Quant Library · Markets and staking · Football
Game-State Strategy Solver
When to press and when to hold, by backward induction on the game state.
A finite-horizon Markov decision process over minute and score difference, with transition rates fitted on half-time and full-time states, solved exactly for the policy that maximises expected points.
Inputs
3What it takes
| Name | Type | Units | Where it comes from |
|---|---|---|---|
| state rates | dict | goals per minute by lead | fitted on your own match states |
| actions | list | - | press, hold, park, with their effect on both sides' rates |
| objective optional | string | - | points, or win probability |
Outputs
3What it gives back
| Name | Type | Units | Notes |
|---|---|---|---|
| policy | action per (minute, lead) | - | |
| value | expected points | points | |
| gain over holding | float | points per match |
Method
How it works
- Discretise the match into minutes and the score into a bounded lead.
- 2 further steps in the licensed specification
Assumptions
- Actions change rates in the way your fit says they do, and a manager can actually execute them.
Limits
- The gain over simply holding a balanced approach is a tenth or two of a point per match. It is real and it is small, and anyone claiming more is selling something.
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:
- 2 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.