67 lines
2.1 KiB
Python
67 lines
2.1 KiB
Python
import pandas as pd
|
||
from common.bet_tools import compute_metrics
|
||
from common.bet_tools import calculate_no_vig_moneyline_power, moneyline_to_prob
|
||
from loguru import logger
|
||
from common.utils import timeit
|
||
|
||
|
||
def get_no_vig_prob(row) -> pd.Series:
|
||
odds = [row["first_price"], row["second_price"]]
|
||
no_vig_odds_power = calculate_no_vig_moneyline_power(odds)
|
||
novig_probs_power = [moneyline_to_prob(o) for o in no_vig_odds_power]
|
||
# 返回两个无水概率
|
||
return pd.Series(
|
||
{
|
||
"first_no_vig_prob": novig_probs_power[0],
|
||
"second_no_vig_prob": novig_probs_power[1],
|
||
}
|
||
)
|
||
|
||
|
||
@timeit
|
||
def calc_metrics(df: pd.DataFrame, cols: list) -> pd.DataFrame:
|
||
data_list = []
|
||
for cs in df[cols].drop_duplicates().values:
|
||
tmp_df = df[cols + ["win_prob", "res"]].copy()
|
||
for i, col in enumerate(cols):
|
||
tmp_df = tmp_df[tmp_df[col] == cs[i]]
|
||
# if len(tmp_df) < 10000:
|
||
# continue
|
||
res = compute_metrics(df=tmp_df, include_draws=False)
|
||
res["filter_cols"] = ",".join(cs)
|
||
data_list.append(res)
|
||
res_df = pd.DataFrame(data_list)
|
||
res_df["reg_alpha"] = abs(res_df["reg_alpha"])
|
||
res_df = res_df.sort_values(by=["brier", "logloss", "ece", "reg_alpha"])
|
||
return res_df
|
||
|
||
|
||
if __name__ == "__main__":
|
||
df = pd.read_csv(
|
||
"/Users/aszer/Documents/vscode/bet/data/pinnical_1xbet_all_api.csv",
|
||
encoding="utf-8-sig",
|
||
)
|
||
df = df[
|
||
[
|
||
"sportsbook",
|
||
"sport",
|
||
"league",
|
||
"fixture_id",
|
||
"game_id",
|
||
"market",
|
||
"first_price",
|
||
"second_price",
|
||
"market_width",
|
||
"result",
|
||
]
|
||
]
|
||
# 防止 SettingWithCopyWarning，推荐使用 .loc 显式分配
|
||
df.loc[:, ["first_no_vig_prob", "second_no_vig_prob"]] = df[
|
||
["first_price", "second_price"]
|
||
].apply(get_no_vig_prob, axis=1)
|
||
df["win_prob"] = df["first_no_vig_prob"]
|
||
df["res"] = df["result"]
|
||
cols = ["sportsbook", "sport"]
|
||
res_df = calc_metrics(df, cols)
|
||
logger.info(f"\n{res_df}")
|