Score the market before the team
The distinctions in DraftKings' current soccer rules frame this checklist: result, goal, handicap, player, cards, corners and tournament products can use different qualifying periods and settlement conditions.
| Criterion | Pass question |
|---|---|
| Contract | Can the event, line, period and settlement source be written exactly? |
| Data | Does one provider cover every required event with a stable definition? |
| Probability | Is there a documented transformation from inputs to a forecast? |
| Validation | Has that forecast been tested on later, unseen events? |
| Price | Is the accepted quote retained at full precision? |
| Dependence | Does another leg change this event's probability? |
| Maintenance | Can rule, provider and competition changes be detected? |
Apply the current local rule and accepted ticket to the actual selection.
Like-for-like market comparison
| Market family | Event to forecast | Common hidden state | Evidence burden | Dependence concern |
|---|---|---|---|---|
| Three-way moneyline | Home, draw or away in the stated period | Regulation draw | Complete outcome vector | Shared team and match context |
| Two-way advance/winner | Which team qualifies or wins the tie | Extra time or shoot-out | Knockout path model | Linked to series or bracket state |
| Goal total | Goals above or below an exact line | Push or split line | Score distribution | Shared attack, defense and game state |
| Handicap or spread | Adjusted result after the line | Push or quarter split | Margin distribution | Linked to moneyline and team totals |
| BTTS | Both teams score in the period | Own-goal and period wording | Joint scoring distribution | Linked to total goals and result |
| Player prop | Named player event or threshold | Non-participation or provider correction | Minutes, role and event model | Linked to team result and teammates |
| Cards or corners | Defined event count | Awarded versus taken; booking points | Provider and referee definitions | Linked to game state and other props |
| Futures | Season or tournament outcome | Field change, dead heat, incomplete event | Full field and path probabilities | Long horizon and shared teams/players |
As scikit-learn's calibration guidance makes clear, probability quality depends on observed reliability rather than a market label. The best-scoring market can differ by competition, provider, model and date. A market with fewer narrative unknowns can still be badly priced; a complex market can be valid when its data and joint probability are genuinely supported.
Result markets
A three-way soccer moneyline must allocate probability to home, draw and away. A two-way "to advance" or "draw no bet" product has different states. Use Soccer Moneyline Parlays to preserve the full outcome vector and distinguish regulation time from qualification.
Goals and handicap markets
Totals and handicaps can be modelled from a score distribution, but the selected line controls payoff. Whole lines can push, half lines cannot, and quarter lines can split where offered. Do not convert a model's expected goal total directly into an over probability without the distribution around it.
Use Soccer Over/Under Parlays for line-by-line settlement and score-model validation.
Player and event props
Provider definitions matter. Opta's football-statistics definitions show why similarly named possession, pressure and defensive events need a named source. Player props also require participation, minutes, role and correction rules.
Same-game combinations
A sportsbook's same-game quote is a contract price, not a disclosure of the underlying dependence model. DraftKings' market rules provide one current example of separately priced related selections. Estimate the joint event directly and compare it with the accepted quote.
Worked decision scorecard
Suppose a reader can model match result and total goals but has no stable player-minutes feed. A transparent scorecard could be:
| Candidate | Contract | Data | Validation | Dependence | Decision |
|---|---|---|---|---|---|
| Three-way result | Pass | Pass | Pass | Review | Eligible for price comparison |
| Over 2.5 goals | Pass | Pass | Pass | Review | Eligible for price comparison |
| Player shots | Pass | Fail | Not tested | High | Exclude |
Exclusion is a useful outcome. Do not fill a parlay with a weaker market merely to reach a leg count.
Probability and expected value
scikit-learn's calibration guidance explains why probability reliability must be evaluated against observed frequencies. Once a joint forecast is available, use OpenStax's expected-value framework with every net payoff state. Market familiarity alone is not evidence of value.
Continue learning
- Next guide: AI Soccer Predictions for Parlays
- Related guide: Bundesliga Parlays
Assumptions and limitations
The framework does not rank live operators, leagues, teams or current prices. Market availability and rules vary by state and sportsbook. A passed scorecard means the leg can be analysed; it does not mean the accepted price is favourable or the forecast is accurate.
GambleAware
GamCare
Gambling Therapy
GAMSTOP