想做哪类“关键球员”统计？给我这几项就能快速落地：

• 联赛与赛季：例如 中超 2024 / 英超 2023-24
• 数据来源：是否已有 CSV/Excel/SQL 或需我帮你爬/接口对接
• 指标口径：只看传统数据（进球/助攻）还是包含高级数据（xG/xA、关键传球、推进传、压迫等）
• 输出形式：CSV/Excel 汇总、排行榜、可视化（雷达图/分布图）、或生成报告

如果你已有逐场球员数据，我可以直接跑下面这段脚本，产出“每90分钟标准化 + 百分位 + 关键球员榜”。你把文件名和列名对上即可。

示例：基于逐场球员统计的一键分析（pandas）

• 预期输入 players_match_stats.csv（逐场一行）
• 必要列：player_id, player_name, team, position, minutes, goals, assists
• 推荐列（有则算）：shots, xg, xa, key_passes, progressive_passes, carries, touches, passes_completed, passes_attempted, pressures, tackles, interceptions

# analyze_league_players.py
import pandas as pd
import numpy as np

# 1) 读取数据
df = pd.read_csv("players_match_stats.csv")

# 2) 确保必要列存在
required = ["player_id","player_name","team","position","minutes","goals","assists"]
for col in required:
    if col not in df.columns:
        raise ValueError(f"缺少必要列: {col}")

optional = {
    "shots":0,"xg":0.0,"xa":0.0,"key_passes":0,"progressive_passes":0,
    "carries":0,"touches":0,"passes_completed":0,"passes_attempted":0,
    "pressures":0,"tackles":0,"interceptions":0
}
for col,fill in optional.items():
    if col not in df.columns:
        df[col] = fill

# 3) 清洗
df = df.copy()
df["minutes"] = pd.to_numeric(df["minutes"], errors="coerce").fillna(0)
df = df[df["minutes"] > 0]

# 4) 按球员聚合（跨比赛）
agg_dict = {
    "team":"last","position":"last",
    "minutes":"sum","goals":"sum","assists":"sum","shots":"sum","xg":"sum","xa":"sum",
    "key_passes":"sum","progressive_passes":"sum","carries":"sum","touches":"sum",
    "passes_completed":"sum","passes_attempted":"sum",
    "pressures":"sum","tackles":"sum","interceptions":"sum"
}
g = df.groupby(["player_id","player_name"], as_index=False).agg(agg_dict)

# 5) 计算每90与比率
m = g["minutes"].replace(0, np.nan)
def per90(x): return x / m * 90

g["g_plus_a"] = g["goals"] + g["assists"]
g["g_plus_a90"] = per90(g["g_plus_a"])
g["xg90"] = per90(g["xg"])
g["xa90"] = per90(g["xa"])
g["shots90"] = per90(g["shots"])
g["kp90"] = per90(g["key_passes"])
g["prog_pass90"] = per90(g["progressive_passes"])
g["pressures90"] = per90(g["pressures"])
g["tack_int90"] = per90(g["tackles"] + g["interceptions"])

g["pass_comp"] = np.where(g["passes_attempted"]>0, g["passes_completed"]/g["passes_attempted"], np.nan)
g["shot_conv"] = np.where(g["shots"]>0, g["goals"]/g["shots"], np.nan)

# 6) 设定样本门槛（避免小样本偏差）
MIN_MINUTES = 600
pool = g[g["minutes"] >= MIN_MINUTES].copy()
if pool.empty:
    raise ValueError("样本为空：请降低 MIN_MINUTES 或检查数据。")

# 7) 百分位与综合评分（通用版）
metrics = ["g_plus_a90","xg90","xa90","shots90","kp90","prog_pass90","tack_int90","pressures90"]
for col in metrics:
    # 百分位排名（0-100）
    pool[col + "_pct"] = pool[col].rank(pct=True) * 100

# 根据位置给权重（可按需调整）
def pos_group(p):
    if pd.isna(p): return "OTH"
    p = str(p).upper()
    if any(k in p for k in ["ST","CF","FW","LW","RW","WF"]): return "ATT"
    if any(k in p for k in ["AM","CM","DM","MID"]): return "MID"
    if any(k in p for k in ["CB","FB","LB","RB","WB","DEF"]): return "DEF"
    return "OTH"

pool["pos_group"] = pool["position"].apply(pos_group)

weights = {
    "ATT": {"g_plus_a90":3,"xg90":2,"xa90":2,"shots90":1,"kp90":1,"prog_pass90":1,"tack_int90":0.5,"pressures90":0.5},
    "MID": {"g_plus_a90":1.5,"xg90":1,"xa90":2,"shots90":0.5,"kp90":2,"prog_pass90":2,"tack_int90":1.5,"pressures90":1.5},
    "DEF": {"g_plus_a90":0.5,"xg90":0.5,"xa90":0.5,"shots90":0.2,"kp90":0.8,"prog_pass90":1.5,"tack_int90":3,"pressures90":2},
    "OTH": {"g_plus_a90":1,"xg90":1,"xa90":1,"shots90":1,"kp90":1,"prog_pass90":1,"tack_int90":1,"pressures90":1}
}

def score_row(r):
    w = weights.get(r["pos_group"], weights["OTH"])
    s = 0.0
    tot = 0.0
    for mcol, wt in w.items():
        v = r[mcol + "_pct"]
        if pd.notna(v) and wt>0:
            s += v * wt
            tot += wt
    return s / tot if tot>0 else np.nan

pool["key_player_score"] = pool.apply(score_row, axis=1)

# 8) 输出榜单
cols_out = [
    "player_id","player_name","team","position","minutes",
    "g_plus_a","g_plus_a90","xg90","xa90","shots90","kp90","prog_pass90",
    "tack_int90","pressures90","pass_comp","shot_conv",
    "key_player_score"
]
pool.sort_values("key_player_score", ascending=False).to_csv("key_players_overall.csv", index=False)

# 分位置前10
topk = []
for grp in ["ATT","MID","DEF","OTH"]:
    part = pool[pool["pos_group"]==grp].nlargest(10, "key_player_score")
    topk.append(part)
pd.concat(topk, ignore_index=True).to_csv("key_players_by_position_top10.csv", index=False)

print("已生成: key_players_overall.csv, key_players_by_position_top10.csv")

可按需扩展

• 加入非点球指标：如果有 np_goals / pen_goals 列，可替换/新增 np_g90。
• 百分位范围：可改用同位置内百分位，避免跨位置偏置。
• 可视化：雷达图（单人）、蜂群图（位置维度）、累积分布（分钟阈值敏感性）。
• 小样本控制：提高分钟阈值，或对每90进行 Empirical Bayes 收缩。

告诉我你的联赛/赛季、是否已有数据文件（发我列名示例），我就按你的口径生成成品榜单和图表。