar(x + 0.2, short_s, 0.4, label="Short", color="#ef5350")
|
||
ax.axhline(0, color="black", linewidth=0.6)
|
||
ax.set_xticks(x)
|
||
ax.set_xticklabels(all_months, rotation=45, ha="right", fontsize=7)
|
||
ax.set_ylabel("USD")
|
||
ax.set_title("SimpleTrendline: Long vs Short Monthly P/L")
|
||
ax.legend()
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "st_long_short_monthly.pdf")
|
||
plt.close()
|
||
|
||
# ST by symbol
|
||
st_sym = st_df.groupby(["month", "symbol"])["net"].sum().unstack(fill_value=0)
|
||
if len(st_sym.columns):
|
||
fig, ax = plt.subplots(figsize=(12, 4))
|
||
st_sym.plot(kind="bar", stacked=True, ax=ax, width=0.85)
|
||
ax.axhline(0, color="black", linewidth=0.6)
|
||
ax.set_title("SimpleTrendline Monthly P/L by Symbol")
|
||
ax.set_ylabel("USD")
|
||
plt.xticks(rotation=45, ha="right", fontsize=7)
|
||
ax.legend(fontsize=7, loc="upper left")
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "st_by_symbol_monthly.pdf")
|
||
plt.close()
|
||
|
||
# Top robots monthly lines
|
||
top6 = strategy_order[:6]
|
||
fig, axes = plt.subplots(3, 2, figsize=(12, 7), sharex=True)
|
||
for ax, strat in zip(axes.flat, top6):
|
||
if strat not in monthly_strat.columns:
|
||
continue
|
||
ser = monthly_strat[strat]
|
||
ax.bar(ser.index, ser.values, color=np.where(ser.values >= 0, "#66bb6a", "#ef5350"), width=0.8)
|
||
ax.axhline(0, color="black", linewidth=0.5)
|
||
ax.set_title(STRATEGY_META.get(strat, {}).get("name", strat), fontsize=8)
|
||
ax.tick_params(axis="x", labelrotation=45, labelsize=6)
|
||
fig.suptitle("Top 6 Robots — Monthly P/L", fontsize=11)
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "top6_monthly.pdf")
|
||
plt.close()
|
||
|
||
# Symbol monthly
|
||
sym_m = df.pivot_table(index="month", columns="symbol", values="net", aggfunc="sum", fill_value=0)
|
||
main_syms = sym_m.sum().abs().sort_values(ascending=False).head(6).index
|
||
fig, ax = plt.subplots(figsize=(12, 5))
|
||
bottom = np.zeros(len(sym_m))
|
||
for sym in main_syms:
|
||
ax.bar(sym_m.index, sym_m[sym].values, bottom=bottom, label=sym)
|
||
bottom += sym_m[sym].values
|
||
ax.axhline(0, color="black", linewidth=0.6)
|
||
ax.legend(fontsize=7)
|
||
ax.set_title("Monthly P/L by Symbol (stacked)")
|
||
plt.xticks(rotation=45, ha="right", fontsize=7)
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "symbol_monthly.pdf")
|
||
plt.close()
|
||
|
||
# XAUUSD regime analysis
|
||
start = datetime(2023, 7, 1)
|
||
end = datetime(2026, 6, 4)
|
||
xau = load_xau_h1(start, end)
|
||
xau = market_features(xau)
|
||
xau.to_csv(DATA_DIR / "xau_h1_features.csv")
|
||
|
||
port_monthly = df.groupby("month")["net"].sum()
|
||
xau_m = xau.resample("ME").agg({"adx14": "mean", "atr_pct": "mean", "chop": "mean", "trend_slope": "mean", "close": "last"})
|
||
xau_m.index = xau_m.index.strftime("%Y-%m")
|
||
|
||
# Regime: portfolio drawdown windows vs calm reference (2024-03..2024-09)
|
||
dd_months: set[str] = set()
|
||
for ep in drawdowns[:3]:
|
||
t0, t1 = pd.Timestamp(ep["start"]), pd.Timestamp(ep["trough"])
|
||
for ts in pd.date_range(t0.to_period("M").to_timestamp(), t1.to_period("M").to_timestamp(), freq="MS"):
|
||
dd_months.add(ts.strftime("%Y-%m"))
|
||
calm_months = [m for m in xau_m.index if "2024-0" in m and m >= "2024-03" and m <= "2024-09"]
|
||
dd_mask = xau_m.index.isin(sorted(dd_months))
|
||
calm_mask = xau_m.index.isin(calm_months)
|
||
|
||
common = sorted(set(port_monthly.index) & set(xau_m.index))
|
||
pm = port_monthly.reindex(common, fill_value=0)
|
||
regime = {
|
||
"dd_adx": float(xau_m.loc[dd_mask, "adx14"].mean()) if dd_mask.any() else 0,
|
||
"calm_adx": float(xau_m.loc[calm_mask, "adx14"].mean()) if calm_mask.any() else 0,
|
||
"dd_atr_pct": float(xau_m.loc[dd_mask, "atr_pct"].mean()) if dd_mask.any() else 0,
|
||
"calm_atr_pct": float(xau_m.loc[calm_mask, "atr_pct"].mean()) if calm_mask.any() else 0,
|
||
"dd_chop": float(xau_m.loc[dd_mask, "chop"].mean()) if dd_mask.any() else 0,
|
||
"calm_chop": float(xau_m.loc[calm_mask, "chop"].mean()) if calm_mask.any() else 0,
|
||
"dd_slope": float(xau_m.loc[dd_mask, "trend_slope"].mean()) if dd_mask.any() else 0,
|
||
"calm_slope": float(xau_m.loc[calm_mask, "trend_slope"].mean()) if calm_mask.any() else 0,
|
||
"dd_months": sorted(dd_months),
|
||
}
|
||
|
||
# Regime monthly chart
|
||
fig, ax1 = plt.subplots(figsize=(12, 4))
|
||
ax2 = ax1.twinx()
|
||
xs = np.arange(len(common))
|
||
ax1.bar(xs, pm.values, color=np.where(pm.values >= 0, "#66bb6a", "#ef5350"), alpha=0.7, label="Portfolio P/L")
|
||
ax2.plot(xs, xau_m.reindex(common)["adx14"].values, color="#1565c0", marker="o", markersize=3, label="ADX")
|
||
ax2.plot(xs, xau_m.reindex(common)["atr_pct"].values * 2, color="#ff8f00", marker="s", markersize=3, label="ATR% x2")
|
||
ax1.set_xticks(xs)
|
||
ax1.set_xticklabels(common, rotation=45, ha="right", fontsize=7)
|
||
ax1.set_ylabel("Portfolio Monthly P/L")
|
||
ax2.set_ylabel("ADX / scaled ATR%")
|
||
ax1.set_title("Monthly Portfolio P/L vs XAUUSD Regime")
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "xau_regime_monthly.pdf")
|
||
plt.close()
|
||
|
||
# Candle windows
|
||
if drawdowns:
|
||
trough = drawdowns[0]["trough"]
|
||
dd_win = xau.loc[trough - pd.Timedelta(days=10): trough + pd.Timedelta(days=10)]
|
||
calm_start = pd.Timestamp("2024-03-01")
|
||
calm_win = xau.loc[calm_start: calm_start + pd.Timedelta(days=10)]
|
||
fig, ax = plt.subplots(figsize=(8, 3.5))
|
||
plot_candles(ax, dd_win, f"Drawdown trough window ~ {trough.strftime('%Y-%m-%d')}")
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "xau_dd_candles.pdf")
|
||
plt.close()
|
||
fig, ax = plt.subplots(figsize=(8, 3.5))
|
||
plot_candles(ax, calm_win, "Calm period reference (2024-03)")
|
||
fig.tight_layout()
|
||
fig.savefig(FIG_DIR / "xau_calm_candles.pdf")
|
||
plt.close()
|
||
|
||
# Issues per strategy
|
||
issues = {}
|
||
monthly_dict = {}
|
||
for strat in strategy_order:
|
||
if strat in monthly_strat.columns:
|
||
ser = monthly_strat[strat]
|
||
monthly_dict[strat] = ser.to_dict()
|
||
issues[strat] = monthly_issue_text(strat, ser, STRATEGY_META.get(strat, {}))
|
||
|
||
summary = {
|
||
"portfolio": portfolio,
|
||
"drawdowns": drawdowns,
|
||
"strategy_order": strategy_order,
|
||
"totals": totals.to_dict(),
|
||
"monthly": monthly_dict,
|
||
"issues": issues,
|
||
"st_monthly": st_monthly,
|
||
"regime": regime,
|
||
}
|
||
(DATA_DIR / "summary.json").write_text(json.dumps(summary, default=str, indent=2), encoding="utf-8")
|
||
|
||
tex = build_latex(summary)
|
||
TEX_FILE.write_text(tex, encoding="utf-8")
|
||
print(f"Wrote {TEX_FILE}")
|
||
print(f"Figures in {FIG_DIR}")
|
||
for _ in range(2):
|
||
subprocess.run(
|
||
["xelatex", "-interaction=nonstopmode", "cluster_evaluation.tex"],
|
||
cwd=ROOT,
|
||
check=False,
|
||
capture_output=True,
|
||
)
|
||
pdf = ROOT / "cluster_evaluation.pdf"
|
||
if pdf.exists():
|
||
print(f"PDF: {pdf}")
|
||
return summary
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|