Python计算结果与Cycle Time Widget平均周期时间不符求助

问题概述 #

我在对比Azure DevOps Cycle Time Widget显示的平均周期时间与自定义计算结果时，发现数值存在差异：

• Cycle Time Widget显示12天
• 用Python/Pandas调用Feed OData API直接计算平均值得到11天
• 按官方逻辑实现移动平均时，结果为9天或11天
• 用Power BI连接Feed OData API计算，结果同样为11天

代码实现（Jupyter Notebook） #

import pandas as pd
import requests
import json
import base64
import math
from datetime import datetime, timedelta

token_do_azure = '{hidden}'
pat_encoded = base64.b64encode((":" + token_do_azure).encode()).decode()
headers = {
    'Content-Type': 'application/json',
    'Authorization': f'Basic {pat_encoded}'
}

hoje = datetime.today()
delta = timedelta(days=90)
dia_resultante = hoje - delta
dia_formatado = dia_resultante.strftime('%Y-%m-%dT00:00:00.00000Z')

url = rf"https://analytics.dev.azure.com/{hidden}/_odata/v4.0-preview/WorkItems?$select=WorkItemId,WorkItemType,Title,CycleTimeDays,ClosedDate&$filter=(Project/ProjectName eq 'Suporte_Torres' AND (WorkItemType eq 'Issue') AND State eq 'Done' AND ClosedOn/Date ge {dia_formatado})"

req = requests.get(url, headers=headers)
req_tabela = json.loads(req.text)
req_valores = req_tabela["value"]

df = pd.DataFrame(req_valores)
df['ClosedDate'] = pd.to_datetime(df['ClosedDate'], format='ISO8601').dt.date

# 直接计算平均值
print(round(df['CycleTimeDays'].mean(), 0))
# 返回11.0，而非预期的12.0

# 移动平均计算（参考官方逻辑）
def calcular_janela_n(n_dias):
    n = int(0.2 * n_dias)
    n = math.floor(n)
    if n % 2 == 0:
        n -= 1
    if n < 1:
        n = 1
    return n

janela_n = calcular_janela_n(90)
df['SMA_n'] = df['CycleTimeDays'].rolling(window=janela_n, min_periods=1).mean()

print(round(df['SMA_n'].tail(1).iloc[0], 0))
# 返回9.0，而非预期的12.0
print(round(df['SMA_n'].mean(), 0))
# 返回11.0，而非预期的12.0

待排查疑问 #

请协助分析可能导致差异的原因：

• OData API返回的CycleTimeDays字段计算逻辑是否与Widget不一致？
• Widget是否对时间范围、工作项筛选有特殊处理（比如状态转换规则、时区差异）？
• 移动平均的窗口计算或数据排序逻辑是否与官方实现存在偏差？

内容的提问来源于stack exchange，提问作者fcoalcantarajr