hsliuping/TradingAgents-CN · error · ValueError

不支持的RSI计算方法: {method}，支持的方法: 'ema', 'sma', 'china'

Error message

不支持的RSI计算方法: {method}，支持的方法: 'ema', 'sma', 'china'

What it means

The rsi function supports exactly three computation methods: 'ema' (exponential smoothing), 'sma' (simple moving average of gains/losses), and 'china' (同花顺/通达信-style SMA(X,N,1) via ewm(com=n-1, adjust=True)). Passing any other method string reaches the final else branch and raises this ValueError. The method parameter exists because different charting platforms produce visibly different RSI values.

Source

Thrown at tradingagents/tools/analysis/indicators.py:117

    gain = delta.where(delta > 0, 0)
    loss = -delta.where(delta < 0, 0)

    if method == 'ema':
        # 国际标准：Wilder's指数移动平均
        avg_gain = gain.ewm(alpha=1 / float(n), adjust=False).mean()
        avg_loss = loss.ewm(alpha=1 / float(n), adjust=False).mean()
    elif method == 'sma':
        # 简单移动平均
        avg_gain = gain.rolling(window=int(n), min_periods=1).mean()
        avg_loss = loss.rolling(window=int(n), min_periods=1).mean()
    elif method == 'china':
        # 中国式SMA：同花顺/通达信风格
        # SMA(X, N, 1) = ewm(com=N-1, adjust=True).mean()
        # 参考：https://blog.csdn.net/u011218867/article/details/117427927
        avg_gain = gain.ewm(com=int(n) - 1, adjust=True).mean()
        avg_loss = loss.ewm(com=int(n) - 1, adjust=True).mean()
    else:
        raise ValueError(f"不支持的RSI计算方法: {method}，支持的方法: 'ema', 'sma', 'china'")

    rs = avg_gain / (avg_loss.replace(0, np.nan))
    rsi_val = 100 - (100 / (1 + rs))
    return rsi_val


def boll(close: pd.Series, n: int = 20, k: float = 2.0, min_periods: int = None) -> pd.DataFrame:
    """
    计算布林带指标（Bollinger Bands）

    Args:
        close: 收盘价序列
        n: 周期，默认20
        k: 标准差倍数，默认2.0
        min_periods: 最小周期数，默认为1（允许前期数据不足时也计算）

    Returns:
        包含 boll_mid, boll_upper, boll_lower 的 DataFrame

View on GitHub (pinned to 74783e8817)

Solutions

1. Use one of the exact strings 'ema', 'sma', or 'china' (lowercase).
2. If you want Wilder's RSI (TA-Lib default), note this library's 'china' method uses ewm(com=n-1) which is equivalent to Wilder smoothing — use that.
3. If a genuinely different smoothing is needed, compute it manually with pandas ewm rather than passing an unsupported method string.

Example fix

# before
rsi_val = rsi(df["close"], n=14, method="wilder")

# after
rsi_val = rsi(df["close"], n=14, method="china")  # ewm(com=n-1) == Wilder-style smoothing

VALID_RSI_METHODS = {"ema", "sma", "china"}
method = (method or "ema").lower()
if method not in VALID_RSI_METHODS:
    method = "china"  # or raise your own config error
rsi_val = rsi(df["close"], n=14, method=method)

def is_valid_rsi_method(m: str) -> bool:
    """Narrow to the library's supported RSI methods."""
    return isinstance(m, str) and m in {"ema", "sma", "china"}

try:
    rsi_val = rsi(close, n, method=method)
except ValueError as e:
    if "不支持的RSI计算方法" in str(e):
        rsi_val = rsi(close, n, method="china")  # sensible default
    else:
        raise

When it happens

Trigger: Calling rsi(close, n, method='wilder'), method='Wilder', method='EMA' (case-sensitive), or compute_indicator(df, 'rsi', method='wma'). Also passing method=None explicitly if the default handling doesn't catch it before the else.

Common situations: Porting code from other libraries where the Wilder/smoothing method is named differently ('rma', 'wilder', 'cutler'); case mismatches; typos like 'cn' or 'zh' instead of 'china'; assuming TradingView/TA-Lib naming applies here.

Related errors

• DataFrame缺少必要列: {missing}, 现有列: {list(df.columns)[:10]}...
• DataFrame缺少收盘价列: {close_col}
• Indicator {indicator} is not supported. Please choose from:
• 不支持的指标: {name}

AI-assisted analysis of hsliuping/TradingAgents-CN@74783e8817 (2026-08-28). Data as JSON: /api/errors/fb4e7f4e0fc77088. Report an issue: GitHub.