What it measures
An illustrative dollar value for a player's card, plus a five-year trajectory. It is a model estimate derived from Balladex's own projection data. It is not a marketplace quote, it is not scraped from completed sales, and it is not investment advice.
The model exists because the two things that drive a real card market pull in opposite directions. A young riser is priced on what he might become. A retired legend is priced on what he already is, and that price keeps climbing while his production sits at zero. A single formula cannot describe both, so the model runs both and takes whichever is higher.
The two forces
- Legacy floor
- Scarcity and all-time stature. It only rises. A legend's card keeps climbing even as his production falls to nothing, which is why the floor — not performance — is what prices a retired player.
- Performance value
- This season's projected production plus hardware actually won. This is the force that moves for an active player, and the one the what-if lab lets you push.
The two are combined with a soft maximum rather than a hard one, so a player crossing from one regime to the other moves smoothly instead of jumping.
From index to dollars
Both forces resolve to a single index, roughly 0 to 115. The index becomes a dollar figure through an exponential curve, because card markets are exponential — the gap between a star and a role player is not additive.
value = 0.9 × e^(0.084 × index)The index is clamped to 0–125 before the curve is applied. The curve is calibrated so that:
| Index | Modeled value |
|---|---|
| 30 | ≈ $11 |
| 50 | ≈ $60 |
| 75 | ≈ $490 |
| 90 | ≈ $1.7k |
| 100 | ≈ $4k |
What hardware is worth, and where
This is the most commonly misread part of the model, so it is worth being exact. The weights below are what the what-if lab applies when you add an award to a player's line yourself. They are not applied to the published value: the server model passes a hardware boost of zero, so adding an MVP does not raise the number you see on the card market by 30%.
Hardware does reach the published value, but through the other side of the model — the legacy floor, where championships, MVPs and All-NBA selections feed the all-time stature component. The effect is real but indirect, and smaller than the table below implies.
| Accolade | Boost |
|---|---|
| MVP | +30% |
| Finals MVP | +13% |
| All-NBA | +12% |
| Champion | +11% |
| DPOY | +8% |
| Scoring title | +7% |
| All-Star | +6% |
Why the numbers match the what-if lab exactly
The arithmetic lives in one place — a module of pure functions with no data access of its own. The server model and the in-browser what-if lab both import it, so dialling in "30 points a game and an MVP" recomputes with the identical formula rather than an approximation of it. There is no second implementation to drift.
How the football side works
Differently from the basketball side, and the difference is worth knowing before quoting a number.
The projection engine that drives the basketball performance value forecasts minutes and per-minute points, rebounds and assists — none of which a football season produces. Football production comes instead from a position-aware grade: a percentile against players at the same position, so a quarterback is measured against quarterbacks. That grade is then quantile-matched onto the basketball distribution, so a football player at the 60th percentile of football is priced like a basketball player at the 60th percentile of basketball. Matching the distributions is what putting two sports on one curve actually requires; mapping the grade straight onto the points axis is not, and an earlier version of this model did exactly that and priced the median footballer like a 23-point NBA star.
The grade is also shrunk toward the previous season rather than taken from the most recent one alone, because a single season of it moves around a great deal — the year-to-year swing has a standard deviation of about ten points on a scale that effectively runs from 38 to 98.
Why the football outlook carries no trend
Because the direction of a player's grade does not predict the next one, and we checked rather than assumed.
Across the 966 football players with three graded seasons, consecutive year-over-year changes in grade correlate at -0.37. Players whose grade rose by six points or more in one year fell by an average of 5.4 the next; those who fell by six or more rose by an average of 3.7. That is mean reversion, which is what a percentile measure usually does.
An earlier version of this model read a rising or falling direction out of that change and extrapolated it across five years. It was confidently projecting the one direction the data says reverses. The five-year window now carries an age slope only, and no claim about momentum. A football card's outlook here tells you where a player sits and roughly what age does to that; it does not tell you he is heating up.
What this is not
Not a price. Nothing here is drawn from eBay, auction results, or any marketplace. A real card's price depends on grade, print run, set, serial number and who else wants it that week — none of which this model sees.
Treat it as a ranking with a dollar sign attached: useful for asking who is rising and who is cooling, not for deciding what to pay.