Research clock
| Stage | Action | Output |
|---|---|---|
| Fixture intake | Import official events and kickoffs | Versioned slate |
| Early model | Run only on available evidence | Initial probabilities and uncertainty |
| News update | Add timestamped injuries and squad news | Revised model version |
| Lineup gate | Confirm required participants | Keep, reprice or exclude decision |
| Acceptance | Save exact product, price and stake | Immutable receipt record |
| Settlement | Reconcile official and operator results | Adjustment ledger |
| Review | Analyse calibration, price and process | Changes for next cycle |
Official schedules can change; the Premier League fixture source is one current example that says fixtures are subject to change. Use each competition's official source.
Information-cutoff rules
Never edit an early probability and leave the same model timestamp. Every update gets a new version and reason. If a lineup is unknown, model participation or wait; do not enter the final lineup retrospectively.
scikit-learn's TimeSeriesSplit guidance explains chronological separation, and its calibration documentation supports probability reliability review.
Leg acceptance gate
- Exact event and settlement are known.
- Data and model cutoffs are preserved.
- Probability and uncertainty are complete.
- Accepted price is available at full precision.
- Dependence with every retained leg is mapped.
- Stake and total open exposure pass the written rule.
Reject unresolved legs instead of replacing them to maintain a target count. OpenStax's independence condition must be justified before combining probabilities.
Settlement and review
Use the current operator contract; Betfair's football rules are one example. Preserve voids, corrections, cash out and promotions separately. Calculate expected and realised net payoffs under OpenStax's framework.
Next step
Use Accumulator Tracking for the next part of this topic.
Post-settlement review
Review process and result separately. A winning receipt can contain a poor price or stale forecast, while a losing receipt can still have followed the declared method.
| Review field | Question |
|---|---|
| Forecast | Was the probability produced from the frozen input packet? |
| Price | Was the accepted price compared with that probability? |
| Execution | Did the saved receipt match the intended market and period? |
| Settlement | Did the operator apply the archived contract correctly? |
| Stake | Did the amount follow the pre-declared bankroll rule? |
Following chronological evaluation guidance, cancel or rebuild the candidate when the fixture, lineup, model version or executable price has changed after its final check. Record the cancellation reason before kickoff so an excluded winner cannot be reinstated after the result.
Aggregate these records over a pre-declared review window. Judge probability quality with calibration diagnostics, price quality against the available market snapshot, and execution quality through preventable error rates rather than accumulator win rate alone.
Continue learning
- Next guide: Weekly Acca Spreadsheet
- Related guide: Acca Insurance and Odds Boosts
Assumptions and limitations
This is an evergreen process, not a source of weekly selections. More frequent research does not guarantee a reliable model or favourable price.
GambleAware
GamCare
Gambling Therapy
GAMSTOP