Findings / Tennis
Tennis: where momentum lives, and where it doesn’t
Tennis momentum is real point-to-point but dies at the game grain; two tiebreak myths are null; altitude and travel effects are confirmed (altitude replicated across two disjoint year slices).
These are our own preregistered tennis tests, surfaced verbatim from the committed forward-claim scoreboard. A CONFIRMED_LOCAL/REPLICATED/NULL_LOCAL verdict is an accuracy/effect finding, not a advantage.
Momentum has a grain limit
Momentum is real at the point grain (CONFIRMED_LOCAL, n=456383) but the same idea dies once you zoom out to game grain: whether a player holds serve after breaking is NULL_LOCAL, confirmed null on both split halves.
| Claim | Verdict | n | Effect | Corpus |
|---|---|---|---|---|
| Point to point momentumReal, small effect: winning the last point raises win probability on the very next point. | CONFIRMED_LOCAL | 456,383 | 0.0216 | slam_points_2011_2015 |
| Post break hold rateDead at game grain: no detectable effect on whether a player holds serve after breaking. | NULL_LOCAL | 21,208 | 0.0051 | slam_points |
| Post break hold rate split ASplit-half corroboration (first half): still null. | NULL_LOCAL | 10,522 | 0.0067 | slam_points__split_half |
| Post break hold rate split BSplit-half corroboration (second half): still null. | NULL_LOCAL | 10,686 | 0.0036 | slam_points__split_half |
Scroll horizontally for all columns.
How to read this — momentum is real point-to-point (CONFIRMED_LOCAL on n=456,383) but goes null the moment you zoom out to the game grain (post-break hold rate, NULL_LOCAL on n=21,208). Same idea, different grain, opposite verdict.
Two tiebreak myths, both null
Two popular tiebreak beliefs -- that some players have a persistent tiebreak-clutch skill, and that serving first in the breaker matters -- both come back NULL_LOCAL.
| Claim | Verdict | n | Effect | Corpus |
|---|---|---|---|---|
| Tiebreak skill persistence partialNo evidence that tiebreak skill persists as a stable player trait. | NULL_LOCAL | 206 | 0.0745 | matches_asof_setdetail_schedule_density_atp_wta |
| Tiebreak serve order win rateThe 'serve first in the breaker' myth: no detectable advantage from serve order. | NULL_LOCAL | 1,281 | 0.0035 | slam_points |
Scroll horizontally for all columns.
Altitude and travel, confirmed
Altitude suppresses serve ace rate and long travel suppresses win probability, both CONFIRMED_LOCAL; the altitude effect was independently REPLICATED on two disjoint year slices (2015-2020 and 2021-2025).
| Claim | Verdict | n | Effect | Corpus |
|---|---|---|---|---|
| Altitude effect on serve ace rateAltitude lowers ace rate (thinner air behaves differently than folklore predicts). | CONFIRMED_LOCAL | 27,463 | -0.0083 | travel_scouting_matches_atp_wta_2015_2025 |
| Altitude effect on serve ace rate replication years 2015 2020Replicates on the 2015-2020 year slice. | REPLICATED | 14,329 | -0.0078 | travel_scouting_disjoint_year_slice |
| Altitude effect on serve ace rate replication years 2021 2025Replicates independently on the 2021-2025 year slice too. | REPLICATED | 13,134 | -0.0089 | travel_scouting_disjoint_year_slice |
| Long travel effect on win prob partialLong travel measurably lowers win probability. | CONFIRMED_LOCAL | 26,950 | -0.0699 | travel_scouting_matches_atp_wta_2015_2025 |
Scroll horizontally for all columns.
- effect signs are raw directional effects on the stated metric, not opponent/surface-adjusted;
- corpora differ (slam point-level 2011-2015 vs match-level 2015-2025); each claim shows its own corpus;
- CONFIRMED/NULL are OUR preregistered leak-free verdicts vs an internal baseline -- descriptive accuracy findings, never a money-advantage claim.
CONFIRMED here means a leak-free accuracy/effect finding against our own baseline. Nulls are published with equal weight: a myth that fails to show an effect is as valid a result as one that confirms.