The Clutch Index · 1991–2024

Nerve, marked point by point

The umpire records every point; almost nobody adds them up correctly. Across 472 careers, break points saved tracks how well a man serves on ordinary points at R² 0.81, slope 0.98. It is mostly a serving statistic.

Deuce

Game

One game, as the chair records it. S is a point to the server, R to the returner. The pencilled column is the break point — the only column this index counts.

Seasons: 1991–2024
Matches measured: 93,720
Players ranked: 472
Scale: 50 ± 10

What it found

Greatness is worth

2.2 points

Players who reached the world's top five average 51.5. Players who never got past 50th average 49.3. Being one of the best of the era buys almost nothing under pressure.

Deciding sets contain

Nothing

The most-quoted clutch statistic in the sport measures 0.0% real signal. Players are bunched more tightly than chance alone would produce, so it earns no weight in the rating.

Most lopsided player

Isner

John Isner stands +2.32 standard deviations above average in tiebreaks — the highest of any player in any component — and -1.38 on return. No one else in the archive splits that far between serving and returning.

At the other end: Gasquet, Enqvist, Monaco close the list of 472.

Where the famous names land

of 472 ranked

Read the columns, not just the rank. The gaps between what these players did under pressure and what the rest of their game predicted are one or two points in a hundred — this is a measure of small margins, not of nerve failing. Federer, for one, came in above expectation on serve and in tiebreaks.

#	Player	Clutch	Serve	Return	Tiebreaks
5	Rafael Nadal	77.3	+1.6	+0.4	-1.2
16	Nick Kyrgios	70.2	+1.4	+2.9	+6.6
20	Pete Sampras	68.1	+0.8	+0.3	+7.2
38	Novak Djokovic	64.1	+0.9	-0.3	+5.1
57	Andre Agassi	61.4	+1.2	-0.3	+0.3
61	John Isner	60.5	+1.7	-2.6	+9.2
329	Roger Federer	44.9	+0.6	-1.2	+3.9
446	Andy Murray	35.2	-0.3	-1.0	+2.7

Columns are percentage points above or below what each player's own ordinary play predicted. Correlation between how good a player was and his clutch rating: r = 0.046 — near zero, so the index is not simply marking down the great.

The problem with the usual number

Above the model Below the model Predicted by ordinary play

Each of the 472 dots is a qualifying career. The stem is the distance from the model — that distance, and nothing else, is what this page ranks.

The leaderboard · top 25

All 472 players ›

Era filters hide rows. They never re-rank: a player's rating is his standing against his own contemporaries either way.

#	Player	Career	Matches	Clutch	Strength
01	Goran Ivanisevic	1991–2004	710	82.8	returning, tiebreaks
Serving +1.0 66.3% v 65.4% expected over 3,806 break points faced Returning +1.9 39.6% v 37.7% expected over 4,373 break points against Tiebreaks +7.1 59.6% v 52.6% expected over 384 tiebreaks Deciders · counts 0 -0.2 59.4% v 59.6% expected over 214 deciding sets Career 464–246 Best rank 2 Best season 1996 · 68.0 282 143 84
02	Feliciano Lopez	1998–2023	943	81.1	serving, tiebreaks
Serving +2.1 65.5% v 63.4% expected over 5,829 break points faced Returning -0.6 35.3% v 35.9% expected over 5,327 break points against Tiebreaks +5.5 54.2% v 48.7% expected over 585 tiebreaks Deciders · counts 0 +3.1 53.8% v 50.7% expected over 338 deciding sets Career 476–467 Best rank 12 Best season 2010 · 72.3 525 260 136
03	Thomas Muster ◂	1991–2011	584	80.6	serving, returning
Serving +2.0 62.9% v 60.9% expected over 3,839 break points faced Returning +0.9 44.8% v 43.9% expected over 4,943 break points against Tiebreaks +3.5 58.0% v 54.5% expected over 193 tiebreaks Deciders · counts 0 +1.9 67.6% v 65.7% expected over 176 deciding sets Career 417–167 Best rank 1 Best season 1997 · 74.7 163 355 13
04	Jo-Wilfried Tsonga	2004–2022	651	80.3	serving, returning
Serving +1.8 67.1% v 65.3% expected over 3,387 break points faced Returning +1.2 39.8% v 38.6% expected over 4,272 break points against Tiebreaks +3.3 57.8% v 54.4% expected over 374 tiebreaks Deciders · counts 0 -1.4 62.1% v 63.5% expected over 198 deciding sets Career 436–215 Best rank 5 Best season 2014 · 68.7 441 128 72
05	Rafael Nadal ▸	2002–2024	1,232	77.3	serving
Serving +1.6 66.0% v 64.4% expected over 6,339 break points faced Returning +0.4 45.0% v 44.6% expected over 10,979 break points against Tiebreaks -1.2 61.0% v 62.2% expected over 413 tiebreaks Deciders · counts 0 -4.0 69.6% v 73.6% expected over 263 deciding sets Career 1022–210 Best rank 1 Best season 2013 · 68.2 624 508 94
06	Nicolas Massu	1997–2013	409	76.8	serving
Serving +2.5 61.4% v 59.0% expected over 3,139 break points faced Returning +0.7 41.7% v 41.0% expected over 2,923 break points against Tiebreaks +2.8 52.3% v 49.5% expected over 174 tiebreaks Deciders · counts 0 -3.9 45.2% v 49.1% expected over 135 deciding sets Career 204–205 Best rank 11 Best season 2009 · 74.1 139 237 17
07	Alexander Volkov ◂	1991–1997	379	76.2	returning, serving
Serving +1.6 60.6% v 59.0% expected over 2,967 break points faced Returning +2.3 43.1% v 40.8% expected over 2,830 break points against Tiebreaks -0.1 48.9% v 49.0% expected over 141 tiebreaks Deciders · counts 0 +3.7 57.7% v 54.0% expected over 156 deciding sets Career 208–171 Best rank 14 Best season 1992 · 73.5 168 58 43
08	Jarkko Nieminen	2001–2015	665	75.7	serving
Serving +1.4 60.6% v 59.2% expected over 4,977 break points faced Returning +0.6 41.5% v 41.0% expected over 4,919 break points against Tiebreaks +4.2 54.1% v 49.9% expected over 290 tiebreaks Deciders · counts 0 +0.3 52.1% v 51.8% expected over 242 deciding sets Career 347–318 Best rank 13 Best season 2006 · 81.6 414 179 53
09	Mikhail Kukushkin ▸	2007–2024	343	75.4	serving, returning
Serving +2.4 60.6% v 58.2% expected over 2,789 break points faced Returning +1.5 41.0% v 39.5% expected over 2,318 break points against Tiebreaks -2.8 43.8% v 46.7% expected over 146 tiebreaks Deciders · counts 0 -2.4 41.9% v 44.3% expected over 105 deciding sets Career 140–203 Best rank 40 Best season 2020 · 68.7 219 85 38
10	Dmitry Tursunov	2001–2017	418	74.2	serving
Serving +1.9 62.6% v 60.7% expected over 3,022 break points faced Returning +1.0 39.8% v 38.9% expected over 2,739 break points against Tiebreaks +4.9 53.9% v 49.0% expected over 193 tiebreaks Deciders · counts 0 +1.4 53.5% v 52.1% expected over 155 deciding sets Career 216–202 Best rank 20 Best season 2014 · 77.8 270 70 64
11	Andreas Vinciguerra	1998–2013	171	74.0	serving
Serving +4.6 64.1% v 59.5% expected over 1,308 break points faced Returning +2.4 42.4% v 40.0% expected over 1,093 break points against Tiebreaks -5.0 43.6% v 48.6% expected over 78 tiebreaks Deciders · counts 0 +7.3 55.6% v 48.2% expected over 45 deciding sets Career 81–90 Best rank 33 Best season 2000 · 77.2 91 64 7
12	Martin Klizan	2007–2020	254	72.1	serving, returning
Serving +2.2 60.5% v 58.3% expected over 2,029 break points faced Returning +2.0 41.7% v 39.7% expected over 1,774 break points against Tiebreaks +4.2 51.7% v 47.6% expected over 114 tiebreaks Deciders · counts 0 +3.0 52.5% v 49.5% expected over 99 deciding sets Career 122–132 Best rank 25 Best season 2019 · 69.2 125 114 15
13	Marcos Ondruska	1991–2001	231	71.9	serving, returning
Serving +2.4 58.5% v 56.0% expected over 2,041 break points faced Returning +2.2 44.6% v 42.4% expected over 1,731 break points against Tiebreaks -4.8 42.5% v 47.3% expected over 73 tiebreaks Deciders · counts 0 +4.1 48.8% v 44.7% expected over 82 deciding sets Career 99–132 Best rank 27 Best season 1995 · 68.7 138 52 26
14	Carlos Moya	1995–2010	835	70.6	serving, tiebreaks
Serving +1.8 63.0% v 61.2% expected over 5,704 break points faced Returning -0.8 40.7% v 41.5% expected over 6,524 break points against Tiebreaks +4.5 57.0% v 52.5% expected over 365 tiebreaks Deciders · counts 0 +0.3 59.9% v 59.6% expected over 294 deciding sets Career 531–304 Best rank 1 Best season 2007 · 72.9 336 445 27
15	Gilles Muller	2002–2018	394	70.4	serving, tiebreaks
Serving +2.4 66.7% v 64.3% expected over 2,365 break points faced Returning -0.2 34.4% v 34.5% expected over 1,979 break points against Tiebreaks +6.0 54.7% v 48.8% expected over 274 tiebreaks Deciders · counts 0 +3.1 52.8% v 49.7% expected over 123 deciding sets Career 193–201 Best rank 21 Best season 2005 · 72.9 262 47 78
16	Nick Kyrgios	2013–2023	279	70.2	returning, tiebreaks
Serving +1.4 67.8% v 66.4% expected over 1,324 break points faced Returning +2.9 39.2% v 36.3% expected over 1,615 break points against Tiebreaks +6.6 59.8% v 53.3% expected over 204 tiebreaks Deciders · counts 0 -4.2 58.0% v 62.1% expected over 88 deciding sets Career 186–93 Best rank 13 Best season 2022 · 75.8 190 42 47
17	Potito Starace	2003–2014	321	69.3	serving
Serving +2.4 61.3% v 59.0% expected over 2,358 break points faced Returning +0.8 40.0% v 39.3% expected over 2,068 break points against Tiebreaks -1.0 46.5% v 47.5% expected over 129 tiebreaks Deciders · counts 0 +0.8 46.3% v 45.5% expected over 108 deciding sets Career 141–180 Best rank 27 Best season 2008 · 66.0 81 223 9
18	Vincent Spadea	1993–2010	633	68.8	serving
Serving +1.7 59.9% v 58.3% expected over 5,039 break points faced Returning +0.4 41.9% v 41.5% expected over 4,613 break points against Tiebreaks -8.4 41.5% v 49.8% expected over 229 tiebreaks Deciders · counts 0 +3.6 50.5% v 46.9% expected over 212 deciding sets Career 294–339 Best rank 19 Best season 2000 · 70.3 378 150 65
19	Felix Mantilla	1994–2007	510	68.6	returning
Serving +0.4 59.1% v 58.7% expected over 3,785 break points faced Returning +1.8 44.4% v 42.6% expected over 3,996 break points against Tiebreaks +3.0 53.6% v 50.6% expected over 192 tiebreaks Deciders · counts 0 +1.2 57.1% v 55.9% expected over 161 deciding sets Career 299–211 Best rank 11 Best season 1998 · 64.1 127 348 13
20	Pete Sampras	1991–2002	782	68.1	tiebreaks
Serving +0.8 67.9% v 67.2% expected over 3,626 break points faced Returning +0.3 40.6% v 40.3% expected over 6,003 break points against Tiebreaks +7.2 63.9% v 56.7% expected over 413 tiebreaks Deciders · counts 0 -2.9 69.1% v 72.0% expected over 217 deciding sets Career 628–154 Best rank 1 Best season 2000 · 66.7 429 115 104
21	Fernando Gonzalez	1999–2012	518	68.1	tiebreaks, serving
Serving +1.1 63.8% v 62.7% expected over 3,162 break points faced Returning +0.7 40.2% v 39.5% expected over 3,730 break points against Tiebreaks +5.9 58.7% v 52.8% expected over 230 tiebreaks Deciders · counts 0 -1.9 58.6% v 60.5% expected over 174 deciding sets Career 333–185 Best rank 5 Best season 2002 · 69.8 248 210 35
22	Jan Siemerink	1991–2002	499	67.8	serving
Serving +2.7 64.8% v 62.1% expected over 3,353 break points faced Returning -1.5 36.5% v 38.0% expected over 3,068 break points against Tiebreaks +3.0 52.8% v 49.8% expected over 248 tiebreaks Deciders · counts 0 +0.8 50.9% v 50.1% expected over 173 deciding sets Career 252–247 Best rank 14 Best season 1997 · 68.3 236 78 55
23	Bernard Tomic	2009–2021	331	67.4	serving
Serving +2.6 63.8% v 61.2% expected over 2,312 break points faced Returning -0.6 36.4% v 37.0% expected over 2,058 break points against Tiebreaks +4.4 52.7% v 48.3% expected over 186 tiebreaks Deciders · counts 0 +0.4 52.3% v 51.8% expected over 111 deciding sets Career 167–164 Best rank 17 Best season 2015 · 72.4 214 54 63
24	Albert Ramos ▸	2010–2024	593	67.3	serving
Serving +1.8 60.8% v 59.0% expected over 4,471 break points faced Returning +0.4 40.3% v 39.9% expected over 4,130 break points against Tiebreaks -10.0 37.9% v 47.9% expected over 253 tiebreaks Deciders · counts 0 +1.5 48.4% v 46.9% expected over 221 deciding sets Career 267–326 Best rank 17 Best season 2023 · 67.0 205 364 24
25	Lucas Pouille ▸	2013–2024	258	66.9	serving
Serving +2.3 62.7% v 60.4% expected over 1,889 break points faced Returning +0.5 38.7% v 38.2% expected over 1,704 break points against Tiebreaks +5.3 54.7% v 49.4% expected over 137 tiebreaks Deciders · counts 0 +4.9 56.6% v 51.8% expected over 83 deciding sets Career 131–127 Best rank 11 Best season 2021 · 65.1 152 72 34

Click any player for the breakdown · See all 472 ranked players ›

Single best performances

Player	Match	Surface	Break points	Above expected
Viktor Troicki	v Evgeny Donskoy · Roland Garros 2017 · R128 · 7-6(4) 6-4 6-0	Clay	saved 20/21, won 5/6	+10.1
Hicham Arazi	v Francisco Clavet · Australian Open 1998 · R32 · 4-6 6-4 3-6 6-2 6-1	Hard	saved 24/31, won 10/16	+9.2
Arnaud Di Pasquale	v Olivier Delaitre · Roland Garros 1998 · R128 · 2-6 7-5 7-6(3) 1-6 6-0	Clay	saved 24/32, won 8/11	+9.0

By season

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024

How the rating is built — formula, weights, thresholds, exclusions

What goes into the rating

The weights are not chosen. Each component is worth the share of its spread between players that binomial sampling noise cannot explain.

Serving: 0.476
Returning: 0.361
Tiebreaks: 0.163
Deciders: 0.000

For each season, each component is fitted as an ordinary least-squares regression of the pressure rate on the corresponding ordinary rate, across that season's qualifying players. A player's residual — how far above or below the fitted line he came in — is the raw measurement.

Residuals are then shrunk toward zero by empirical Bayes: each component's spread is split into the part binomial sampling noise would produce and the part it cannot, and a player keeps only his share of the second. Without this the leaderboard fills with players who had one good year over a few hundred break points. Career figures are opportunity-weighted across seasons, restandardised across the pool, combined with the weights above, and expressed as 50 + 10 standard deviations.

Standardising within each season is what makes the eras comparable. Between 1991 and 2024 the tour-wide break-point-save rate rose from 58.5% to 61.8% and aces per match from 7.9 to 12.9. A single pool across all 34 seasons would hand modern players an advantage that has nothing to do with nerve.

What this supports: relative to the men he actually played, X performed N standard deviations above the line. What it does not support: that one era's clutch equals another's in absolute terms.

Qualification

Career

100+ matches, 400+ break points faced, 3+ qualifying seasons

Season

15+ matches, 60+ break points faced

What is excluded, and why

Team events and exhibitions	10,244
Retired, walkover or defaulted	2,592
No serve statistics recorded	1,785
Score could not be parsed	6
Failed a consistency check	28
Excluded of 108,375 seen	14,655

Team events go because Davis Cup, the Laver Cup and the United Cup use inconsistent formats and stat coverage, and the Next Gen Finals uses no-ad scoring — which makes every deuce a break point and would poison both break-point components outright. Retirements go entirely rather than partially: their serve figures stop mid-set and their set count breaks deciding-set detection.

Match records compiled by Jeff Sackmann, published as tennis_atp under CC BY-NC-SA 4.0. Computed 2026-08-27.