Grand Slam Point-by-Point Analysis

Tennis
R
Data Analysis
Serve patterns, rally length, break point pressure, distance run, and long-point fatigue across all four Grand Slams (2011–2020)
Published

March 12, 2026

Data

Point-by-point data from Jeff Sackmann’s Grand Slam repository. Covers matches on courts with the Hawkeye tracking system — typically from the second week onward plus select early-round showcourts. Australian Open runs 2011–2020; the other three slams run 2011–2019.

Code 
data_dir <- here::here("data-projects", "data")

slams     <- c("ausopen", "frenchopen", "usopen", "wimbledon")
ao_years  <- 2011:2020
other_years <- 2011:2019

slam_years <- bind_rows(
  tibble(slam = "ausopen",    year = ao_years),
  tibble(slam = "frenchopen", year = other_years),
  tibble(slam = "usopen",     year = other_years),
  tibble(slam = "wimbledon",  year = other_years)
)

base_url <- "https://raw.githubusercontent.com/JeffSackmann/tennis_slam_pointbypoint/master/"

# ── Matches ──────────────────────────────────────────────────────────────────
matches_cache <- file.path(data_dir, "slam_matches.csv")

if (!file.exists(matches_cache)) {
  matches_raw <- pmap_dfr(slam_years, function(slam, year) {
    url <- paste0(base_url, year, "-", slam, "-matches.csv")
    tryCatch(
      read_csv(url, show_col_types = FALSE, col_types = cols(.default = "c")) |>
        mutate(slam = slam, year = year),
      error = \(e) tibble()
    )
  })
  write_csv(matches_raw, matches_cache)
} else {
  matches_raw <- read_csv(matches_cache, show_col_types = FALSE)
}

# ── Points ────────────────────────────────────────────────────────────────────
points_cache <- file.path(data_dir, "slam_points.csv")

keep_cols <- c(
  "match_id", "SetNo", "GameNo", "PointNumber", "PointServer",
  "P1GamesWon", "P2GamesWon", "P1PointsWon", "P2PointsWon",
  "SetWinner", "GameWinner", "PointWinner",
  "P1Score", "P2Score",
  "Speed_KMH", "ServeWidth", "ServeIndicator", "ServeDepth", "ReturnDepth",
  "RallyCount",
  "P1Ace", "P2Ace", "P1Winner", "P2Winner",
  "P1DoubleFault", "P2DoubleFault",
  "P1UnfErr", "P2UnfErr",
  "P1NetPoint", "P2NetPoint",
  "P1BreakPoint", "P2BreakPoint",
  "P1BreakPointWon", "P2BreakPointWon",
  "P1BreakPointMissed", "P2BreakPointMissed",
  "P1DistanceRun", "P2DistanceRun"
)

if (!file.exists(points_cache)) {
  points_raw <- pmap_dfr(slam_years, function(slam, year) {
    url <- paste0(base_url, year, "-", slam, "-points.csv")
    tryCatch({
      df <- read_csv(url, show_col_types = FALSE, col_types = cols(.default = "c"))
      # keep only columns that exist in this file
      df[intersect(keep_cols, names(df))] |>
        mutate(slam = slam, year = as.integer(year))
    }, error = \(e) tibble())
  })
  write_csv(points_raw, points_cache)
} else {
  points_raw <- read_csv(points_cache, show_col_types = FALSE)
}

# ── Clean & type-cast ─────────────────────────────────────────────────────────
pts <- points_raw |>
  mutate(
    across(any_of(c("year", "SetNo", "GameNo", "PointNumber", "PointServer",
                    "PointWinner", "GameWinner", "RallyCount",
                    "P1Ace", "P2Ace", "P1Winner", "P2Winner",
                    "P1DoubleFault", "P2DoubleFault",
                    "P1UnfErr", "P2UnfErr",
                    "P1BreakPoint", "P2BreakPoint",
                    "P1BreakPointWon", "P2BreakPointWon")),
           \(x) suppressWarnings(as.integer(x))),
    across(any_of(c("Speed_KMH", "P1DistanceRun", "P2DistanceRun")),
           \(x) suppressWarnings(as.numeric(x))),
    slam_label = slam_labels[slam]
  )

matches <- matches_raw |>
  mutate(year = as.integer(year))

Coverage overview

Not every match is tracked — Hawkeye data is only available for select courts. This heatmap shows how many matches have point data per slam per year.

Code 
pts |>
  distinct(match_id, slam, year) |>
  count(slam, year) |>
  mutate(slam_label = slam_labels[slam]) |>
  ggplot(aes(factor(year), slam_label, fill = n)) +
  geom_tile(colour = "white", linewidth = 0.5) +
  geom_text(aes(label = n), size = 3.5, colour = "white", fontface = "bold") +
  scale_fill_gradient(low = "#9ECAE1", high = "#08519C") +
  labs(
    title = "Matches with point-by-point data",
    x = NULL, y = NULL, fill = "Matches"
  ) +
  theme(legend.position = "right", axis.text.x = element_text(angle = 45, hjust = 1))

Rally length by slam

Clay courts at Roland Garros produce the longest rallies; grass at Wimbledon the shortest. The distributions below pool all points by slam.

Code 
pts |>
  filter(!is.na(RallyCount), RallyCount >= 1, RallyCount <= 40) |>
  mutate(slam_label = fct_reorder(slam_label, RallyCount, median, .desc = TRUE)) |>
  ggplot(aes(RallyCount, fill = slam)) +
  geom_histogram(binwidth = 1, colour = NA, alpha = 0.85) +
  scale_fill_manual(values = slam_cols, labels = slam_labels) +
  facet_wrap(~slam_label, scales = "free_y") +
  labs(
    title = "Rally length distribution by slam",
    subtitle = "All tracked points, 2011–2020",
    x = "Rally length (shots)", y = "Points", fill = NULL
  ) +
  theme(legend.position = "none")

Code 
pts |>
  filter(!is.na(RallyCount), RallyCount >= 1) |>
  group_by(slam, year) |>
  summarise(
    median_rally = median(RallyCount),
    mean_rally   = mean(RallyCount),
    .groups = "drop"
  ) |>
  mutate(slam_label = slam_labels[slam]) |>
  ggplot(aes(year, median_rally, colour = slam)) +
  geom_line(linewidth = 0.9) +
  geom_point(size = 2) +
  scale_colour_manual(values = slam_cols, labels = slam_labels) +
  labs(
    title = "Median rally length over time by slam",
    x = NULL, y = "Median rally length (shots)", colour = NULL
  )

Serve speed

First serves are substantially faster than second serves at every slam. Hard-court slams (AO, USO) tend to produce slightly higher serve speeds than clay or grass.

Code 
speed_df <- pts |>
  filter(!is.na(Speed_KMH), Speed_KMH > 100, Speed_KMH < 260,
         ServeIndicator %in% c("1", "2")) |>
  mutate(
    serve = if_else(ServeIndicator == "1", "1st serve", "2nd serve"),
    slam_label = slam_labels[slam]
  )

if (nrow(speed_df) > 0) {
  speed_df |>
    ggplot(aes(Speed_KMH, fill = slam)) +
    geom_density(alpha = 0.65, colour = NA) +
    scale_fill_manual(values = slam_cols, labels = slam_labels) +
    facet_grid(serve ~ slam_label) +
    labs(
      title = "Serve speed distribution by slam and serve number",
      x = "Speed (km/h)", y = "Density", fill = NULL
    ) +
    theme(legend.position = "none")
}

Code 
pts |>
  filter(!is.na(Speed_KMH), Speed_KMH > 100, Speed_KMH < 260,
         ServeIndicator %in% c("1", "2")) |>
  mutate(serve = if_else(ServeIndicator == "1", "1st serve", "2nd serve")) |>
  group_by(slam, year, serve) |>
  summarise(median_speed = median(Speed_KMH), .groups = "drop") |>
  mutate(slam_label = slam_labels[slam]) |>
  ggplot(aes(year, median_speed, colour = slam, linetype = serve)) +
  geom_line(linewidth = 0.9) +
  geom_point(size = 2) +
  scale_colour_manual(values = slam_cols, labels = slam_labels) +
  labs(
    title = "Median serve speed over time",
    x = NULL, y = "Median speed (km/h)", colour = NULL, linetype = NULL
  )

Serve direction under pressure

Does a server change their target when facing a break point? ServeWidth encodes: W (wide), B/BC/BW (body), C (centre/T).

Code 
dir_df <- pts |>
  filter(!is.na(ServeWidth), ServeWidth != "", ServeIndicator == "1") |>
  mutate(
    direction = case_when(
      ServeWidth %in% c("W", "BW") ~ "Wide",
      ServeWidth %in% c("B", "BC") ~ "Body",
      ServeWidth == "C"            ~ "Centre/T",
      TRUE                         ~ NA_character_
    ),
    pressure = case_when(
      PointServer == 1 & P1BreakPoint == 1 ~ "Break point against server",
      PointServer == 2 & P2BreakPoint == 1 ~ "Break point against server",
      TRUE                                  ~ "Normal point"
    )
  ) |>
  filter(!is.na(direction)) |>
  count(slam, pressure, direction) |>
  group_by(slam, pressure) |>
  mutate(pct = n / sum(n), slam_label = slam_labels[slam]) |>
  ungroup()

if (nrow(dir_df) > 0) {
  dir_df |>
    ggplot(aes(direction, pct, fill = pressure)) +
    geom_col(position = "dodge", width = 0.7) +
    scale_fill_manual(values = c("Normal point" = "#2171B5",
                                  "Break point against server" = "#CB4335")) +
    scale_y_continuous(labels = percent) +
    facet_wrap(~slam_label) +
    labs(
      title = "1st serve direction: normal vs break point",
      subtitle = "Does pressure shift the target?",
      x = NULL, y = "Share of 1st serves", fill = NULL
    )
}

Break point conversion

Code 
bp_data <- pts |>
  filter(!is.na(P1BreakPoint), !is.na(P2BreakPoint)) |>
  mutate(
    bp_opp  = P1BreakPoint + P2BreakPoint,
    bp_won  = P1BreakPointWon + P2BreakPointWon
  ) |>
  filter(bp_opp == 1) |>
  group_by(slam, year) |>
  summarise(
    conversion = mean(bp_won == 1, na.rm = TRUE),
    n_bp = n(),
    .groups = "drop"
  ) |>
  filter(n_bp >= 20) |>
  mutate(slam_label = slam_labels[slam])

if (nrow(bp_data) > 0) {
  bp_data |>
    ggplot(aes(year, conversion, colour = slam)) +
    geom_line(linewidth = 0.9) +
    geom_point(size = 2) +
    scale_colour_manual(values = slam_cols, labels = slam_labels) +
    scale_y_continuous(labels = percent) +
    labs(
      title = "Break point conversion rate over time",
      subtitle = "Share of break point opportunities converted",
      x = NULL, y = "Conversion rate", colour = NULL
    )
}

Code 
if (nrow(bp_data) > 0) {
  bp_data |>
    group_by(slam, slam_label) |>
    summarise(avg_conversion = mean(conversion), .groups = "drop") |>
    mutate(slam_label = fct_reorder(slam_label, avg_conversion)) |>
    ggplot(aes(avg_conversion, slam_label, fill = slam)) +
    geom_col(width = 0.6) +
    scale_fill_manual(values = slam_cols) +
    scale_x_continuous(labels = percent) +
    labs(title = "Average break point conversion by slam", x = "Conversion rate", y = NULL) +
    theme(legend.position = "none")
}

Distance run: winners vs losers

Do match winners consistently run more or less than their opponents? Distance run data reflects how much each player was pushed around the court.

Code 
# winner column in matches may be a player name or "1"/"2" — try both
matches_winner <- matches_raw |>
  mutate(winner_int = suppressWarnings(as.integer(winner))) |>
  filter(!is.na(winner_int)) |>
  select(match_id, winner_int)

dist_data <- pts |>
  filter(!is.na(P1DistanceRun), !is.na(P2DistanceRun),
         P1DistanceRun > 0, P2DistanceRun > 0) |>
  inner_join(matches_winner, by = "match_id") |>
  mutate(
    winner_dist = if_else(winner_int == 1L, P1DistanceRun, P2DistanceRun),
    loser_dist  = if_else(winner_int == 1L, P2DistanceRun, P1DistanceRun)
  ) |>
  group_by(match_id, slam, year) |>
  summarise(
    winner_total = sum(winner_dist, na.rm = TRUE),
    loser_total  = sum(loser_dist,  na.rm = TRUE),
    .groups = "drop"
  ) |>
  filter(winner_total > 0, loser_total > 0) |>
  mutate(
    diff = winner_total - loser_total,
    slam_label = slam_labels[slam]
  )

if (nrow(dist_data) > 0) {
  dist_data |>
    pivot_longer(c(winner_total, loser_total), names_to = "result", values_to = "dist") |>
    mutate(result = if_else(result == "winner_total", "Match winner", "Match loser")) |>
    ggplot(aes(dist / 1000, fill = result)) +
    geom_density(alpha = 0.6, colour = NA) +
    scale_fill_manual(values = c("Match winner" = "#27AE60", "Match loser" = "#CB4335")) +
    scale_x_continuous(labels = \(x) paste0(x, "k")) +
    facet_wrap(~slam_label) +
    labs(
      title = "Total distance run per match: winners vs losers",
      subtitle = "Metres run by winner and loser across tracked matches",
      x = "Distance run (metres, thousands)", y = "Density", fill = NULL
    ) |> print()
} else {
  message("Distance run data not available in this dataset.")
}
Code 
if (nrow(dist_data) > 0) {
  dist_data |>
    ggplot(aes(diff, fill = slam)) +
    geom_vline(xintercept = 0, linetype = "dashed", colour = "grey50") +
    geom_density(alpha = 0.65, colour = NA) +
    scale_fill_manual(values = slam_cols, labels = slam_labels) +
    labs(
      title = "Winner's distance advantage per match",
      subtitle = "Positive = winner ran more; negative = winner ran less",
      x = "Winner metres − Loser metres", y = "Density", fill = NULL
    )
}

After a long point: does the loser fade?

When a rally goes long, one player wins and one loses. Does the losing player carry any fatigue into the next point — or do elite players reset completely?

The test: for every point following a rally of length ≥ threshold, compute the win rate of the player who lost the previous long point, split by whether they are serving or returning. Compare against their overall win rate in that role.

Code 
if (nrow(fatigue_df) == 0) stop("No fatigue data available.")

fatigue_df |>
  ggplot(aes(threshold, edge, colour = prev_loser_role)) +
  geom_hline(yintercept = 0, linetype = "dashed", colour = "grey50") +
  geom_line(linewidth = 1.1) +
  geom_point(aes(size = n)) +
  scale_colour_manual(values = c("Serving next" = "#2171B5", "Returning next" = "#CB4335")) +
  scale_size_continuous(range = c(2, 6), labels = comma) +
  scale_x_continuous(breaks = thresholds) +
  scale_y_continuous(labels = \(x) paste0(round(x * 100, 1), " pp")) +
  labs(
    title = "Long-point fatigue: does losing a long rally hurt the next point?",
    subtitle = "Win rate of previous-point loser on next point, relative to their baseline\nNegative = they win next point less often than expected",
    x = "Minimum rally length of previous point (shots)",
    y = "Win rate vs baseline (percentage points)",
    colour = "Role on next point",
    size = "Points (n)"
  )

Code 
thr_main <- 10  # focus on rallies of 10+ shots

pts_lagged |>
  filter(prev_rally >= thr_main) |>
  group_by(slam, prev_loser_role) |>
  summarise(
    win_rate = mean(loser_won_next, na.rm = TRUE),
    n = n(),
    .groups = "drop"
  ) |>
  left_join(baseline, by = "prev_loser_role") |>
  mutate(
    edge = win_rate - baseline_win,
    slam_label = slam_labels[slam]
  ) |>
  ggplot(aes(slam_label, edge, fill = slam)) +
  geom_col(width = 0.6) +
  geom_hline(yintercept = 0, linetype = "dashed") +
  scale_fill_manual(values = slam_cols) +
  scale_y_continuous(labels = \(x) paste0(round(x * 100, 1), " pp")) +
  facet_wrap(~prev_loser_role) +
  labs(
    title = "Fatigue effect by slam (rallies ≥ 10 shots)",
    subtitle = "Win rate of long-point loser on next point vs. baseline",
    x = NULL, y = "vs. baseline (pp)", fill = NULL
  ) +
  theme(legend.position = "none", axis.text.x = element_text(angle = 25, hjust = 1))

Code 
fatigue_df |>
  filter(threshold %in% c(5, 10, 15)) |>
  mutate(
    win_pct      = percent(win_rate, accuracy = 0.1),
    baseline_pct = percent(baseline_win, accuracy = 0.1),
    edge_pp      = paste0(round(edge * 100, 2), " pp"),
    n            = comma(n)
  ) |>
  select(
    `Min rally` = threshold,
    Role = prev_loser_role,
    `Win rate (next point)` = win_pct,
    `Baseline win rate` = baseline_pct,
    `Difference` = edge_pp,
    `N` = n
  ) |>
  arrange(`Min rally`, Role) |>
  gt() |>
  tab_header(
    title    = "Long-point fatigue effect",
    subtitle = "Win rate of previous-point loser on the immediately following point"
  ) |>
  tab_style(
    style = cell_text(color = "#CB4335", weight = "bold"),
    locations = cells_body(
      columns = Difference,
      rows    = as.numeric(str_remove(Difference, " pp")) < 0
    )
  )
Long-point fatigue effect
Win rate of previous-point loser on the immediately following point
Min rally Role Win rate (next point) Baseline win rate Difference N
5 Returning next 38.8% 37.5% 1.3 pp 82,993
5 Serving next 59.7% 60.5% -0.81 pp 71,015
10 Returning next 39.8% 37.5% 2.28 pp 17,020
10 Serving next 59.4% 60.5% -1.12 pp 21,661
15 Returning next 40.7% 37.5% 3.22 pp 5,490
15 Serving next 58.9% 60.5% -1.59 pp 4,401

Playstyle Clusters — Last 30 Years

Every player has a statistical “recipe” — the blend of serving aggression, consistency, and defensive resilience that defines how they win matches. K-means clustering on six serve-and-match metrics groups ATP players (1996–2025, ≥ 50 matches) into four distinct archetypes.

Features:

Metric What it captures
Ace rate Serving firepower
Double-fault rate Serve risk tolerance
1st serve in % Serve consistency
1st serve win % Dominance when 1st serve lands
2nd serve win % Baseline / net quality under pressure
BP save rate Clutch performance defending break points
Code 
wss_df |>
  ggplot(aes(k, wss)) +
  geom_line(linewidth = 1) +
  geom_point(size = 3, colour = "#2171B5") +
  geom_vline(xintercept = 4, linetype = "dashed", colour = "#CB4335", linewidth = 0.7) +
  scale_x_continuous(breaks = 1:8) +
  labs(
    title    = "Choosing k: within-cluster sum of squares",
    subtitle = "Elbow at k = 4 — four distinct playstyle archetypes",
    x = "Number of clusters (k)", y = "Total within-cluster SS"
  )

Code 
centroids |>
  left_join(style_map, by = "cluster") |>
  pivot_longer(all_of(feat_cols), names_to = "feature", values_to = "z") |>
  mutate(feature = factor(feat_labels[feature], levels = feat_labels)) |>
  ggplot(aes(feature, z, fill = style)) +
  geom_col(width = 0.7) +
  geom_hline(yintercept = 0, linetype = "dashed", colour = "grey60") +
  scale_fill_manual(values = cluster_cols) +
  facet_wrap(~style, nrow = 2) +
  labs(
    title    = "Playstyle fingerprints — cluster centroids (z-scored)",
    subtitle = "Above zero = above average for that metric across all players",
    x = NULL, y = "Z-score vs. population mean", fill = NULL
  ) +
  theme(legend.position = "none",
        axis.text.x = element_text(angle = 30, hjust = 1))

Code 
player_profiles |>
  ggplot(aes(ace_rate, second_win_pct, colour = style, size = n_matches)) +
  geom_point(alpha = 0.5) +
  scale_colour_manual(values = cluster_cols) +
  scale_x_continuous(labels = percent) +
  scale_y_continuous(labels = percent) +
  scale_size_continuous(range = c(1, 4), guide = "none") +
  labs(
    title    = "Player playstyle clusters: Ace Rate vs 2nd Serve Win%",
    subtitle = "Dot size = career matches played  ·  ATP 1996–2025",
    x = "Ace Rate", y = "2nd Serve Win%", colour = NULL
  )

Code 
player_profiles |>
  group_by(style) |>
  summarise(
    Players        = n(),
    `Win %`        = mean(win_pct),
    `Ace Rate`     = mean(ace_rate),
    `DF Rate`      = mean(df_rate),
    `1st In%`      = mean(first_in_pct),
    `1st Win%`     = mean(first_win_pct),
    `2nd Win%`     = mean(second_win_pct),
    `BP Save%`     = mean(bp_save_rate),
    .groups = "drop"
  ) |>
  arrange(desc(`Win %`)) |>
  gt() |>
  tab_header(
    title    = "Playstyles ranked by win %",
    subtitle = "ATP 1996–2025 · players with ≥ 50 matches · k-means (k = 4)"
  ) |>
  fmt_percent(
    columns  = c(`Win %`, `Ace Rate`, `DF Rate`, `1st In%`,
                 `1st Win%`, `2nd Win%`, `BP Save%`),
    decimals = 1
  ) |>
  fmt_number(columns = Players, decimals = 0) |>
  cols_label(style = "Playstyle") |>
  data_color(
    columns = `Win %`,
    palette = c("#FDECEA", "#A93226")
  ) |>
  tab_style(
    style     = cell_text(weight = "bold"),
    locations = cells_body(rows = 1)
  ) |>
  tab_footnote(
    footnote  = "Raw win % across all tour-level ATP matches; not adjusted for opposition strength.",
    locations = cells_column_labels(columns = `Win %`)
  )
Playstyles ranked by win %
ATP 1996–2025 · players with ≥ 50 matches · k-means (k = 4)
Playstyle Players Win %1 Ace Rate DF Rate 1st In% 1st Win% 2nd Win% BP Save%
Big Server 92 55.3% 12.3% 4.2% 60.9% 76.1% 51.6% 64.5%
Aggressive Baseliner 187 48.0% 6.2% 3.2% 61.8% 69.9% 51.1% 60.7%
All-Court 135 43.7% 8.0% 5.0% 56.6% 72.4% 48.4% 59.9%
Counter-Puncher 158 38.1% 4.4% 3.9% 61.5% 66.5% 48.0% 56.8%
1 Raw win % across all tour-level ATP matches; not adjusted for opposition strength.
Code 
player_profiles |>
  group_by(style) |>
  slice_max(win_pct, n = 6, with_ties = FALSE) |>
  ungroup() |>
  arrange(style, desc(win_pct)) |>
  select(style, player, n_matches, win_pct,
         ace_rate, second_win_pct, bp_save_rate) |>
  gt() |>
  tab_header(
    title    = "Top players by playstyle",
    subtitle = "Highest win % within each cluster (min 50 matches · ATP 1996–2025)"
  ) |>
  fmt_percent(columns = c(win_pct, ace_rate, second_win_pct, bp_save_rate),
              decimals = 1) |>
  fmt_number(columns  = n_matches, decimals = 0) |>
  cols_label(
    style         = "Playstyle",  player        = "Player",
    n_matches     = "Matches",    win_pct       = "Win %",
    ace_rate      = "Ace Rate",   second_win_pct = "2nd Win%",
    bp_save_rate  = "BP Save%"
  ) |>
  tab_row_group(label = "Big Server",           rows = style == "Big Server") |>
  tab_row_group(label = "Aggressive Baseliner", rows = style == "Aggressive Baseliner") |>
  tab_row_group(label = "Counter-Puncher",      rows = style == "Counter-Puncher") |>
  tab_row_group(label = "All-Court",            rows = style == "All-Court") |>
  tab_style(
    style     = cell_fill(color = "#F4F4F4"),
    locations = cells_row_groups()
  ) |>
  data_color(
    columns = win_pct,
    palette = c("#FDECEA", "#A93226")
  )
Top players by playstyle
Highest win % within each cluster (min 50 matches · ATP 1996–2025)
Playstyle Player Matches Win % Ace Rate 2nd Win% BP Save%
All-Court
All-Court Lleyton Hewitt 817 70.1% 7.7% 52.7% 61.9%
All-Court Yevgeny Kafelnikov 641 66.6% 7.4% 50.4% 60.9%
All-Court Tim Henman 696 64.9% 7.4% 50.8% 63.1%
All-Court Petr Korda 200 64.5% 6.3% 49.4% 62.3%
All-Court Thomas Enqvist 506 60.3% 10.2% 50.2% 60.0%
All-Court Magnus Gustafsson 262 59.2% 4.9% 49.4% 60.6%
Counter-Puncher
Counter-Puncher Guillermo Coria 321 66.4% 3.8% 50.1% 58.6%
Counter-Puncher Gaston Gaudio 444 57.4% 3.5% 49.9% 58.3%
Counter-Puncher Alberto Berasategui 250 56.0% 2.3% 47.5% 56.5%
Counter-Puncher Juan Ignacio Chela 574 53.5% 4.0% 50.3% 58.4%
Counter-Puncher Diego Schwartzman 466 53.2% 2.5% 49.8% 56.8%
Counter-Puncher Todd Woodbridge 192 52.6% 4.0% 47.0% 57.8%
Aggressive Baseliner
Aggressive Baseliner Rafael Nadal 1,280 82.5% 4.3% 57.1% 66.0%
Aggressive Baseliner Carlos Alcaraz 263 79.1% 5.1% 55.7% 63.8%
Aggressive Baseliner Andy Murray 970 73.3% 8.5% 51.7% 62.0%
Aggressive Baseliner Marcelo Rios 460 67.6% 6.5% 52.9% 61.2%
Aggressive Baseliner Kei Nishikori 639 65.9% 4.1% 53.2% 60.9%
Aggressive Baseliner David Nalbandian 544 65.8% 4.6% 52.3% 59.3%
Big Server
Big Server Novak Djokovic 1,299 83.9% 7.3% 55.5% 65.6%
Big Server Roger Federer 1,458 82.0% 10.0% 56.8% 67.3%
Big Server Pete Sampras 400 77.5% 13.9% 52.5% 68.2%
Big Server Jannik Sinner 340 76.8% 7.5% 54.8% 66.6%
Big Server Andre Agassi 567 76.0% 7.1% 53.9% 65.1%
Big Server Andy Roddick 773 74.5% 15.5% 55.9% 67.6%