source .venv/bin/activate.fish

summary = {
    "annual_return": annual_return(net_ret),
    "annual_volatility": annual_volatility(net_ret),
    "sharpe": sharpe_ratio(net_ret),
    "max_drawdown": max_drawdown(net_ret),
}
print(summary)

quant-backtest-mini/
├── data/
│   └── prices.csv
├── figures/
│   ├── equity_curve.png
│   ├── drawdown.png
│   ├── split_equity.png
│   └── cost_sensitivity.png
├── reports/
│   ├── week15_backtest_daily_results.csv
│   ├── week15_backtest_notes.md
│   ├── week16_metrics_by_split.csv
│   ├── week16_cost_sensitivity.csv
│   └── week16_quant_project_report.md
├── src/
│   └── quant_backtest_mini/
│       ├── __init__.py
│       ├── backtest.py
│       ├── metrics.py
│       └── report.py
├── README.md
└── pyproject.toml

cd ~/Code/ustc-stat-ai-quant/quant-backtest-mini
source .venv/bin/activate.fish

python -c "import pandas as pd; import numpy as np; import matplotlib; print('report env ok')"

from pathlib import Path
import matplotlib.pyplot as plt
import pandas as pd
from quant_backtest_mini.backtest import load_prices, run_backtest
from quant_backtest_mini.metrics import drawdown, equity_curve, summarize
FIGURE_DIR = Path(“figures”)
REPORT_DIR = Path(“reports”)
FIGURE_DIR.mkdir(exist_ok=True)
REPORT_DIR.mkdir(exist_ok=True)
def split_periods(result: pd.DataFrame) -> dict[str, pd.DataFrame]:
n = len(result)
train_end = int(n * 0.6)
test_start = int(n * 0.8)
return {
“in_sample”: result.iloc[
],
“validation”: result.iloc[train_end
],
“out_of_sample”: result.iloc[test_start:],
“full_period”: result,
}

def metrics_by_split(result: pd.DataFrame) -> pd.DataFrame:
rows = []
for name, part in split_periods(result).items():
strategy = summarize(part[“net_strategy_return”], part[“turnover”])
benchmark = summarize(part[“benchmark_return”], pd.Series(0, index=part.index))
rows.append({“period”: name, “portfolio”: “strategy”, **strategy})
rows.append({“period”: name, “portfolio”: “benchmark”, **benchmark})
return pd.DataFrame(rows)
def cost_sensitivity(prices: pd.DataFrame, lookback: int) -> pd.DataFrame:
rows = []
for cost_bps in [0, 5, 10, 20, 50]:
result = run_backtest(prices, lookback=lookback, cost_bps=cost_bps)
summary = summarize(result[“net_strategy_return”], result[“turnover”])
rows.append({“cost_bps”: cost_bps, **summary})
return pd.DataFrame(rows)
def plot_split_equity(result: pd.DataFrame) -> None:
fig, ax = plt.subplots(figsize=(10, 5))
for name, part in split_periods(result).items():
if name == “full_period”:
continue
equity_curve(part[“net_strategy_return”]).plot(ax=ax, label=name)
ax.set_title(“Strategy Equity by Split”)
ax.set_ylabel(“growth of 1 within split”)
ax.legend()
fig.tight_layout()
fig.savefig(FIGURE_DIR / “split_equity.png”, dpi=150)
plt.close(fig)
def plot_cost_sensitivity(table: pd.DataFrame) -> None:
fig, axes = plt.subplots(2, 1, figsize=(10, 7), sharex=True)
axes[0].plot(table[“cost_bps”], table[“annual_return”], marker=“o”)
axes[0].set_title(“Annual Return vs Transaction Cost”)
axes[0].set_ylabel(“annual return”)
plain axes[1].plot(table[“cost_bps”], table[“sharpe_ratio”], marker=“o”)
axes[1].set_title(“Sharpe vs Transaction Cost”)
axes[1].set_xlabel(“cost bps”)
axes[1].set_ylabel(“sharpe”)
plain fig.tight_layout()
fig.savefig(FIGURE_DIR / “cost_sensitivity.png”, dpi=150)
plt.close(fig)
def main() -> None:
prices = load_prices()
result = run_backtest(prices, lookback=20, cost_bps=5)
plain split_metrics = metrics_by_split(result)
split_metrics.to_csv(REPORT_DIR / “week16_metrics_by_split.csv”, index=False)
print(split_metrics)
plain cost_table = cost_sensitivity(prices, lookback=20)
cost_table.to_csv(REPORT_DIR / “week16_cost_sensitivity.csv”, index=False)
print(cost_table)
plain plot_split_equity(result)
plot_cost_sensitivity(cost_table)
print(“saved Week 16 report tables and figures”)
if name == “main”:
main()

python -m quant_backtest_mini.report

PYTHONPATH=src python src/quant_backtest_mini/report.py

python -c "import pandas as pd; print(pd.read_csv('reports/week16_metrics_by_split.csv').round(4))"

样本内策略 Sharpe 高于 benchmark，但样本外 Sharpe 明显下降，说明策略可能依赖前期市场结构。样本外最大回撤接近 benchmark，没有明显风险优势。因此不能认为该动量规则已经稳定有效。

def drawdown_details(daily_returns: pd.Series) -> dict[str, object]:
    equity = equity_curve(daily_returns)
    dd = drawdown(equity)
    trough = dd.idxmin()
    peak = equity.loc[:trough].idxmax()
    recovery_candidates = equity.loc[trough:][equity.loc[trough:] >= equity.loc[peak]]
    recovery = recovery_candidates.index[0] if not recovery_candidates.empty else None
    return {
        "peak": peak,
        "trough": trough,
        "recovery": recovery,
        "max_drawdown": float(dd.loc[trough]),
    }

策略最大回撤从 YYYY-MM-DD 的净值高点开始，到 YYYY-MM-DD 达到最低点。样本结束前是否恢复：是/否。这个回撤说明策略在连续震荡或趋势反转时承压明显。

in_sample：用于观察和开发策略
validation：用于比较参数或模型
out_of_sample：最终检验，不反复调参

全样本收益很好，所以策略有效。

全样本表现不能单独证明策略有效。样本外表现更接近未来未知数据，因此本文主要依据 out_of_sample 指标判断策略稳定性。

样本外年化收益和 Sharpe 均下降，说明策略可能对样本内走势过拟合，或该动量规则在后期市场状态中不适用。

python -c "import pandas as pd; print(pd.read_csv('reports/week16_cost_sensitivity.csv').round(4))"

xdg-open figures/cost_sensitivity.png

策略在无成本条件下略优于 benchmark，但随着成本从 0 bps 增加到 20 bps，年化收益和 Sharpe 快速下降，说明该策略对交易成本敏感，不适合高频换仓或高摩擦市场。

code reports/week16_quant_project_report.md

# Mini Quant Backtest Report
1. Research question
本文研究一个简单 N 日动量规则是否能在样本外获得稳定风险调整收益。
2. Data

标的：
数据来源：
日期范围：
频率：日频
缺失值处理：

3. Strategy rule

过去 N 日收益为正：持有
否则：空仓
t 日收盘生成信号，t+1 日持仓

4. Backtest assumptions

初始资金：1
交易成本：5 bps
不考虑滑点
不考虑停牌、涨跌停、成交量限制
benchmark：buy and hold

5. Sample split

in_sample：
validation：
out_of_sample：

6. Metrics
插入 metrics_by_split 表格，并解释 annual return、volatility、Sharpe、max drawdown、turnover。
7. Equity and drawdown
引用 equity_curve、drawdown、split_equity 图，并解释最大回撤时间段。
8. Cost sensitivity
引用 cost_sensitivity 表格和图片。
9. Findings

样本内是否有效？
样本外是否有效？
是否优于 benchmark？
是否对成本敏感？

10. Limitations

数据样本有限
策略过于简单
未考虑真实交易滑点和流动性
未考虑参数挖掘惩罚
未进行多标的验证

11. Conclusion
用谨慎语言判断：可以继续研究、需要改进，或暂不认为有效。

本回测只使用单一标的日频数据，无法证明策略在其它市场或更长周期中有效。交易成本仅用固定 bps 近似，没有考虑滑点、成交量、涨跌停和资金容量，因此实盘表现可能显著低于回测。

如果以后优化一下，策略肯定会更好。

![Equity Curve](../figures/equity_curve.png)
![Drawdown](../figures/drawdown.png)
![Cost Sensitivity](../figures/cost_sensitivity.png)

![Equity Curve](figures/equity_curve.png)

pwd
ls reports figures src/quant_backtest_mini

reports/week16_metrics_by_split.csv
reports/week16_cost_sensitivity.csv
reports/week16_quant_project_report.md
figures/split_equity.png
figures/cost_sensitivity.png
src/quant_backtest_mini/report.py

source .venv/bin/activate.fish
python -m quant_backtest_mini.report
python -c "from pathlib import Path; required = ['reports/week16_metrics_by_split.csv', 'reports/week16_cost_sensitivity.csv', 'reports/week16_quant_project_report.md', 'figures/split_equity.png', 'figures/cost_sensitivity.png']; missing = [p for p in required if not Path(p).exists()]; print('missing:', missing); raise SystemExit(1 if missing else 0)"