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_CSV = os.path.join(
    OUTPUT_DIR,
    "relative_band_threshold_analysis.csv"
)

OUTPUT_FILTERED_CSV = os.path.join(
    OUTPUT_DIR,
    "relative_band_threshold_filtered.csv"
)

OUTPUT_HEATMAP = os.path.join(
    OUTPUT_DIR,
    "relative_band_pf_heatmap.png"
)

OUTPUT_LOWER_GRAPH = os.path.join(
    OUTPUT_DIR,
    "relative_band_lower_sensitivity.png"
)

OUTPUT_UPPER_GRAPH = os.path.join(
    OUTPUT_DIR,
    "relative_band_upper_sensitivity.png"
)

OUTPUT_TOP_CSV = os.path.join(
    OUTPUT_DIR,
    "relative_band_threshold_top.csv"
)


# 元Q3のおおよその境界
BASE_LOWER = 0.44
BASE_UPPER = 0.88


# 探索範囲
LOWER_VALUES = np.round(
    np.arange(
        0.30,
        0.62,
        0.02
    ),
    2
)

UPPER_VALUES = np.round(
    np.arange(
        0.70,
        1.12,
        0.02
    ),
    2
)


# 最低取引数
MIN_TRADES = 100


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

df = pd.read_csv(
    INPUT_FILE
)


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

print(
    "RELATIVE BAND THRESHOLD 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()


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"
    ]
)


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(
    f"分析対象件数: {len(analysis_df)}"
)


print(
    "\nRelative Band Distance 基本統計量"
)

print(
    analysis_df[
        "relative_band_distance"
    ].describe()
)


# ============================================================
# 集計関数
# ============================================================

def calculate_metrics(
    selected,
    lower,
    upper
):

    trade_count = len(
        selected
    )

    if trade_count == 0:

        return {
            "lower": lower,
            "upper": upper,
            "trade_count": 0,
            "win_rate": np.nan,
            "total_profit": np.nan,
            "average_profit": np.nan,
            "median_profit": np.nan,
            "gross_profit": np.nan,
            "gross_loss": np.nan,
            "profit_factor": np.nan
        }


    win_count = (
        selected[
            "profit"
        ]
        > 0
    ).sum()


    win_rate = (
        win_count
        /
        trade_count
        *
        100
    )


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


    average_profit = (
        selected[
            "profit"
        ].mean()
    )


    median_profit = (
        selected[
            "profit"
        ].median()
    )


    gross_profit = (
        selected.loc[
            selected[
                "profit"
            ]
            > 0,
            "profit"
        ]
        .sum()
    )


    gross_loss = -(
        selected.loc[
            selected[
                "profit"
            ]
            < 0,
            "profit"
        ]
        .sum()
    )


    if gross_loss > 0:

        profit_factor = (
            gross_profit
            /
            gross_loss
        )

    else:

        profit_factor = np.nan


    return {
        "lower": lower,
        "upper": upper,
        "trade_count": trade_count,
        "win_rate": win_rate,
        "total_profit": total_profit,
        "average_profit": average_profit,
        "median_profit": median_profit,
        "gross_profit": gross_profit,
        "gross_loss": gross_loss,
        "profit_factor": profit_factor
    }


# ============================================================
# 全組み合わせを分析
# ============================================================

results = []


for lower in LOWER_VALUES:

    for upper in UPPER_VALUES:

        if lower >= upper:

            continue


        mask = (
            (
                analysis_df[
                    "relative_band_distance"
                ]
                >= lower
            )
            &
            (
                analysis_df[
                    "relative_band_distance"
                ]
                <= upper
            )
        )


        selected = analysis_df[
            mask
        ]


        metrics = calculate_metrics(
            selected,
            lower,
            upper
        )


        results.append(
            metrics
        )


result_df = pd.DataFrame(
    results
)


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

result_df.to_csv(
    OUTPUT_CSV,
    index=False,
    encoding="utf-8-sig"
)


print(
    f"\n保存: {OUTPUT_CSV}"
)


# ============================================================
# 最低取引数以上に限定
# ============================================================

filtered_df = result_df[
    result_df[
        "trade_count"
    ]
    >= MIN_TRADES
].copy()


filtered_df.to_csv(
    OUTPUT_FILTERED_CSV,
    index=False,
    encoding="utf-8-sig"
)


print(
    f"保存: {OUTPUT_FILTERED_CSV}"
)


# ============================================================
# PF上位条件
# ============================================================

top_df = (
    filtered_df
    .sort_values(
        "profit_factor",
        ascending=False
    )
    .head(
        30
    )
    .copy()
)


top_df.to_csv(
    OUTPUT_TOP_CSV,
    index=False,
    encoding="utf-8-sig"
)


print(
    f"保存: {OUTPUT_TOP_CSV}"
)


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

print(
    f"PF上位条件"
    f"（取引数{MIN_TRADES}以上）"
)

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


print(
    top_df[
        [
            "lower",
            "upper",
            "trade_count",
            "win_rate",
            "total_profit",
            "median_profit",
            "profit_factor"
        ]
    ]
    .head(
        20
    )
    .to_string(
        index=False
    )
)


# ============================================================
# 元Q3相当条件
# ============================================================

base_mask = (
    (
        analysis_df[
            "relative_band_distance"
        ]
        >= BASE_LOWER
    )
    &
    (
        analysis_df[
            "relative_band_distance"
        ]
        <= BASE_UPPER
    )
)


base_df = analysis_df[
    base_mask
]


base_metrics = calculate_metrics(
    base_df,
    BASE_LOWER,
    BASE_UPPER
)


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

print(
    "元Q3相当条件"
)

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


for key, value in base_metrics.items():

    print(
        f"{key}: {value}"
    )


# ============================================================
# PF 1以上 / 1.2以上 / 1.3以上の条件数
# ============================================================

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

print(
    "安定領域確認"
)

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


valid_count = len(
    filtered_df
)


pf_100_count = (
    filtered_df[
        "profit_factor"
    ]
    >= 1.0
).sum()


pf_120_count = (
    filtered_df[
        "profit_factor"
    ]
    >= 1.2
).sum()


pf_130_count = (
    filtered_df[
        "profit_factor"
    ]
    >= 1.3
).sum()


print(
    f"評価対象組み合わせ数: "
    f"{valid_count}"
)

print(
    f"PF >= 1.0: "
    f"{pf_100_count}"
)

print(
    f"PF >= 1.2: "
    f"{pf_120_count}"
)

print(
    f"PF >= 1.3: "
    f"{pf_130_count}"
)


if valid_count > 0:

    print(
        f"PF1以上割合: "
        f"{pf_100_count / valid_count * 100:.1f}%"
    )

    print(
        f"PF1.2以上割合: "
        f"{pf_120_count / valid_count * 100:.1f}%"
    )

    print(
        f"PF1.3以上割合: "
        f"{pf_130_count / valid_count * 100:.1f}%"
    )


# ============================================================
# Q3近傍だけを確認
# ============================================================

near_q3_df = filtered_df[
    (
        filtered_df[
            "lower"
        ]
        >= 0.38
    )
    &
    (
        filtered_df[
            "lower"
        ]
        <= 0.50
    )
    &
    (
        filtered_df[
            "upper"
        ]
        >= 0.82
    )
    &
    (
        filtered_df[
            "upper"
        ]
        <= 0.94
    )
].copy()


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

print(
    "Q3近傍"
)

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


print(
    near_q3_df[
        [
            "lower",
            "upper",
            "trade_count",
            "total_profit",
            "median_profit",
            "profit_factor"
        ]
    ]
    .sort_values(
        [
            "lower",
            "upper"
        ]
    )
    .to_string(
        index=False
    )
)


# ============================================================
# ヒートマップ用データ
# ============================================================

heatmap_df = (
    filtered_df
    .pivot(
        index="lower",
        columns="upper",
        values="profit_factor"
    )
    .sort_index(
        ascending=False
    )
)


# ============================================================
# PFヒートマップ
# ============================================================

fig, ax = plt.subplots(
    figsize=(
        14,
        9
    )
)


image = ax.imshow(
    heatmap_df.values,
    aspect="auto",
    interpolation="nearest"
)


ax.set_title(
    "Profit Factor Heatmap\n"
    "Relative Band Distance Threshold"
)

ax.set_xlabel(
    "Upper Threshold"
)

ax.set_ylabel(
    "Lower Threshold"
)


ax.set_xticks(
    np.arange(
        len(
            heatmap_df.columns
        )
    )
)

ax.set_xticklabels(
    [
        f"{value:.2f}"
        for value
        in heatmap_df.columns
    ],
    rotation=90
)


ax.set_yticks(
    np.arange(
        len(
            heatmap_df.index
        )
    )
)

ax.set_yticklabels(
    [
        f"{value:.2f}"
        for value
        in heatmap_df.index
    ]
)


colorbar = fig.colorbar(
    image,
    ax=ax
)

colorbar.set_label(
    "Profit Factor"
)


plt.tight_layout()


plt.savefig(
    OUTPUT_HEATMAP,
    dpi=150
)


plt.close()


print(
    f"\n保存: {OUTPUT_HEATMAP}"
)


# ============================================================
# 上限を元Q3付近に固定し下限を動かす
# ============================================================

fixed_upper_candidates = result_df[
    np.isclose(
        result_df[
            "upper"
        ],
        BASE_UPPER
    )
].copy()


fixed_upper_candidates = (
    fixed_upper_candidates
    .sort_values(
        "lower"
    )
)


plt.figure(
    figsize=(
        11,
        6
    )
)


plt.plot(
    fixed_upper_candidates[
        "lower"
    ],
    fixed_upper_candidates[
        "profit_factor"
    ],
    marker="o"
)


plt.axhline(
    1.0,
    linestyle="--"
)


plt.axvline(
    BASE_LOWER,
    linestyle="--"
)


plt.title(
    f"Lower Threshold Sensitivity "
    f"(Upper = {BASE_UPPER:.2f})"
)

plt.xlabel(
    "Lower Threshold"
)

plt.ylabel(
    "Profit Factor"
)

plt.grid()


plt.tight_layout()


plt.savefig(
    OUTPUT_LOWER_GRAPH,
    dpi=150
)


plt.close()


print(
    f"保存: {OUTPUT_LOWER_GRAPH}"
)


# ============================================================
# 下限を元Q3付近に固定し上限を動かす
# ============================================================

fixed_lower_candidates = result_df[
    np.isclose(
        result_df[
            "lower"
        ],
        BASE_LOWER
    )
].copy()


fixed_lower_candidates = (
    fixed_lower_candidates
    .sort_values(
        "upper"
    )
)


plt.figure(
    figsize=(
        11,
        6
    )
)


plt.plot(
    fixed_lower_candidates[
        "upper"
    ],
    fixed_lower_candidates[
        "profit_factor"
    ],
    marker="o"
)


plt.axhline(
    1.0,
    linestyle="--"
)


plt.axvline(
    BASE_UPPER,
    linestyle="--"
)


plt.title(
    f"Upper Threshold Sensitivity "
    f"(Lower = {BASE_LOWER:.2f})"
)

plt.xlabel(
    "Upper Threshold"
)

plt.ylabel(
    "Profit Factor"
)

plt.grid()


plt.tight_layout()


plt.savefig(
    OUTPUT_UPPER_GRAPH,
    dpi=150
)


plt.close()


print(
    f"保存: {OUTPUT_UPPER_GRAPH}"
)


# ============================================================
# 最終表示
# ============================================================

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

print(
    "分析完了"
)

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

print(
    "\n出力ファイル:"
)

print(
    OUTPUT_CSV
)

print(
    OUTPUT_FILTERED_CSV
)

print(
    OUTPUT_TOP_CSV
)

print(
    OUTPUT_HEATMAP
)

print(
    OUTPUT_LOWER_GRAPH
)

print(
    OUTPUT_UPPER_GRAPH
)