ランダムウォークへの挑戦⑲｜BAND_DISTANCEをATRで割って「相対的なブレイクの強さ」を分析する（2026/8/19）

import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt


# ============================================================
# 設定
# ============================================================

INPUT_FILE = "results/trade_base.csv"

OUTPUT_DIR = "results"

os.makedirs(
    OUTPUT_DIR,
    exist_ok=True
)


OUTPUT_ANALYSIS_CSV = os.path.join(
    OUTPUT_DIR,
    "relative_band_distance_analysis.csv"
)

OUTPUT_RANGES_CSV = os.path.join(
    OUTPUT_DIR,
    "relative_band_distance_ranges.csv"
)

OUTPUT_CORRELATION_CSV = os.path.join(
    OUTPUT_DIR,
    "band_atr_correlation.csv"
)

OUTPUT_CROSS_COUNT_CSV = os.path.join(
    OUTPUT_DIR,
    "band_atr_cross_count.csv"
)

OUTPUT_CROSS_RATIO_CSV = os.path.join(
    OUTPUT_DIR,
    "band_atr_cross_ratio.csv"
)

OUTPUT_HISTOGRAM = os.path.join(
    OUTPUT_DIR,
    "relative_band_distance_histogram.png"
)

OUTPUT_PF_GRAPH = os.path.join(
    OUTPUT_DIR,
    "relative_band_distance_pf.png"
)

OUTPUT_AVG_PROFIT_GRAPH = os.path.join(
    OUTPUT_DIR,
    "relative_band_distance_average_profit.png"
)


# ============================================================
# データ読み込み
# ============================================================

df = pd.read_csv(
    INPUT_FILE
)

print(
    "========================================"
)

print(
    "RELATIVE BAND DISTANCE ANALYSIS"
)

print(
    "========================================"
)

print(
    f"\n読み込み件数: {len(df)}"
)


# ============================================================
# 必要列確認
# ============================================================

required_columns = [
    "entry_band_distance_pips",
    "entry_atr14_pips",
    "profit"
]

missing_columns = [
    column
    for column in required_columns
    if column not in df.columns
]

if missing_columns:
    raise ValueError(
        "必要な列がありません: "
        + ", ".join(missing_columns)
    )


# ============================================================
# 数値型へ変換
# ============================================================

for column in required_columns:

    df[column] = pd.to_numeric(
        df[column],
        errors="coerce"
    )


# ============================================================
# 分析対象データ
# ============================================================

analysis_df = df[
    df["entry_band_distance_pips"].notna()
    &
    df["entry_atr14_pips"].notna()
    &
    df["profit"].notna()
    &
    (
        df["entry_atr14_pips"]
        > 0
    )
].copy()


print(
    f"分析対象件数: {len(analysis_df)}"
)


if analysis_df.empty:

    raise ValueError(
        "分析対象データがありません。"
    )


# ============================================================
# Relative Band Distance
# ============================================================

analysis_df[
    "relative_band_distance"
] = (
    analysis_df[
        "entry_band_distance_pips"
    ]
    /
    analysis_df[
        "entry_atr14_pips"
    ]
)


# ============================================================
# 無限大等を除外
# ============================================================

analysis_df[
    "relative_band_distance"
] = (
    analysis_df[
        "relative_band_distance"
    ]
    .replace(
        [
            np.inf,
            -np.inf
        ],
        np.nan
    )
)


analysis_df = analysis_df[
    analysis_df[
        "relative_band_distance"
    ]
    .notna()
].copy()


print(
    "Relative Band Distance計算後:",
    len(analysis_df)
)


# ============================================================
# 基本統計量
# ============================================================

print(
    "\n========================================"
)

print(
    "RELATIVE BAND DISTANCE 基本統計量"
)

print(
    "========================================"
)


relative_description = (
    analysis_df[
        "relative_band_distance"
    ]
    .describe()
)


print(
    relative_description
)


# ============================================================
# BAND_DISTANCEとATRの相関
# ============================================================

band_atr_corr = (
    analysis_df[
        [
            "entry_band_distance_pips",
            "entry_atr14_pips"
        ]
    ]
    .corr()
    .iloc[
        0,
        1
    ]
)


# ============================================================
# Relative Band Distanceとprofitの相関
# ============================================================

relative_profit_corr = (
    analysis_df[
        [
            "relative_band_distance",
            "profit"
        ]
    ]
    .corr()
    .iloc[
        0,
        1
    ]
)


# ============================================================
# BAND_DISTANCEとprofit
# ============================================================

band_profit_corr = (
    analysis_df[
        [
            "entry_band_distance_pips",
            "profit"
        ]
    ]
    .corr()
    .iloc[
        0,
        1
    ]
)


# ============================================================
# ATRとprofit
# ============================================================

atr_profit_corr = (
    analysis_df[
        [
            "entry_atr14_pips",
            "profit"
        ]
    ]
    .corr()
    .iloc[
        0,
        1
    ]
)


# ============================================================
# 相関結果
# ============================================================

correlation_df = pd.DataFrame(
    {
        "correlation": [
            band_atr_corr,
            relative_profit_corr,
            band_profit_corr,
            atr_profit_corr
        ]
    },
    index=[
        "band_distance_vs_atr",
        "relative_band_distance_vs_profit",
        "band_distance_vs_profit",
        "atr_vs_profit"
    ]
)


print(
    "\n========================================"
)

print(
    "相関係数"
)

print(
    "========================================"
)

print(
    correlation_df.round(6)
)


correlation_df.to_csv(
    OUTPUT_CORRELATION_CSV,
    encoding="utf-8-sig"
)


# ============================================================
# ヒストグラム
# ============================================================

fig, ax = plt.subplots(
    figsize=(9, 6)
)


ax.hist(
    analysis_df[
        "relative_band_distance"
    ],
    bins=40
)


ax.set_xlabel(
    "Relative Band Distance"
)


ax.set_ylabel(
    "Number of Trades"
)


ax.set_title(
    "Distribution of Relative Band Distance"
)


ax.grid(
    axis="y",
    alpha=0.3
)


plt.tight_layout()


plt.savefig(
    OUTPUT_HISTOGRAM,
    dpi=150
)


plt.close()


# ============================================================
# 四分位に分類
# ============================================================

analysis_df[
    "relative_band_quartile_no"
] = pd.qcut(
    analysis_df[
        "relative_band_distance"
    ],
    q=4,
    labels=False,
    duplicates="drop"
)


analysis_df[
    "relative_band_group"
] = (
    analysis_df[
        "relative_band_quartile_no"
    ]
    .apply(
        lambda x:
            f"Q{int(x) + 1}"
            if pd.notna(x)
            else np.nan
    )
)


quartile_order = [
    "Q1",
    "Q2",
    "Q3",
    "Q4"
]


# ============================================================
# 四分位ごとの範囲
# ============================================================

relative_range = (
    analysis_df
    .groupby(
        "relative_band_group",
        observed=True
    )
    [
        "relative_band_distance"
    ]
    .agg(
        min_relative_distance="min",
        max_relative_distance="max",
        mean_relative_distance="mean",
        median_relative_distance="median",
        trades="count"
    )
)


relative_range = (
    relative_range
    .reindex(
        quartile_order
    )
)


print(
    "\n========================================"
)

print(
    "RELATIVE BAND DISTANCE 四分位"
)

print(
    "========================================"
)


print(
    relative_range
)


# ============================================================
# 勝敗判定
# ============================================================

analysis_df[
    "is_win"
] = (
    analysis_df[
        "profit"
    ]
    > 0
)


# ============================================================
# 基本成績
# ============================================================

summary = (
    analysis_df
    .groupby(
        "relative_band_group",
        observed=True
    )
    .agg(
        trades=(
            "profit",
            "count"
        ),
        total_profit=(
            "profit",
            "sum"
        ),
        average_profit=(
            "profit",
            "mean"
        ),
        median_profit=(
            "profit",
            "median"
        ),
        win_rate=(
            "is_win",
            "mean"
        )
    )
)


summary[
    "win_rate"
] = (
    summary[
        "win_rate"
    ]
    * 100
)


# ============================================================
# 勝ち数
# ============================================================

wins = (
    analysis_df[
        analysis_df[
            "profit"
        ]
        > 0
    ]
    .groupby(
        "relative_band_group",
        observed=True
    )
    .size()
)


# ============================================================
# 負け数
# ============================================================

losses = (
    analysis_df[
        analysis_df[
            "profit"
        ]
        < 0
    ]
    .groupby(
        "relative_band_group",
        observed=True
    )
    .size()
)


summary[
    "wins"
] = wins


summary[
    "losses"
] = losses


summary[
    "wins"
] = (
    summary[
        "wins"
    ]
    .fillna(0)
    .astype(int)
)


summary[
    "losses"
] = (
    summary[
        "losses"
    ]
    .fillna(0)
    .astype(int)
)


# ============================================================
# 総利益
# ============================================================

gross_profit = (
    analysis_df[
        analysis_df[
            "profit"
        ]
        > 0
    ]
    .groupby(
        "relative_band_group",
        observed=True
    )
    [
        "profit"
    ]
    .sum()
)


# ============================================================
# 総損失
# ============================================================

gross_loss = (
    analysis_df[
        analysis_df[
            "profit"
        ]
        < 0
    ]
    .groupby(
        "relative_band_group",
        observed=True
    )
    [
        "profit"
    ]
    .sum()
    .abs()
)


summary[
    "gross_profit"
] = (
    gross_profit
)


summary[
    "gross_loss"
] = (
    gross_loss
)


summary[
    "gross_profit"
] = (
    summary[
        "gross_profit"
    ]
    .fillna(0)
)


summary[
    "gross_loss"
] = (
    summary[
        "gross_loss"
    ]
    .fillna(0)
)


# ============================================================
# PF
# ============================================================

summary[
    "profit_factor"
] = np.where(
    summary[
        "gross_loss"
    ]
    > 0,
    summary[
        "gross_profit"
    ]
    /
    summary[
        "gross_loss"
    ],
    np.nan
)


# ============================================================
# Q1～Q4順に整理
# ============================================================

summary = (
    summary
    .reindex(
        quartile_order
    )
)


summary = summary[
    [
        "trades",
        "wins",
        "losses",
        "win_rate",
        "total_profit",
        "average_profit",
        "median_profit",
        "gross_profit",
        "gross_loss",
        "profit_factor"
    ]
]


# ============================================================
# 小数点整理
# ============================================================

summary[
    "win_rate"
] = (
    summary[
        "win_rate"
    ]
    .round(2)
)


for column in [
    "total_profit",
    "average_profit",
    "median_profit",
    "gross_profit",
    "gross_loss"
]:

    summary[column] = (
        summary[column]
        .round(2)
    )


summary[
    "profit_factor"
] = (
    summary[
        "profit_factor"
    ]
    .round(3)
)


for column in [
    "min_relative_distance",
    "max_relative_distance",
    "mean_relative_distance",
    "median_relative_distance"
]:

    relative_range[column] = (
        relative_range[column]
        .round(4)
    )


# ============================================================
# 結果表示
# ============================================================

print(
    "\n========================================"
)

print(
    "RELATIVE BAND DISTANCE別 取引成績"
)

print(
    "========================================"
)


print(
    summary
)


# ============================================================
# CSV保存
# ============================================================

relative_range.to_csv(
    OUTPUT_RANGES_CSV,
    encoding="utf-8-sig"
)


summary.to_csv(
    OUTPUT_ANALYSIS_CSV,
    encoding="utf-8-sig"
)


# ============================================================
# PFグラフ
# ============================================================

fig, ax = plt.subplots(
    figsize=(8, 6)
)


summary[
    "profit_factor"
].plot(
    kind="bar",
    ax=ax
)


ax.axhline(
    y=1.0,
    linestyle="--",
    linewidth=1.5
)


ax.set_xlabel(
    "Relative Band Distance Quartile"
)


ax.set_ylabel(
    "Profit Factor"
)


ax.set_title(
    "Profit Factor by Relative Band Distance"
)


ax.tick_params(
    axis="x",
    rotation=0
)


ax.grid(
    axis="y",
    alpha=0.3
)


plt.tight_layout()


plt.savefig(
    OUTPUT_PF_GRAPH,
    dpi=150
)


plt.close()


# ============================================================
# 平均損益グラフ
# ============================================================

fig, ax = plt.subplots(
    figsize=(8, 6)
)


summary[
    "average_profit"
].plot(
    kind="bar",
    ax=ax
)


ax.axhline(
    y=0,
    linestyle="--",
    linewidth=1.5
)


ax.set_xlabel(
    "Relative Band Distance Quartile"
)


ax.set_ylabel(
    "Average Profit"
)


ax.set_title(
    "Average Profit by Relative Band Distance"
)


ax.tick_params(
    axis="x",
    rotation=0
)


ax.grid(
    axis="y",
    alpha=0.3
)


plt.tight_layout()


plt.savefig(
    OUTPUT_AVG_PROFIT_GRAPH,
    dpi=150
)


plt.close()


# ============================================================
# BAND_DISTANCEを四分位化
# ============================================================

analysis_df[
    "band_group"
] = pd.qcut(
    analysis_df[
        "entry_band_distance_pips"
    ],
    q=4,
    labels=quartile_order
)


# ============================================================
# ATRを四分位化
# ============================================================

analysis_df[
    "atr_group"
] = pd.qcut(
    analysis_df[
        "entry_atr14_pips"
    ],
    q=4,
    labels=quartile_order
)


# ============================================================
# BAND_DISTANCE × ATR 件数
# ============================================================

cross_count = pd.crosstab(
    analysis_df[
        "band_group"
    ],
    analysis_df[
        "atr_group"
    ]
)


cross_count = (
    cross_count
    .reindex(
        index=quartile_order,
        columns=quartile_order
    )
)


print(
    "\n========================================"
)

print(
    "BAND_DISTANCE × ATR 取引件数"
)

print(
    "========================================"
)


print(
    cross_count
)


# ============================================================
# BAND_DISTANCE × ATR 構成比
# ============================================================

cross_ratio = (
    pd.crosstab(
        analysis_df[
            "band_group"
        ],
        analysis_df[
            "atr_group"
        ],
        normalize="index"
    )
    * 100
)


cross_ratio = (
    cross_ratio
    .reindex(
        index=quartile_order,
        columns=quartile_order
    )
)


cross_ratio = (
    cross_ratio
    .round(2)
)


print(
    "\n========================================"
)

print(
    "BAND_DISTANCE × ATR 構成比（％）"
)

print(
    "========================================"
)


print(
    cross_ratio
)


cross_count.to_csv(
    OUTPUT_CROSS_COUNT_CSV,
    encoding="utf-8-sig"
)


cross_ratio.to_csv(
    OUTPUT_CROSS_RATIO_CSV,
    encoding="utf-8-sig"
)


# ============================================================
# 特にBAND_DISTANCE Q4のATR構成を表示
# ============================================================

if "Q4" in cross_ratio.index:

    print(
        "\n========================================"
    )

    print(
        "BAND_DISTANCE Q4 のATR構成"
    )

    print(
        "========================================"
    )

    print(
        cross_ratio.loc[
            "Q4"
        ]
    )


# ============================================================
# 全体成績
# ============================================================

total_trades = len(
    analysis_df
)


total_wins = (
    analysis_df[
        "profit"
    ]
    > 0
).sum()


total_losses = (
    analysis_df[
        "profit"
    ]
    < 0
).sum()


total_profit = (
    analysis_df[
        "profit"
    ]
    .sum()
)


overall_win_rate = (
    total_wins
    /
    total_trades
    * 100
)


overall_gross_profit = (
    analysis_df.loc[
        analysis_df[
            "profit"
        ]
        > 0,
        "profit"
    ]
    .sum()
)


overall_gross_loss = abs(
    analysis_df.loc[
        analysis_df[
            "profit"
        ]
        < 0,
        "profit"
    ]
    .sum()
)


if overall_gross_loss > 0:

    overall_pf = (
        overall_gross_profit
        /
        overall_gross_loss
    )

else:

    overall_pf = np.nan


print(
    "\n========================================"
)

print(
    "分析対象取引 全体成績"
)

print(
    "========================================"
)


print(
    f"取引回数 : {total_trades}"
)


print(
    f"勝ち     : {total_wins}"
)


print(
    f"負け     : {total_losses}"
)


print(
    f"勝率     : {overall_win_rate:.2f}%"
)


print(
    f"総損益   : {total_profit:.2f}"
)


print(
    f"PF       : {overall_pf:.3f}"
)


# ============================================================
# 出力ファイル
# ============================================================

print(
    "\n========================================"
)

print(
    "分析完了"
)

print(
    "========================================"
)


print(
    "\n出力ファイル:"
)


print(
    OUTPUT_ANALYSIS_CSV
)


print(
    OUTPUT_RANGES_CSV
)


print(
    OUTPUT_CORRELATION_CSV
)


print(
    OUTPUT_CROSS_COUNT_CSV
)


print(
    OUTPUT_CROSS_RATIO_CSV
)


print(
    OUTPUT_HISTOGRAM
)


print(
    OUTPUT_PF_GRAPH
)


print(
    OUTPUT_AVG_PROFIT_GRAPH
)