Dominating Tennis

Is momentum real in tennis, and how to stop losing streaks

Tell a real run from a coin flip. Stop a slide anyway.

Traced from real footage

See it, eight instants at a time.

Every frame is traced from real match footage, so you see the shape of the stroke, not a drawing of an idea.

Alcaraz’s serve, instant 1 of 8
FIG. 49 · ALCARAZ’S SERVE
Alcaraz’s serve: eight instants traced from the footage, frame for frame. Hover or tap to flip.

Key facts

  1. 2023

    A local audit of 334 points from the 2023 Wimbledon final found no disagreement with the MCP cross-check. In 20,000 seeded simulations with the real serving order and fitted serve-win rates, each observed run-count bar for lengths 1–20 fell inside its individual 5th–95th percentile band. This is not a joint test or proof of independence.

    scripts/momentum/build-final.mjs; seed 20230716; local audit 2026-10-01; commercial permission pending

  2. 70%

    Good win-probability models are calibrated: of all the moments they call 70 %, the player wins about 70 %.

    Kovalchik

  3. 1985

    Gilovich, Vallone and Tversky called the hot hand a cognitive illusion in 1985.

    Gilovich, Vallone & Tversky, 1985

  4. 1985

    Miller and Sanjurjo found a selection bias in the 1985 hot-hand analysis that partially reverses its conclusion.

    Miller & Sanjurjo, 2018

  5. 25s

    ITF Rule 29 sets a maximum of 25 seconds between points, ending when the first serve is struck. Rule 21 requires the receiver to follow the server’s reasonable pace; the server must wait for readiness. A reset is not an extra pause.

    ITF Rules of Tennis 2026, Rules 21 and 29

  6. 2023

    Carlos Alcaraz won the 2023 Wimbledon final 1–6, 7–6, 6–1, 3–6, 6–4. The held point log reconstructs the same five set scores.

    Roland-Garros, Reem Abulleil, 17 July 2023; COMAP/MCP score replay

  7. 07

    The fifth game of the third set lasted 32 points. Alcaraz broke Djokovic to lead 4–1. This game is separate from the six-point fifth-set game shown in the film study.

    Wimbledon, “26 Minutes and 32 Points”, official complete-game upload; COMAP/MCP game count

  8. 08

    At 1–1 in the fifth set, Djokovic lost the 30–40 point and Alcaraz led 2–1. Djokovic then struck his racquet against the net post. The action does not establish a causal momentum effect.

    COMAP/MCP point 297; reviewed ESPN Brasil aftermath; Roland-Garros match report, 17 July 2023

  9. 1992

    Klaassen and Magnus studied 481 Wimbledon singles matches from 1992–1995, excluding tiebreak points. Their model found small departures from independence and constant point-win probabilities.

    Klaassen & Magnus (2001), JASA 96(454), 500–509, sections 2 and 4.2–4.3

  10. 2025

    A prediction made before a point cannot use that point’s ace, winner or error. The 2025 momentum framework includes these current-point features; its reported performance does not establish an advance forecast.

    Du, Zhang & Zhou (2025), arXiv:2509.01243v1, Table 6 and sections 3.2.2, 3.4–3.5; local source review 2026-10-06

Why it works

With fixed serve-win probabilities, a tennis scoring model can calculate a match-win chance from each score and serving order. Point leverage is the difference between the match-win chances after winning and after losing the upcoming point. The ribbon’s actual jump records only the outcome that happened; it is not the full leverage.

Independence is an approximation to test. Klaassen and Magnus’s Wimbledon study found small departures from it. Our single-match run comparison cannot identify fatigue, changing tactics or psychological causes.

A slide can change your arousal and attention. Use the routine to give the next point a clear task: take the reset and name one outward change.

The line to say on court

“Read the run. Reset and choose one observable adjustment.”

Drill: The slide answer

20 min · with a partner
  1. Play a scored practice set. After three consecutive losses, mark one reset opportunity before the next point. Count once per losing run; restart the loss count after you win a point.
  2. Complete your reset and name one adjustment before the next point. When serving, leave time to strike within the applicable limit. When receiving, be ready at the server’s reasonable pace.
  3. Use a short cue, such as body serve, higher over the net or return cross. In this practice task, say it before play without delaying the next point.
  4. Count reset completed and adjustment named separately; both are needed to score the opportunity. Keep missed resets in the total and the next-point result separate. The four-opportunity score counts only after four opportunities.
Number to beat4 of 4 three-point slides answered with the full reset and one named change

More drills like this: mental game

Faults and fixes

Speeding up to get it back
Rushing can skip the preparation you use to choose a target.
FixKeep a brief, repeatable routine that fits the applicable clock and your serving or receiving role.
Abandoning the game plan after one run
Three consecutive losses can occur in an independent model; the run alone does not identify a tactical problem.
FixChange one variable and keep the plan.
Reading the opponent's fist pump as data
A fist pump does not measure the opponent’s point-win probability. Its absence from this model is not proof that behaviour has no effect.
FixReview the incoming ball, your target and the reply instead of diagnosing from the gesture.

Questions players ask

Does momentum exist in tennis?

A streak alone does not answer it. The comparison here tests one fixed-rate model and cannot establish causes.

What should I do after losing three points in a row?

Take your full between-point reset, then change one thing you can name, such as serve target or net height. Evaluate the next pattern rather than assuming the previous result determines it.

What does a big swing in a win-probability chart mean?

It means the score moved at a high-leverage moment. A crossing of the 50 % line isn't evidence that one point caused the next.

Is the hot hand a myth?

Gilovich, Vallone and Tversky called it an illusion in 1985. Miller and Sanjurjo later found a selection bias in that analysis that partially reverses the conclusion.

What this doesn't cover

The model uses constant serve-win rates estimated from this same match and preserves its serving order. Individual simulation bands are not a joint hypothesis test. It ignores changing tactics and fatigue; the result neither proves nor disproves momentum. Source data remains in local review pending commercial permission. The cited 2025 framework has unresolved table arithmetic and uses information unavailable before the predicted point. Its reported performance is not evidence that this lesson predicts the next point.

Sources

Read next

Know the gameWhat makes a big point in tennis: win probability and leverageTennis win probability is each player's live chance of winning the match, computed from the score and how often each wins points on serve. Leverage is how much one point moves that chance, and it's what makes a big point. A 30–40 break point at 5–6 in a deciding set carries huge leverage. 40–0 at 3–0 carries almost none.MindWhat to do between points in tennisBetween points, practise a short repeatable routine: release the last point, breathe, choose a target and begin your familiar service ritual. The four phases are a coaching structure. Timing depends on what you measure: a study of 18 men’s matches at the 2018 US Open found an average 21.6 seconds from point end to the next serve strike. That is an observation, not a routine target.MindHandling pressure in tennis: attention, routines and practicePressure can pull attention toward checking a familiar movement or away from the task into worry. These are two studied explanations, not a way to diagnose every miss. Try a practised target cue and familiar routine, then test them in more than one session. The studies here do not establish one routine that fixes both pathways in tennis.MindHow to play big points in tennisUse the score to understand the stakes, then choose a target and next-ball plan you can execute. The match model compares winning and losing the next point using both players’ assumed service point-win rates. It does not establish a universally best shot.