Quant Library · Markets and staking · Any

Walk-Forward Backtest Harness

The test that refits before every week and scores against the closing line.

A harness that refits on a schedule, prices the week ahead, scores Brier, log loss, calibration and skill against the market, and simulates a staking rule with closing-line value attached. The point of it is to make lying to yourself difficult.

Read the specification (PDF) Run it on the API The markets and staking family · £699

Inputs

What it takes

Name	Type	Units	Where it comes from
fixtures	list with results and prices	-	your database
refit_days optional	int	days	how often to refit; weekly is the platform's
min_train optional	int	matches	before which no price is emitted
staking rule optional	callable	-	optional, to simulate returns as well as scores

Outputs

What it gives back

Name	Type	Units	Notes
metrics	dict	Brier, log loss, accuracy, ECE
market metrics	dict	the same, for the closing line
brier_skill_vs_market	float	fraction
betting	dict	bets, staked, profit, ROI, CLV, max drawdown
predictions	list	-	every priced match, kept so a later rule can be replayed without refitting

Method

How it works

Walk the calendar. At each refit date, fit on everything strictly before it and price the following window.
3 further steps in the licensed specification

Assumptions

The stored prices are what you could actually have taken. If your feed's opening price lagged the market, the harness will flatter you and the closing-line value column is how you find out.

Limits

It cannot know about a limit you would have hit, a price you would have been refused, or a market that would have moved as you bet.
1 further limits in the licensed specification

Validation

What it was measured at

The harness's own output, over every league the platform covers.

Matches walked forward	419
Of those, with a closing price to score against	419
Refit cadence	weekly	fit on everything strictly before the week
Reported per league	Brier, log loss, accuracy, calibration, skill, ROI, CLV	including when they are negative
Predictions kept	yes	so a staking rule can be replayed later without refitting

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.