How SignalXI selects its analyses
The process is designed to reduce forced publications and focus on situations where probability, price and data quality are aligned.
1. Context and recent data
Before evaluating a market, the available context is reviewed: recent performance, attacking and defensive production, opponent, competition and data sufficiency. Not every variable carries the same weight across different markets.
2. Independent probability
The model produces an estimate for the event being analyzed. That probability is a statistical assessment, not a guarantee. The aim is for the estimate to remain useful across repeated observations, not simply to predict one isolated match.
3. Price verification
The available odds are converted into an implied probability reference. SignalXI compares that reference with its own estimate and avoids treating a selection as attractive simply because its estimated probability is high.
4. Edge and expected value
Edge measures the gap between estimated probability and the probability implied by the price. EV combines probability and odds into a theoretical expected-value measure. Both are comparison tools, not promises of return.
5. Publication filters
An opportunity must clear the system's minimum criteria before appearing as a published analysis. SignalXI may publish anywhere from zero to several analyses per day; it does not force a fixed total or a quota by market type.
6. One recommendation per match
When several options from the same fixture compete with each other, the system prioritizes a single final recommendation. This reduces exposure to correlated selections and keeps later evaluation easier to interpret.
7. Classification and observation
Labels such as HIGH and WATCH help organize published analyses by their characteristics. These groups can be reviewed separately to see whether their behavior differs as the sample grows.
8. Tracking and model review
Published analyses are recorded and later settled. Wins, losses, ROI, units, average odds, probabilities and segment behavior are observed over time. Small samples are not treated as sufficient evidence for changing the model based on one short-term result.
What SignalXI does not do
SignalXI does not accept bets, hold gambling funds or guarantee profits. The platform provides statistical and informational analysis and keeps the modeling process separate from any later user decision.
View published analyses →