~/tennis.computer $
tennis.computer
The smartest, best-designed tennis stats on the web — player profiles, head-to-heads, tournament forecasts, and a surface-aware model doing the analysis.
this week
as of 2026-06-07
as of 2026-06-07
the model's world top 5
- 1Jannik Sinner2526
- 2Carlos Alcaraz2444
- 3Novak Djokovic2321
- 4Alexander Zverev2301
- 5Daniil Medvedev2201
the feed · latest atoms
All insights →Mover · 2026-06-07
+128
Matteo Arnaldi is the tour's biggest riser — up 128 Elo in eight weeks
Elo gained · 8 weeks
tennis.computer
Model read · 2026-06-07
+82 Elo
The model's clear No. 1: Sinner leads Alcaraz by 82 Elo
Sinner's Elo lead over Alcaraz
tennis.computer
Mover · 2026-06-07
2192
Rafael Jodar already carries a near-elite rating off just 34 tour matches
Elo · 34 tour matches
tennis.computer
Story · 2026-06-07
4
Four players have gained 100+ Elo in eight weeks
players +100 Elo in 8 weeks
tennis.computer
Discovery · 2026-06-07
Clay still lifts Tsitsipas most — the model rates him 142 Elo higher on the dirt
tennis.computer
Discovery · 2026-06-07
The model and the rankings disagree most about Gaël Monfils — 145 places apart
tennis.computer
held-out accuracy
log-loss
0.5907
lower is sharper
calibration error
0.0074
ECE · |pred − actual|
matches scored
32,091
held-out
Every probability is scored against what actually happened. Low log-loss means sharp predictions; a calibration error near zero means the percentages mean what they say.
How the model works & full track record →calibration curve
predicted % (x) vs actual win-rate (y) · dashed = perfect