openquant.volatility

Compiled from Rust (crates/pyopenquant). Signatures and types are from the shipped .pyi stubs; descriptions are the functions’ own docstrings, as help() shows them. All Python modules.

Functions

get_daily_vol

get_daily_vol(
    close_timestamps: Sequence[str],
    close_prices: Sequence[float],
    lookback: int,
) -> list[tuple[str, float]]

Daily volatility: exponentially weighted standard deviation of daily returns.

AFML Snippet 3.1, matching mlfinlab.util.volatility.get_daily_vol. For each bar, the return is the simple return p_t / p_j - 1 against the last bar j strictly more than one day earlier; the volatility is pandas' ewm(span=lookback).std() of those returns (adjust=True, unbiased). Bars with no bar more than a day before them are omitted, and the first returned value is NaN (one return has no sample variance). Bars must be in increasing time. A typical use is as the target of the labeling functions, which drop NaN targets.

Parameters

close_timestamps : list[str]

Bar timestamps as "%Y-%m-%d %H:%M:%S" (an optional fractional second is accepted), in increasing order.

close_prices : list[float]

Close price of each bar.

lookback : int

EWM span, in number of returns.

Returns

list[tuple[str, float]]

(timestamp, volatility) rows; empty when there are fewer than two bars or lookback is 0.

Raises

ValueError

If the timestamps and prices differ in length, or a timestamp does not parse.

get_garman_class_vol

get_garman_class_vol(
    open: Sequence[float],
    high: Sequence[float],
    low: Sequence[float],
    close: Sequence[float],
    window: int,
) -> list[float]

Garman-Klass volatility over a rolling window.

Each output is the square root of the mean over the last window bars of 0.5 ln(high / low)^2 - (2 ln 2 - 1) ln(close / open)^2, per bar (not annualised), matching mlfinlab.util.volatility.get_garman_class_vol. The first window - 1 outputs are NaN, as is any output whose window contains a NaN; window=0 gives all NaN.

Parameters

open : list[float]

Open price of each bar, oldest first.

high : list[float]

High price of each bar.

low : list[float]

Low price of each bar.

close : list[float]

Close price of each bar.

window : int

Number of bars in the rolling window.

Returns

list[float]

Volatility per bar, same length as open.

Raises

ValueError

If high, low or close differs in length from open.

get_parkinson_vol

get_parkinson_vol(
    high: Sequence[float],
    low: Sequence[float],
    window: int,
) -> list[float]

Parkinson high-low volatility over a rolling window.

Each output is sqrt(mean(ln(high / low)^2) / (4 ln 2)) over the last window bars, per bar (not annualised), matching mlfinlab's get_parksinson_vol. The first window - 1 outputs are NaN, as is any output whose window contains a NaN; window=0 gives all NaN.

Parameters

high : list[float]

High price of each bar, oldest first.

low : list[float]

Low price of each bar.

window : int

Number of bars in the rolling window.

Returns

list[float]

Volatility per bar, same length as high.

Raises

ValueError

If low differs in length from high.

get_yang_zhang_vol

get_yang_zhang_vol(
    open: Sequence[float],
    high: Sequence[float],
    low: Sequence[float],
    close: Sequence[float],
    window: int,
) -> list[float]

Yang-Zhang (2000) volatility over a rolling window.

For the n = window bars ending at each bar, returns sqrt(var_o + k var_c + (1 - k) var_rs) with k = 0.34 / (1.34 + (n + 1) / (n - 1)), where var_o is the sample variance of the overnight returns ln(open_i / close_{i-1}), var_c that of the open-to-close returns ln(close_i / open_i), and var_rs the mean Rogers-Satchell term. The result is per bar, not annualised. The overnight return needs the previous close, so the first window outputs are NaN; a window containing a NaN gives NaN and window < 2 gives all NaN. This differs from mlfinlab's get_yang_zhang_vol, which uses ln(close_i / open_{i-1}) and does not demean.

Parameters

open : list[float]

Open price of each bar, oldest first.

high : list[float]

High price of each bar.

low : list[float]

Low price of each bar.

close : list[float]

Close price of each bar.

window : int

Number of bars in the rolling window.

Returns

list[float]

Volatility per bar, same length as open.

Raises

ValueError

If high, low or close differs in length from open.