ATP Paired-Comparison Panels, 2000–2022

A yearly panel of head-to-head counts for the top 105 ATP players, suitable for Bradley–Terry/Stochastic Block Model analyses. For each season, the data include (i) the number of wins by player \(i\) against player \(j\) and (ii) the number of matches played between \(i\) and \(j\), along with a per-season player metadata tibble.

Usage

ATP_2000_2022

Format

A named list of length 23, with elements "2000", "2001", …, "2022". Each yearly element is a list of length 3:

Y_ij

A numeric matrix \(105 \times 105\). Entry Y_ij[i, j] is the count of matches in which player \(i\) defeated player \(j\) in that calendar year (nonnegative integer; diagonal is zero).

N_ij

A numeric matrix \(105 \times 105\). Entry N_ij[i, j] is the total number of matches between players \(i\) and \(j\) that year (nonnegative integer; symmetric by construction; diagonal is zero).

players_df

A tibble/data frame with 105 rows and 7 columns describing the player index used in the matrices for that year:

player

Integer player identifier (row/column index used in Y_ij and N_ij).

worst_rank

Worst (numerically largest) ATP ranking attained by the player during the year.

median_rank

Median ATP ranking across the player's ranking snapshots in that year.

last_rank

ATP ranking at the last snapshot available in the year (e.g., year-end ranking).

age_year

Approximate age (in years) for that season (e.g., at mid-season).

ht_year

Player height in centimeters (season-level value).

player_slug

Character identifier (URL-safe or underscored name) for the player.

Source

Aggregated by the package author from public ATP results (e.g., the tennis\_atp datasets by Jeff Sackmann) and internal preprocessing. See the package vignette for provenance and cleaning steps.

Details

The player ordering in players_df defines the row/column indexing of Y_ij and N_ij for the corresponding year. The diagonal entries of both matrices are zero by definition. In typical usage for Bradley–Terry-type models, one can treat Y_ij[i, j] as the number of “successes” for \(i\) vs. \(j\), with the binomial denominator N_ij[i, j].

Note

The matrices may be sparse for many player pairs. Ensure any model code guards against divisions by zero when N_ij[i, j] = 0.

Examples

 if (FALSE) { # \dontrun{
data(ATP_2000_2022)
names(ATP_2000_2022)
year <- "2000"
str(ATP_2000_2022[[year]])

# Player i's total wins that year:
i <- 1
sum(ATP_2000_2022[[year]]$Y_ij[i, ], na.rm = TRUE)

# Total matches between i and j:
j <- 2
ATP_2000_2022[[year]]$N_ij[i, j]

# Join player metadata to indices used in the matrices:
head(ATP_2000_2022[[year]]$players_df)
} # }