PH$ Market

These labels exist so a list of numbers can be scanned quickly. They describe a position in a distribution — not a judgement of a career.

Pending is not a low value. A player ProbHoops has not observed enough of carries no PH$ at all and is never placed in a band.

PH$ is a function of one input: where a player sits in his own season's AF-Player distribution. Nothing else enters it. There is no editorial adjustment, no reputation term and no manual override.

1. AF-Player — a descriptive 0–100 index of observed production, era-normalized against the player's own season. It is not a win probability.
2. Percentile — the player's position in that season's ranked field. Using percentile rather than the raw index is what keeps PH$ comparable across seasons that had different league distributions.
3. The curve — percentile raised to 1.8, mapped onto 5M–50M. The curve is convex on purpose: separation grows near the top of the observed percentile field without adding any salary or market input.

The index behind PH$ was not chosen by preference. Five formulations were declared in advance, fitted on 2019–2023 and gated before the 2024–2025 holdout was opened.

The first selection rule picked whichever candidate was most stable season to season. That rule turned out to be confounded: scoring is the most persistent thing a player carries between seasons, so a scoring-only index measured more stable (0.8213) than every composite. At the limit, that stability rule was effectively ranking by points, which is why it was replaced.

So stability became a floor and the objective became differentiation. The published formulation, efficiency_tilt_v1, correlates 0.5764 with a per-36 scoring ranking, against 0.7924 for the runner-up.