12 months · same surface (Hard) · BO3 · before 2026-10-14. Historical rates, not forecasts or model confidence. Each server's rates sit beside the opponent's return evidence.
Tommy Paul: 27 played matches; 0 retirements retained where counts are valid; 0 walkovers excluded. Missing serve / opponent stats: 2 / 2. Invalid serve / opponent stats: 0 / 0.
Zizou Bergs: 26 played matches; 0 retirements retained where counts are valid; 0 walkovers excluded. Missing serve / opponent stats: 5 / 5. Invalid serve / opponent stats: 0 / 0.
0 target rows, 0 future rows, 0 rows with unproven timing, and 0 duplicate perspectives excluded.
Upcoming match: every stored result counts, since each was recorded before this match can start. Results dated after the match day and later rounds at this event are still excluded. Stored tournament dates can approximate match days; the 52-week boundary uses those dates. Current statistics and context may include later corrections and are not a historical snapshot.
Percentages use summed counts, with separate exposures for each metric. Ace rates use service points; aces conceded use return points. Second serve and return counts include double faults. Holds and breaks count games. A small sample means fewer than 10 observed matches and is a coverage flag.
Match structure: completed sets include sets finished before a retirement. Tiebreaks, straight sets, deciding sets and total games use completed matches only; "mid" is the middle half of total games. Service and return games are per observed match.
Loaded history since 2022-01-01; the History tab can load the full archive.
Hard · Outdoor · Best of 3. Published CPI: 32.8 (medium slow), season 2025 — older-season fallback. Source: user seed (unverified). This is the stored season value, not a reading for this match or week.
Season speed aggregates and retrospectively updated player records are not a fully historical snapshot. Speed is not used to select past evidence. Verified injury reports, weather, precise rest days and travel workload are unavailable. The Edge Lab found no useful model information beyond Pinnacle; model reports remain secondary research.