Model vs market: what an edge actually means
Two probabilities can tell different stories. Here is how to read the comparison without mistaking it for a promise.
An alert is a starting point for understanding a disagreement. It tells you that a model and a market are assigning different probabilities to the same outcome. Before deciding what that difference means, it helps to be clear about what each number represents.
One number comes from the market
On Kalshi, the price of an event contract can be read as the market’s implied probability for its outcome. It reflects what participants are willing to pay. That is a useful summary of the market’s view, rather than a statement that the outcome is certain. Kalshi explains this price-to-probability relationship in its own guide.
The other comes from a model
Deviate’s model comparison starts with the game situation. Depending on the sport, that can include the score, time remaining, possession, inning, or set and game score. The historical models look at outcomes in comparable situations. Some leagues use an estimate formula instead; the Sports page explains the difference.
This gives a separate view of the outcome. It is an estimate based on the model’s inputs and assumptions. It does not capture everything another participant might know about an event.
The difference is measured in points
Model probability − market-implied probability = the displayed edge in percentage points.
The word “edge” names that model-to-market difference. It is not a return, a win rate, or a promise of profit. The original comparison in a Deviate alert records a moment; a market can move after that moment, so a past alert is not a current quote.
A disagreement can have more than one explanation
Markets can incorporate information that a historical game-state comparison does not. A participant may be responding to an injury, a player’s condition, or another detail that the score alone cannot describe. A difference between the numbers does not establish which one is right.
The underlying information can also be delayed or incomplete. Reading the game state and the explanation matters just as much as reading the size of the gap. A large number deserves context, rather than automatic confidence.
What an alert gives you
Deviate brings the model number, the market number, the gap, and a plain-English reason together. You choose the threshold that should trigger a notification. Once the game finishes, a resolved alert’s outcome stays in History alongside the original comparison.
Use those details to understand what was observed and when. A winning example does not establish typical results, and a historical probability never removes uncertainty. Deviate provides information and does not place trades. Any decision to use an exchange remains yours.
Market-price background: Kalshi’s guide to reading probabilities. Product details: How Deviate works and sport-by-sport coverage.