EdgeworthArena
Live dataSign in
Quant Library · The machine layer · Any

Bayesian Online Changepoint Detector

Has this side actually changed, or is it a run of coins?

Adams and MacKay's run-length posterior on the residuals of a price, so the detector judges a side against what was expected of it rather than against its own recent form.

Inputs

What it takes

3
NameTypeUnitsWhere it comes from
residualslist of floatstandard deviationsgoal difference minus the pre-match expectation, standardised
hazard optionalfloat1/matchesyour prior on how often a real change happens; 1/40 is the platform's
sigma optionalfloatstandard deviationsobservation noise
Outputs

What it gives back

3
NameTypeUnitsNotes
run_length_posteriorarrayprobability
p_change_last_kfloat for k in 3, 5, 10probability
regime_meanfloatstandard deviations
Method

How it works

  1. Keep a distribution over the run length, the number of observations since the last change.
  2. 2 further steps in the licensed specification
Assumptions
  • Residuals are normal around a piecewise-constant level. Heavy tails make the detector jumpy; the platform standardises first for that reason.
Limits
  • It is a detector, not a forecast. It tells you a level has moved; it does not tell you where the level is going.
  • 1 further limits in the licensed specification
Validation

What it was measured at

ok

Run on the residuals of the walk-forward prices, so a side is judged against what was expected of it.

Sides tracked18
Flagged above 50% in the last ten1a detector that fires on everything is not a detector
Hazard1 in 40the prior on how often a real change happens
Observation modelnormal, conjugate priorso the update is closed form
In the package
  • The full specification: inputs, method, assumptions and limits
  • The validation panel as measured on the day of purchase
  • A reference implementation extracted from the running source
  • Perpetual commercial use for one entity

Reference modules:

  • 1 module(s), named in the licensed specification and shipped in the package
Read this before you buy. The platform's own walk-forward tests say the closing line forecasts football better than these models do. What they give you is a coherent price for every market and match, before the market opens, with the band and the working attached. Any model here that has been beaten by something simpler says so in its limits, and the validation panel above is generated from the platform's result files rather than typed in.