Live decision sequence
- Record event ID, score, match clock, period and feed timestamp.
- Record players, substitutions, cards and any suspended or corrected state.
- Define every proposed leg and qualifying period.
- Update probabilities from the full current state, not a momentum label.
- Map dependence within and across the live matches.
- Capture the quoted price, submission time and actual acceptance response.
- Recalculate open exposure only after acceptance is confirmed.
- Preserve final settlement and every correction for replay testing.
One state, not mixed snapshots
Sportmonks' livescore documentation describes live fixtures, events, statistics, lineups, odds and prediction includes. Store each source timestamp. Combining a score from one feed with a lineup, clock or price from a later state creates a decision that never existed.
| Live field | Required record |
|---|---|
| Match | Event ID, teams, period and score |
| Time | Provider clock, stoppage context and retrieval time |
| Players | Lineup, substitutions, dismissals and roles |
| Market | Exact line, period and settlement rule |
| Forecast | Model version, state inputs and probability |
| Quote | Price, limit, request and acceptance timestamps |
Latency and suspension
The FanDuel rules published by the Massachusetts Gaming Commission explicitly warn that broadcasts may be delayed, requests can take longer in play, displayed live information may be inaccurate, and markets can be suspended. Those rules are state-specific, but the operational questions apply to any live product.
Do not assume a selected quote was accepted. Preserve rejection, suspension, price-change and partial product responses as part of the method.
Probability update
A live model should update from observable state and pre-match strength. Published Bayesian in-play research provides one model-specific example of updating football outcome probabilities and checking assumptions; it does not establish a universal live edge.
For an illustrative two-leg cross-match ticket with current probabilities 0.64 and 0.58 under an independence assumption:
Joint probability = 0.64 * 0.58 = 0.3712, or 37.12%
If the accepted combined decimal price is 2.55:
Raw break-even probability = 1 / 2.55 = 0.3922, or 39.22%
Under those hypothetical inputs, the forecast is below the raw break-even level before costs and uncertainty. The correct action can be no wager.
Same-match dependence
Score, time, red cards, substitutions and tactical state affect result, totals, props and corners together. Model the complete joint event for a live same-game parlay. A price update by the sportsbook is not evidence that the reader's model has handled dependence correctly.
Settlement and exposure
DraftKings' soccer rules provide one current example of qualifying periods and soccer settlement. Retain the local operator rules for abandoned matches, already-determined events, player props and corrections.
Only accepted stake enters open exposure. An unconfirmed click, stale quote or potential return is not cash position. Expected value uses current probabilities and net states under OpenStax's framework.
Stop conditions
Stop when the feed conflict cannot be resolved, the market is repeatedly suspended, the accepted price differs materially from the analysed quote, or the decision is being rushed to recover a loss. The AGA responsible-play guidance provides US account controls and support routes.
Next step
Use In Play Betting Football for the next part of this topic.
Continue learning
- Next guide: How Soccer Parlays Work
- Related guide: Soccer Futures Parlays
Assumptions and limitations
The worked values are hypothetical. Live feeds, clocks, broadcasts and odds can differ or lag; operators can suspend or reject markets; and in-play models are vulnerable to unobserved state and rapid change. This page does not publish live picks or claim a speed advantage.

