Source: https://tennis-db.com/research/nick-kyrgios-against-the-best

# Nick Kyrgios Played His Best Against the Best

> His largest surplus over pre-match Elo expectations came against stronger opponents. Kyrgios won 25 matches against the Top 10, compared with about 15 expected.

- Published: 2026-10-02
- Updated: 2026-10-02
- Data as of: 2026-09-27
- Authors: [Mika Honkasalo](https://x.com/mhonkasalo)
- Canonical: https://tennis-db.com/research/nick-kyrgios-against-the-best
- Data: https://tennis-db.com/static/research/nick-kyrgios-against-the-best/results.csv
- PDF: https://tennis-db.com/static/research/nick-kyrgios-against-the-best/nick-kyrgios-against-the-best.pdf

## Key findings

- Kyrgios won 25 of 58 completed matches against opponents ranked in the Top 10, versus 14.91 expected. He ranks first by standardized surplus among 241 eligible ATP players since 1990.
- Against opponents more than 100 Elo points above him, Kyrgios won 27 of 63 matches, versus 13.03 expected. Against opponents more than 100 points below him, he won 122 of 159, versus 120.21 expected.
- After a Top 10 win, Kyrgios won 13 of 20 eligible next matches in the same tournament, versus 9.77 expected. His upsets were often followed by another victory.

## Abstract

Nick Kyrgios reached a career-high ranking of No. 13, yet repeatedly beat players near the top of the sport. The interesting question is how surprising those wins were given his own strength at the time. The [ATP's rankings history](https://www.atptour.com/en/players/nick-kyrgios/ke17/rankings-history) records that career high in October 2016.

This study compares completed ATP main-draw singles with a benchmark using **pre-match overall Elo, surface Elo, and match format**. Elo estimates strength from previous results. Adding the benchmark's win probabilities across matches gives expected wins. Top 10 status always means the opponent's ranking at match time.

Among 241 eligible players in the 1990-onward study, Kyrgios has the largest standardized surplus against Top 10 opponents: **25 wins from 58 matches, compared with 14.91 expected**. The player-level analysis then examines all 287 eligible Kyrgios matches in the September 27, 2026 snapshot.

## First in the Top 10 overperformance screen

The opening table ranks players by **standardized surplus**: actual wins minus expected wins, divided by the amount of variation the benchmark predicts for that set of matches. This accounts for both the number of opportunities and their difficulty. Eligibility requires at least 30 completed Top 10 matches, five distinct Top 10 opponents, and 100 matches against other ranked opponents.

| Rank and player | Wins / matches | Expected wins | Extra wins | Score |
|---|---:|---:|---:|---:|
| 1. Nick Kyrgios | 25 / 58 | 14.91 | +10.09 | 3.26 |
| 2. Carlos Alcaraz | 53 / 78 | 40.99 | +12.01 | 2.96 |
| 3. Marat Safin | 45 / 91 | 33.27 | +11.73 | 2.70 |
| 4. Jannik Sinner | 68 / 102 | 56.20 | +11.80 | 2.68 |
| 5. Alexei Popyrin | 12 / 33 | 6.26 | +5.74 | 2.67 |
| 6. Stan Wawrinka | 60 / 161 | 47.77 | +12.23 | 2.25 |
| 7. Dušan Lajović | 10 / 37 | 5.45 | +4.55 | 2.19 |
| 8. Gustavo Kuerten | 37 / 69 | 28.84 | +8.16 | 2.11 |
| 9. Francisco Cerúndolo | 16 / 38 | 10.48 | +5.52 | 2.06 |
| 10. Jaime Yzaga | 12 / 30 | 7.45 | +4.55 | 1.97 |

Score is standardized surplus; larger values indicate a greater positive departure from expectation.

Kyrgios won **43.1%** of these matches, against an expected **25.7%**. On a scale of 100 matches with the same mix of opponents and conditions, that is approximately **43 actual wins versus 26 expected**, a surplus of 17.4. He also leads this eligible sample on surplus per 100 matches, although he ranks seventh on the raw number of extra wins.

The No. 1 position is specific to the stated model and historical window. Starting the comparison in 2013, the first season in his sample, puts Kyrgios second among 81 eligible players, behind Wawrinka. Using a simpler overall-Elo benchmark puts him third in the historical cohort. The consistent result is a place near the top; the precise rank depends on the comparison.

## Stronger opponents brought the largest surplus

Official rankings identify the elite. Elo supplies a separate comparison with Kyrgios's own level before each match. Here, **stronger** means an opponent more than 100 overall-Elo points above him; **similar** means within 100 points; and **weaker** means more than 100 points below him. These groups cover all 287 matches.

| Opponent relative to Kyrgios | Wins / matches | Actual win rate | Expected win rate | Extra wins per 100 |
|---|---:|---:|---:|---:|
| Stronger | 27 / 63 | 42.9% | 20.7% | +22.2 |
| Similar | 39 / 65 | 60.0% | 49.6% | +10.4 |
| Weaker | 122 / 159 | 76.7% | 75.6% | +1.1 |

The stronger group produced **27 wins against 13.03 expected**. Against weaker opponents, the corresponding figures were **122 against 120.21**. His actual win rate still fell as opponents became stronger, but it fell much less than the benchmark anticipated.

That distinction changes the interpretation of his uneven reputation. Within this completed-match sample, his record against weaker players was close to expectation. His exceptional results came mainly from winning difficult matches more often than his rating suggested.

All three groups can exceed expectation because the benchmark is calibrated across the tour, rather than to Kyrgios's own career win rate. It underestimated him overall. The distinction of interest is how much larger his surplus was against stronger opponents, beyond that general tendency.

![Scatter plot of Kyrgios's extra wins per 100 matches against opponents' average pre-match Elo advantage. All 287 matches contribute to the plotted groups. Labels show sample sizes and vertical whiskers show approximate 95% intervals, with substantial uncertainty in the small groups.](/static/research/nick-kyrgios-against-the-best/opponent-strength.svg)

The scatter divides the strength gap into 100-point groups and places each dot at the group's average gap. The tails pool gaps below -500 and at least +500. **All 287 matches are included**, regardless of how often he faced an individual opponent. Labels show the number of matches, and whiskers show approximate uncertainty around each group's surplus. The upper tail contains only two matches and has a wide interval; retaining it makes the full sample visible without treating that point as a reliable pattern on its own.

Changing the main grouping threshold to 75 or 150 Elo points preserves the broad contrast. The stronger group remains 15.9 to 22.9 wins per 100 above expectation; the weaker group ranges from -0.9 to +2.7. The middle group's result is more sensitive to where its boundaries are drawn.

Against ranked opponents outside the Top 10, Kyrgios was also above expectation overall: **162 wins in 227 matches, versus 148.85 expected**. That group includes some players stronger than him at the time, so it should not be equated with the weaker-Elo group. Across all opponents, he won **188 of 287 matches, versus 165.50 expected**, a surplus of 7.8 wins per 100.

## The matchups behind the reputation

Across completed meetings with Federer, Nadal, and Djokovic at individual tour events, Kyrgios went **6–10**, against just **2.16 expected wins**. Those 16 matches contribute to the aggregate even though the individual series are small.

The table shows his five most frequent opponents in the completed-match sample, plus Federer, Djokovic, and Medvedev as references for his record against the elite. This is a selected comparison panel, with no minimum meeting count. Sample size determines how much weight an individual record deserves.

| Opponent | Matches | Kyrgios W-L | Expected wins | Extra wins |
|---|---:|---:|---:|---:|
| Rafael Nadal | 9 | 3–6 | 1.31 | +1.69 |
| Andy Murray | 7 | 1–6 | 0.87 | +0.13 |
| Milos Raonic | 7 | 3–4 | 2.05 | +0.95 |
| Richard Gasquet | 7 | 2–5 | 2.19 | -0.19 |
| Alexander Zverev | 6 | 4–2 | 2.93 | +1.07 |
| Daniil Medvedev | 5 | 4–1 | 1.78 | +2.22 |
| Roger Federer | 4 | 1–3 | 0.49 | +0.51 |
| Novak Djokovic | 3 | 2–1 | 0.36 | +1.64 |

Murray provides a useful counterpoint. A 1–6 record looks like a difficult personal matchup, but the benchmark expected fewer than one Kyrgios win. Gasquet's 5–2 advantage also sits close to expectation. Neither series supplies a substantial negative residual once the players' strength at the time is considered.

Medvedev is an especially striking smaller sample: Kyrgios won four of five completed meetings, against **1.78 expected**. The [ATP's review of their 2022 rivalry](https://www.atptour.com/en/news/best-of-2022-rivalries-medvedev-kyrgios/) documents the Montreal and US Open victories, both against the reigning No. 1, and describes the attacking, serve-and-volley approach used in Montreal. That offers a plausible tactical account of a particular match; this study does not identify the cause of the broader surplus.

Medvedev also contributed the largest share of Kyrgios's Top 10 surplus. Even after removing those meetings, Kyrgios has **22 wins against 13.61 expected** in the remaining 54 Top 10 matches.

## A pattern across years, with a format split

The Top 10 surplus predates Kyrgios's 2022 Wimbledon final season. Removing 2022 leaves **19 wins against 11.47 expected**. Removing 2016, the season contributing the largest surplus, leaves **19 against 11.82**. It also persists after restricting the sample to matches in which both players had at least 20 previous observed main-tour appearances.

![Cumulative Kyrgios wins against Top 10 opponents and cumulative expected wins by season from 2013 through 2022. The actual line finishes at 25 and the expectation line at 14.91, with the gap developing across multiple seasons.](/static/research/nick-kyrgios-against-the-best/top-ten-timeline.svg)

Match format reveals a sharper distinction:

| Format, against Top 10 opponents | Wins / matches | Expected wins | Extra wins per 100 |
|---|---:|---:|---:|
| Best of three | 21 / 39 | 12.07 | +22.9 |
| Best of five | 4 / 19 | 2.85 | +6.1 |

Most of the excess Top 10 victories came in best-of-three matches. The best-of-five sample remained above expectation, but by a much smaller amount. Surface splits were positive on hard courts, grass, and clay, although the latter two contained only 12 and eight Top 10 matches, respectively.

His broader performance also changed over time. Across all opponents, 2013–2016 produced 69 wins against 50.54 expected; 2017–2021 produced 81 against 78.65. Early improvement can run ahead of a rating built from previous matches. That makes the persistence of the later Top 10 surplus more informative than a career-wide total alone.

## What happened after the upset?

The next match provides a direct check on whether the big wins tended to stand alone. After each Top 10 victory, this analysis locates Kyrgios's immediately next recorded match in the same tournament edition and evaluates it using its own pre-match expectation.

There were **20 eligible completed next matches**. Kyrgios won **13**, compared with **9.77 expected**: a 65.0% win rate against an estimated 48.8%. Two Top 10 victories ended title runs and had no next match; two were followed by walkovers and one by a retirement, which remain outside the completed-match comparison.

The aggregate includes both outcomes familiar from his biggest runs. After defeating Medvedev at the 2022 US Open, he lost to Karen Khachanov. After defeating Nadal in Acapulco in 2019, he beat Wawrinka in the next round. Across the eligible sample, there is no observed collapse below the benchmark immediately after a Top 10 win. Twenty matches cannot establish a general response to winning, but they are enough to make the next-match question worth checking rather than assuming its answer.

## Conclusion

Kyrgios's strongest statistical distinction is the frequency with which he won matches that looked difficult beforehand. He leads the stated historical Top 10 screen and remains near its top under alternative comparisons. The larger surplus against stronger opponents coexists with roughly expected performance against weaker ones, and many of his elite wins were followed by another victory.

This describes the results he produced in completed matches. It does not estimate the career he could have had with a different schedule, fewer interruptions, or different choices.

## Methodology

### Scope and expected wins

The frozen September 27, 2026 ATP extract supplies 102,326 completed main-draw singles matches since 1990 at individual main-tour events, including the Olympics. Team events, qualifying, lower tours, exhibitions, walkovers, retirements, defaults, and incomplete results are excluded. Kyrgios contributes 287 completed matches dated May 29, 2013 through June 22, 2026. The same source population contains 15 retirements, nine walkovers, and one default involving him that do not enter the study. These definitions differ from full career records that include team events and awarded results.

Two Kyrgios matches lack an opponent ranking and are excluded from the Top 10-versus-other comparison, but included in all-match and Elo-based analyses. All 58 eligible Top 10 meetings occur between 2013 and 2022. Rankings and ratings are taken before the match, never from career peaks or current standings.

For each calendar year, a symmetric logistic model is fitted only to the preceding ten seasons. It uses overall and surface Elo differences, scaled by 400, and their interactions with best-of-five format. A unit ridge penalty stabilizes the four coefficients. Missing surface Elo falls back to overall Elo. The existing warehouse Elo histories are reused; earlier awarded results can therefore affect later ratings even when excluded from the study outcomes.

Expected wins are the sum of individual probabilities. Extra wins are actual minus expected wins. Standardized surplus divides that difference by the square root of the sum of each probability multiplied by one minus that probability. Kyrgios's aggregate and scatter include every eligible match, with no minimum count per opponent or strength group. The historical player screen retains the eligibility thresholds stated above so that each compared player has substantial exposure to Top 10 and other opponents.

### Uncertainty and sensitivity

Scatter whiskers use approximate Wilson score intervals for each group's observed win proportion, then subtract its mean expected win proportion to express the interval on the surplus-per-100 scale. The expected benchmark is held fixed. These intervals assume independent matches and do not include model-estimation uncertainty or career-phase effects; they primarily show how imprecise the sparse groups are.

Conditional on the estimated probabilities and independent results, Kyrgios's 58 Top 10 matches produce a 95% prediction range of **nine to 21 wins**, versus 25 observed. However, he was selected after screening 241 players. His two-sided Poisson-binomial probability is approximately 0.0028 before adjustment and **0.678 after a conservative Holm correction** across that screen. The ranking is a descriptive discovery, not family-wide statistical confirmation of a stable personal trait.

The Top 10 surplus exceeds his surplus against other ranked opponents by 11.6 wins per 100 matches. A 20,000-replicate bootstrap resampling his 13 observed calendar seasons gives a 95% percentile interval of **-3.4 to +25.5** for that difference. This wide interval, based on few seasons, does not account for selecting Kyrgios from the earlier screen.

The historical ranking is checked with the simple overall-Elo probability and with a 2013-onward comparison window. The player pattern is checked at 75-, 100-, and 150-point relative-strength thresholds, after removing the strongest contributing season or opponent, after excluding early-career meetings, and after excluding gaps above 180 days. The gap restriction removes none of his eligible Top 10 meetings.

A separate calibration check groups other tour appearances since 2013 into predicted-probability deciles, separately for Top 10 and other opponents, excluding matches involving Kyrgios. Weighting those residuals by Kyrgios's own decile distribution leaves approximately **10.29 extra Top 10 wins**, close to the original 10.09. This argues against a generic underdog calibration error explaining the entire result; it is not a fully matched comparison.

### Limitations

Elo can lag improvement and retain stale strength after inactivity. Opponent choice, surface, format, tournament stage, and career phase are interrelated. The subgroup comparisons are descriptive, and the best-of-five contrast is not an isolated effect of match length. Excluding retirements and defaults changes the question to performance in completed matches and omits part of availability and durability. Neither conditional prediction ranges nor the bootstrap include all model uncertainty.

The scatter aggregates different opponents and periods; its small groups are volatile and should not be read as a fitted response curve. The next-match sample is also selected: it requires first winning a Top 10 match and then having an eligible recorded successor. It measures neither effort nor motivation. Ranking position depends on participation and ranking rules as well as match performance, so the analysis cannot translate the surplus directly into a hypothetical career ranking.

## Data and attribution

Historical ATP results are supplied by Jeff Sackmann's tennis data project and Tennis Abstract, with current-era coverage from BALLDONTLIE, reconciled through tennis-db.com's ATP serving data. The frozen predictions, analysis specification, aggregate tables, and reproducible code are preserved in the accompanying research record.

Study questions and suggestions for future Research can be sent to [contact@tennis-db.com](mailto:contact@tennis-db.com) or through the [contact form](/contact).

## Citations

- [Jeff Sackmann and Tennis Abstract, historical ATP data](https://github.com/JeffSackmann/tennis_atp)
- [BALLDONTLIE tennis data documentation](https://www.balldontlie.io/docs/)
- [ATP Tour, Nick Kyrgios rankings history](https://www.atptour.com/en/players/nick-kyrgios/ke17/rankings-history)
- [ATP Tour, Medvedev and Kyrgios rivalry in 2022](https://www.atptour.com/en/news/best-of-2022-rivalries-medvedev-kyrgios/)
- [R statistical documentation, Holm adjustment for multiple comparisons](https://stat.ethz.ch/R-manual/R-devel/library/stats/html/p.adjust.html)

Questions about this study or requests for future tennis-db.com Research can be sent to [contact@tennis-db.com](mailto:contact@tennis-db.com) or through the [contact form](https://tennis-db.com/contact).