Fantasy Score Props: The Same Name, a Different Product

A rebound is a rebound. Whatever app you open, whatever league you are in, the box score records the same event and every product built on it agrees about what happened. That shared definition is why most player prop research travels: a projection of 8.4 rebounds means the same thing to everyone who reads it.

Fantasy score has no such shared definition. It is not an event that occurs on a field. It is an accounting output, produced by multiplying a set of real events by a set of coefficients that one platform chose and can change. Two apps can post the same player at the same number on the same night and be selling genuinely different products, and unlike most cases of cross platform disagreement, no amount of line shopping resolves it, because the disagreement is not about the number. It is about what the number counts.

A composite, not a statistic

The scoring systems used across fantasy sport all share one structure: take a list of countable events, assign each a coefficient, and sum. A hitter's fantasy total might weight a single one way, a home run another, a walk another, a stolen base another, with something subtracted for an out. A basketball total weights points, rebounds, assists, steals, blocks, and turnovers. A pitcher total weights outs and strikeouts and penalizes runs allowed, sometimes with a bonus for a start that meets a quality threshold.

Every one of those coefficients is a product decision. There is no natural scale for fantasy score, no governing body that standardizes it, and no reason for two competing platforms to converge on the same numbers. So the composite is not a property of the player's performance. It is a property of the performance and the chart together, and only one of those two things is on the box score.

This is why the entry in the glossary for fantasy score has to be written differently from the entry for rebounds. Rebounds has a definition. Fantasy score has a formula, and the formula has an owner.

Why you cannot convert a projection between platforms

Faced with a projection built for one platform and a line posted on another, the instinct is to scale. Find a ratio between the two charts and multiply. This is wrong in a way that is worth understanding precisely, because the failure is structural rather than approximate.

A composite is a weighted sum over several components, and two charts generally differ in more than one weight. The ratio between the two totals is therefore not a constant; it depends on the mix of components the player actually produces. A player who accumulates his score mostly through one component converts at a different ratio than a player who spreads it across several, even though both are described by the same two charts. Scaling by a single factor is an assumption that every player has the same component mix, and that assumption fails hardest exactly where the money is: on the unusual players whose profile is the reason you looked at them.

The correct method is unglamorous. Project the components, then apply the target platform's own chart to them. The components are the part of the problem that is shared across platforms and therefore worth modeling carefully; the chart is a lookup applied at the end. Doing it in that order also means adding a new platform is a data task rather than a modeling task, which is a good sign that the decomposition is the right one.

Composites inherit correlation, and that changes their shape

Once you are summing components, the relationships between those components stop being a footnote. A composite's variance is not the sum of its parts' variances. It is the sum plus a term for every pair of components that move together, scaled by their weights. That extra term can be large, and its sign is not always the one people assume.

Many components inside a fantasy total are positively correlated for the obvious shared reason: opportunity. A player who is on the floor longer, sees more plate appearances, or plays more rounds accumulates more of nearly everything at once. Positive correlation among components widens the composite's distribution relative to a naive independent calculation, because the good games are unusually good and the quiet games are unusually quiet. Other pairs push the other way. A pitcher's outs and his runs allowed are entangled through the same events, and a chart that rewards one while penalizing the other partially cancels, narrowing the spread.

The practical consequence is that you cannot get a composite's probability right by projecting each component's mean and adding them up. The mean will be fine. The distribution will not, and probability is a statement about the distribution, not the mean. This is the general point made in the correlation article applied inside a single offer instead of across an entry: the components share a game, so they share the game's randomness.

There is a related trap in how the shape of a composite differs from a single stat. A single counting stat is often lumpy and discrete, which is why pushes and integer thresholds matter so much on those markets. A composite built from several weighted components lands on a much finer grid and behaves more continuously, which changes where the mass sits near a line and how sensitive the probability is to a small move in the number. A researcher used to reading single stat distributions should expect fantasy totals to feel different, and should read them at their own scale rather than by analogy.

The chart is not a constant, and it is not yours

Nothing in this article states any platform's current weights, and that omission is deliberate. Scoring charts are revised. Platforms add stats, retire them, adjust coefficients when a market behaves in a way they did not intend, and occasionally run different charts for different contest types on the same app. A published chart is authoritative on the day you read it and on no other day.

This is also the cleanest pass rule in prop research. If you cannot confirm the weights, you cannot price the offer. Not with difficulty, not with a wide band: at all. Every other input can be estimated under uncertainty, but the chart is the function that turns your projection into the quantity being settled, and an unknown function has no uncertainty band because there is nothing to be uncertain around. A fantasy score line with an unverified chart is not a hard research problem. It is an undefined one.

How Slateline handles this

We model each platform's scoring as a separate versioned profile rather than as a conversion of some house standard. The projection is built from the components inside the same simulations that produce every other market for that player, so the correlation between components is present by construction rather than approximated afterward, and the composite's distribution comes out of the same simulated games rather than from a formula fitted to its mean.

Versioning matters as much as modeling. When a platform changes a chart, the old profile does not get quietly overwritten; a new version supersedes it with an effective date, so a projection graded against last month's settlement is graded under the chart that was actually in force. Grading a past result under a current chart would manufacture a record that never existed.

Where a chart is not verified against the platform's own published rules, we decline to settle it. Those offers carry an explicit flag rather than a confident probability, and they do not contribute to our published accuracy. That is the same discipline described in the platform differences article, applied to the one market where the platform does not just set the line but defines the unit.

Working rules for composite offers

• Read the chart on the platform before anything else, and confirm it applies to the contest type you are looking at.
• Project the components and apply the chart. Never scale a total from another platform.
• Expect the distribution to be shaped by correlation between components, not by the components taken one at a time.
• Compare fantasy lines across apps only after confirming the charts match, which is rarer than it sounds.
• Check the sold sides. Some fantasy offers exist over only, so a projection that dislikes the number produces a pass rather than the other direction.
• If the chart cannot be verified, pass. There is no version of this offer you can price without it.

Composite markets reward the least exciting part of research: reading the rules carefully and building from parts. They punish the habit that works elsewhere, which is carrying a good projection from one app to another. Slateline is independent of every platform named or unnamed here, and our interest in charts is purely mechanical, since a projection cannot be honest about a number whose definition it has guessed. The offers our projections are currently being compared against are on the Signal Board, the graded record is public in the Model Room, and if you participate anywhere, keep it inside limits you set first. Support is available at 1 800 GAMBLER and through the National Council on Problem Gambling.