3rdServe
ATP

Clutch index

How much better a player performs on break points than on ordinary points — against their own baseline, not the field's. Saving 70% of break points isn't clutch if you win 78% of your service points normally; it's an 8-point drop under pressure.

serve_clutch = BPsaved ÷ BPfaced − svPtsWon ÷ svPts
return_clutch = BPconv ÷ BPchances − rtnPtsWon ÷ rtnPts
CLUTCH_pp = (serve_clutch × BPfaced + return_clutch × BPchances) ÷ (BPfaced + BPchances) × 100
clutch_index = 100 + CLUTCH_pp
Window
Career (full archive) and rolling 18 months, computed for both.
Rung
A — ATP, 1968→, all levels. Needs full serve statistics.
Floor
100 total break points (faced + created). Below it no rating is issued and the UI says so.
Cannot
It cannot say why a player is clutch, and it cannot weight a break point at 5–5 in a final above one at 5–0 in a first round. Leverage weighting arrives with the point-level ingest, behind the same interface.
Source
Method after Sackmann, “Measuring the Clutchness of Everything”.

The Cannot line appears on every metric in this product. Stating a measure's limits is the cheapest credibility available, and it is the reason these numbers are computed here rather than taken from someone else.