Card guide

How to read a SnapFactor card

A card estimates the chance of a stated player outcome. A market-comparison card also shows how that estimate compares with a captured sportsbook offer.

Start with these fields

Model probability
SnapFactor’s estimated chance that the exact outcome printed on the row occurs. A 58% forecast still leaves a 42% chance it does not.
Line and outcome
Over 3.5 receptions means at least 4; Under 3.5 means at most 3. For these half-point lines, the under probability is the complement of the at-least-4 forecast. A 4+ receptions milestone means at least 4. Read the stat and side together.
Odds and implied probability
American odds show the offer’s payout: +150 profits $150 on a $100 stake; −110 needs a $110 stake to profit $100. The raw implied chance is 100 ÷ (odds + 100) for positive odds, or absolute odds ÷ (absolute odds + 100) for negative odds. It includes the book’s pricing margin.
Edge
Model probability minus the offer’s raw implied probability, shown in percentage points (pp). It is not a percent return.
EV
Estimated net return as a percent of the stake, using the model probability and the displayed odds. The calculation weights the profit if the outcome wins and the full stake loss if it loses.
Sportsbook and freshness
The book names the source of the displayed offer. “Market captured” records when the offer was collected; several rows may have different capture times. “Generated” is when the card was built. “Data cutoff” is the forecast input boundary. Prices and availability can change after capture.
Fictional example

One row, worked through

FICTIONAL — NOT A LIVE PICK
Example Player
Over 3.5 receptions · Example Sportsbook · −110
Model probability58.0%
Raw implied probability52.4%
Edge+5.6 pp
Estimated EV+10.7%

At −110, implied probability is 110 ÷ 210 = 52.38%. Edge is 58% − 52.38% = 5.62 percentage points. A $100 stake would profit $90.91 on a win, so estimated EV is (0.58 × $90.91) − (0.42 × $100) = $10.73 per $100 staked, or +10.7%. Values on cards are rounded for display.

Evidence labels

What CAL and SUP mean

CAL: VALIDATED

A historical evaluation selected a calibration method for this stat family, and the card uses its adjusted probability. Calibration checks how probabilities behave across evaluated outcomes; it cannot certify this individual pick.

UNCALIBRATED uses a raw probability when no adjustment is selected. NOT APPLICABLE appears when a probability cannot be produced. UNKNOWN means the display lacks a status source.

SUP: STRONG

The current support calculation requires at least 12 prior player games and at least 500 evaluated outcomes in the relevant probability bucket. MODERATE requires at least 6 games and 200 bucket outcomes; LIMITED is below those levels. INSUFFICIENT means the underlying forecast cannot provide a usable probability. Support measures available evidence, not the likelihood that one result wins.

Why can CAL and SUP differ?

CAL concerns a stat family’s evaluated probability adjustment. SUP counts the player history and comparable historical outcomes behind a specific forecast. A validated adjustment can coexist with moderate or limited support.

Card types

Forecast versus market

A forecast-only card presents a player outcome and its probability without a sportsbook offer. A market-comparison card adds a captured line, odds, sportsbook, edge and estimated EV. Tier 1, Tier 2 and Tier 3 currently select the same rows and calculations; their artwork and labels vary. A higher tier does not mean a stronger forecast.

Other labels on a card or its message

“Model” or “Forecast” identifies the forecast version. Board and market snapshot SHA-256 strings identify the frozen forecast and captured market artifacts; they are integrity identifiers, not quality scores. A row’s “Evidence: supported” corresponds to strong support on private market cards; “caution” corresponds to moderate or limited support. Private market rows are ordered by estimated EV among eligible offers, not by certainty.

Keep the limits in view

Positive estimated EV depends on model accuracy; it does not establish proven profitability. Support and calibration are evidence labels, not guarantees. Individual prop probabilities alone do not determine a parlay’s joint probability. Check the live sportsbook before acting on a captured price.