Module `Algostream_pairs.Hedge_ratio`

β estimator for the spread regression y ≈ β·x + α.

Three modes (selected via Config.beta_mode):

Static of float — fixed β, no online update; intercept fixed at 0.
Rolling_ols — closed-form rolling cov(x,y) / var(x) over Config.beta_window, with periodic full recompute to bound floating-point drift.
Kalman_smoothed — runs Filters.Kalman1d over the rolling-OLS β estimate, so the reported beta is a low-pass-filtered version of the noisy OLS estimate.

If Rolling_var(x) < 1e-12 (a flat regressor), β is held at its previous value and beta_frozen_ticks increments — avoids divide-by-near-zero blow-ups in calm regimes.

type t

val create : Config.t -> t

val update : t -> x:float -> y:float -> float * float

Returns the updated (beta, intercept) after observing one paired sample.

val beta : t -> float

val intercept : t -> float

val beta_stdev : t -> float

Trailing standard deviation of β across the most recent beta_window/4 updates. Used by Selection to screen out pairs with an unstable hedge ratio.

val n_updates : t -> int

val beta_frozen_ticks : t -> int

val ready : t -> bool