1 Background

Basketball is one of the most widely played and watched sports in the world, with professional and collegiate competitions attracting millions of fans across North America and beyond. This report examines game-by-game team performance data across four competitions: the National Basketball Association (NBA), the Women’s National Basketball Association (WNBA), NCAA Division I Men’s Basketball (NCAAM), and NCAA Division I Women’s Basketball (NCAAW), spanning the 2016 to 2025 seasons.

The NBA is the premier professional men’s basketball league in the world, featuring 30 franchises competing across an 82-game regular season, followed by a playoff series culminating in the NBA Finals. The WNBA is the premier professional women’s basketball league, featuring 14 teams competing across a shorter regular season and playoffs. The NCAAM and NCAAW competitions represent the highest level of collegiate basketball in the United States, governed by the National Collegiate Athletic Association (NCAA), with over 1,400 men’s and 1,000 women’s Division I programs competing each season, culminating in the annual NCAA Tournament — commonly known as March Madness.

The data used in this report was sourced from ESPN’s publicly available game data, capturing team-level box score statistics for every game across all four competitions from 2016 to 2025. The dataset records a wide range of team performance metrics including scoring, shooting percentages, rebounds, assists, turnovers, steals, blocks, and fouls, providing a comprehensive picture of team performance across competitions and seasons.

2 Data Description

2.1 Dataset Overview

The dataset contains 246,748 game-by-game team observations across four competitions spanning ten seasons from 2016 to 2025. Each row represents one team’s performance in a single game, meaning every game contributes two rows to the dataset — one for each competing team. The dataset covers 45 variables capturing a wide range of team performance metrics including scoring, shooting, rebounding, assists, turnovers, and defensive statistics.

Table 1 provides a high level summary of the dataset across the four competitions.

Table 1: Dataset Overview by Competition
Competition Teams Games Rows Seasons
NBA 30 ~12,678 25,356 2016-2025
WNBA 14 ~2,310 4,620 2016-2025
NCAAM 1455 ~58,603 117,206 2016-2025
NCAAW 1018 ~49,783 99,566 2016-2025
Total 1,517 ~123,374 246,748 2016-2025

Notable points regarding the dataset:

  • The 2020 and 2021 seasons were shortened across all competitions due to the COVID-19 pandemic, resulting in fewer games in those seasons
  • The NCAAW 2016 season appears incomplete with significantly fewer games than subsequent seasons
  • fastbreak_points, points_in_paint and turnover_points are only available for NBA games — these columns are structurally missing for NCAAM, NCAAW and partially WNBA
  • largest_lead and flagrant_fouls have structural missingness in NCAAW, reflecting data collection differences across competitions

2.2 Data Processing

Prior to analysis, several data processing steps were undertaken to clean and prepare the dataset. The raw dataset contained 57 variables, which was reduced to 45 following the removal of redundant and unnecessary columns. Columns removed included multiple team and opponent identifier fields (team IDs, slugs, UIDs), redundant naming columns where a single display name was retained, and visual/branding columns for opponent teams including colours and logos. The game_date_time column was also removed as the game_date column provided sufficient temporal information.

Several variables were renamed to improve clarity and consistency. team_display_name and opponent_team_display_name were shortened to team and opponent respectively. Shooting percentage columns were renamed to fg_pct, ft_pct, and three_p_pct, and team_home_away was simplified to home_away. American English spellings were also corrected to Australian English, with color renamed to colour. Three point field goal columns were renamed with the prefix three_p_ for consistency.

Several value and type conversions were applied. The competition column was converted to uppercase to ensure consistency across all four competitions. The game_date column was converted from a character string to a proper Date type using the format dd/mm/yyyy. The season_type column was recoded from numeric codes to descriptive labels, where 2 = Regular Season, 3 = Playoffs, and 5 = Play-In.

Five new variables were created to extend the dataset for analysis. point_differential was calculated as the difference between a team’s score and their opponent’s score. result was derived from the team_winner column as a W/L character variable. two_p_made and two_p_attempted were derived by subtracting three point field goals from total field goals, and two_p_pct was calculated as the percentage of two point attempts converted. Additionally, Dean Oliver’s four factors were calculated: Effective FG% (eFG), Turnover Rate (TOV_rate), Offensive Rebound Rate (ORB_rate), and Free Throw Rate (FTR). A champion flag (is_champion) was also added to identify championship winning teams in each competition and season.

Non-team entries were removed from the NBA and WNBA competitions, including All-Star teams, exhibition squads, and national team entries.

2.3 Missing Values & Outliers

The dataset was assessed for missing, erroneous, and outlier values prior to analysis.

The largest category of missing values was identified in three NBA-specific columns: fastbreak_points, points_in_paint, and turnover_points. These columns were missing entirely for NCAAM and NCAAW, and largely missing for WNBA, reflecting the fact that these statistics are not collected or reported in the same way across all competitions. These missing values are structural rather than erroneous and the columns were retained, with the understanding that they are only used in NBA-specific analysis.

Missing values were also identified in largest_lead (27,076 missing, concentrated in NCAAW) and flagrant_fouls (23,048 missing, concentrated in NCAAW), again reflecting structural differences in data collection across competitions rather than genuine data quality issues. These columns were retained with their missing values intact.

Two erroneous values were identified and corrected. A single observation for the South Carolina Upstate Spartans in the 2019 NCAAW season recorded 727 turnovers — clearly a data entry error — and was set to NA. Additionally, 36 game observations across the 2017 NCAAW season recorded a team score of 0, concentrated on a single date (7 January 2017), suggesting a data collection failure rather than genuine results. These values were also set to NA.

An investigation of the 2018 NBA season revealed widespread data quality issues in the fg_pct, steals, and blocks columns, with values substantially above realistic ranges for professional basketball. The fg_pct erroneous values were set to NA, and the 2018 season was excluded from the Pace & Space vs Defence use case analysis due to the impact on steals and blocks classification. No duplicate rows were identified in the dataset.

2.4 Variable Definitions

The dataset contains 45 variables capturing team level game performance across the four competitions. Key variables relevant to this analysis are defined in Table 2 below.

Table 2: Key Variable Definitions
Variable Description Type
competition The competition the game belongs to (NBA, WNBA, NCAAM, NCAAW) Categorical
season The season year the game was played in Numeric
season_type The stage of the season (Regular Season, Playoffs, Play-In) Categorical
game_date The date the game was played Date
team The name of the team Categorical
opponent The name of the opposing team Categorical
home_away Whether the team was playing at home or away Categorical
result The result of the game for the team (W or L) Categorical
team_winner Whether the team won the game (TRUE or FALSE) Logical
team_score The total points scored by the team Numeric
opponent_team_score The total points scored by the opposing team Numeric
point_differential The difference between team score and opponent score Numeric
field_goals_made The number of field goals made Numeric
field_goals_attempted The number of field goals attempted Numeric
fg_pct The percentage of field goals converted Numeric
two_p_made The number of two point field goals made (derived) Numeric
two_p_attempted The number of two point field goals attempted (derived) Numeric
two_p_pct The percentage of two point field goals converted (derived) Numeric
three_p_made The number of three point field goals made Numeric
three_p_attempted The number of three point field goals attempted Numeric
three_p_pct The percentage of three point field goals converted Numeric
free_throws_made The number of free throws made Numeric
free_throws_attempted The number of free throws attempted Numeric
ft_pct The percentage of free throws converted Numeric
offensive_rebounds The number of offensive rebounds Numeric
defensive_rebounds The number of defensive rebounds Numeric
total_rebounds The total number of rebounds Numeric
assists The number of assists Numeric
steals The number of steals Numeric
blocks The number of blocks Numeric
turnovers The number of turnovers Numeric
fouls The number of personal fouls Numeric
largest_lead The largest lead held by the team during the game Numeric
fastbreak_points Points scored in transition/fastbreak situations (NBA only) Numeric
points_in_paint Points scored in the paint (NBA only) Numeric
turnover_points Points scored off turnovers (NBA only) Numeric
eFG Effective Field Goal Percentage — accounts for the added value of three pointers (derived) Numeric
TOV_rate Turnover Rate — proportion of possessions ending in a turnover (derived) Numeric
ORB_rate Offensive Rebound Rate — proportion of available offensive rebounds secured (derived) Numeric
FTR Free Throw Rate — ratio of free throw attempts to field goal attempts (derived) Numeric
is_champion Whether the team was the championship winner in that competition and season (derived) Categorical

2.5 Data Visualisations

2.5.1 Competition & Season Overview

The dataset spans ten seasons across four competitions, with a total of approximately 123,374 unique games. Figure 1 displays the distribution of games across the four competitions, highlighting the significantly larger volume of college basketball games relative to the professional leagues.

Figure 1 displays the number of unique games per competition across the 2016 to 2025 seasons.

NCAAM has by far the largest number of games (~58,603), followed by NCAAW (~49,783), reflecting the vast number of Division I programs competing each season. The NBA (~12,678) and WNBA (~2,310) have significantly fewer games, consistent with their smaller number of teams and shorter seasons.

Figure 2 displays the number of games per season broken down by competition from 2016 to 2025.

A clear dip is visible across all four competitions in 2020 and 2021, reflecting the impact of the COVID-19 pandemic which resulted in shortened or suspended seasons. The NCAAW 2016 season shows notably fewer games than subsequent seasons, consistent with the incomplete data identified during the data quality assessment. The WNBA shows a gradual increase in games from 2022 onwards, reflecting league expansion with new franchises entering the competition.

2.5.2 Scoring

Figure 3 displays the distribution of team scores across the four competitions from 2016 to 2025.

The NBA records the highest average team score (110.0 points) reflecting the longer 48 minute game compared to the 40 minute college game. The WNBA averages 81.7 points, followed by NCAAM (72.1) and NCAAW (65.1). Notably, NCAAW has the largest spread in scoring (SD = 14.0), reflecting the wide range of team quality across over 1,000 Division I programs. The NBA and NCAAM show the most symmetric distributions, while NCAAW shows a wider, more spread out shape consistent with greater variability in team quality.

2.5.3 Scoring Breakdown

Figure 4 displays the average points contributed by each shot type per competition from 2016 to 2025.

The NBA scores significantly more across all categories — 57.9 points from two pointers, 34.7 from three pointers and 17.4 from free throws. However, this is partly explained by the longer 48 minute game compared to the 40 minute college game. To account for this, Figure 5 presents the proportional breakdown of scoring by shot type.

Figure 5 displays the proportional contribution of each shot type to total scoring per competition from 2016 to 2025.

When controlling for total scoring, the composition is more similar across competitions than the raw numbers suggest. NBA and NCAAM show near identical three point reliance (~31%), while women’s competitions rely less on the three ball (WNBA 26.2%, NCAAW 27.5%). Two pointers remain the dominant scoring source across all competitions (50-56%), and free throw proportions are remarkably consistent (15-19%), suggesting that the scoring differences between competitions are largely explained by game length and pace rather than style of play.

2.5.4 Shooting

Figure 6 displays average shooting percentages by competition from 2016 to 2025.

The NBA leads in field goal efficiency (49.6%) and two point shooting (52.6%), reflecting the skill level of professional players. The WNBA leads all competitions in free throw shooting (79.6%), while NCAAW is the least efficient competition across all four categories. Notably, WNBA and NCAAM share identical FG% (44.1%) and 3P% (34.0%) despite the significant difference in level of play. College averages are based on far more games (~58,000 NCAAM, ~49,000 NCAAW) compared to the NBA (~12,600) and WNBA (~2,300), reflecting a much wider range of team quality from elite programs to smaller Division I schools.

Figure 7 displays the average two point field goal percentage by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

All competitions show a general upward trend in 2P% from 2016 to 2025. The NBA peaked in 2023 (54.9%) and has slightly declined since, while NCAAM and NCAAW peaked in 2025 (51.2% and 45.8% respectively). The WNBA shows the most volatility, with a notable dip in 2019 (46.5%) before recovering, likely attributable to the smaller sample size relative to other competitions.

Figure 8 displays the average two point field goal attempts per game by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

Two point attempts have declined across all competitions, most dramatically in the NBA (60.3 in 2016 to 51.6 in 2025) and WNBA (51.5 to 43.4). This decline mirrors the rise in three point attempts observed in Figure 10, suggesting teams are replacing low quality two point shots with three point attempts while retaining only high percentage two point looks such as layups and dunks.

Figure 9 displays the average three point field goal percentage by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

The NBA peaked in three point efficiency in 2021 (36.6%) before a slight dip and recovery, while both college competitions show a declining trend despite increasing attempts — NCAAM dropping from 34.9% in 2017 to 33.6% in 2025, and NCAAW from 31.6% in 2018 to 30.8% in 2025. This suggests that as the three pointer becomes more fashionable at the college level, shot selection may be suffering as teams attempt more threes than their skill level can sustain.

Figure 10 displays the average three point field goal attempts per game by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

NBA three point attempts have increased dramatically from 24.2 per game in 2016 to 37.4 in 2025 — a 54% increase — while efficiency has remained relatively stable (35.2% to 35.9%), suggesting elite players are maintaining accuracy at higher volumes. This is in contrast to college basketball where volume increases have come at the cost of efficiency, as discussed in Figure 9.

Figure 11 displays the average free throw attempts per game by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

Contrary to the popular narrative of “foul baiting” in modern basketball, free throw attempts have declined or remained flat across all four competitions from 2016 to 2025. The NBA peaked in 2023 (23.4 attempts) before declining again, while the WNBA shows the largest overall decline (21.1 to 18.2). The shift towards perimeter play is a likely driver — it is harder to draw fouls on three point attempts than in the paint, and more floor spacing means fewer physical contests. Some analysts also argue that defensive intensity has declined in the modern game, with teams prioritising switching and drop coverage over physical man-to-man defence.

2.5.5 Rebounding

Figure 12 displays the average total rebounds per 40 minutes by competition from 2016 to 2025. All rebounding statistics were normalised per 40 minutes to account for the difference in game length between the NBA (48 minutes) and all other competitions (40 minutes).

The NBA leads in total rebounds per 40 minutes (40.2) despite the normalisation, reflecting superior athleticism and court awareness. The WNBA has the fewest rebounds per 40 minutes (34.3), suggesting a more structured half court style with fewer missed shots and therefore fewer rebound opportunities.

Figure 13 displays the average offensive and defensive rebounds per 40 minutes by competition from 2016 to 2025.

NCAAW leads all competitions in offensive rebounding (11.7 per 40 minutes), suggesting a more aggressive glass crashing style that may compensate for lower shooting efficiency. The NBA leads in defensive rebounding (28.1 per 40 minutes), reflecting superior athleticism and defensive organisation. Across all competitions, defensive rebounds significantly outnumber offensive rebounds, consistent with the general principle that teams secure possession on the defensive end more often than they extend possessions offensively.

2.5.6 Assists

Figure 14 displays the average assists per 40 minutes by competition from 2016 to 2025.

The WNBA leads all competitions in assists per 40 minutes (19.4) ahead of the NBA (18.7), suggesting the WNBA plays a more team oriented, ball movement style of basketball. Both professional leagues significantly outperform college basketball (NCAAM 13.3, NCAAW 13.1 per 40 minutes), likely reflecting more structured offensive systems and higher skill levels in passing and decision making. College basketball may also feature more isolation based play, where individual players create their own shots rather than generating assists through ball movement. The large range of team quality in college basketball — from elite programs to smaller Division I schools — further contributes to the lower average assist numbers.

2.5.7 Defence

Figure 15 displays the average steals and blocks per 40 minutes by competition from 2016 to 2025.

Steals and blocks — sometimes referred to collectively as “stocks” — reveal interesting contrasts across competitions. The NBA leads in blocks (5.3 per 40 minutes), consistent with the greater average height and athleticism of NBA players who are better positioned to contest shots at the rim. Surprisingly, NCAAW leads all competitions in steals (7.7 per 40 minutes) despite having the lowest blocks. Figure 16 provides important context for this finding.

Figure 16 displays the average steals versus turnovers per 40 minutes by competition from 2016 to 2025.

NCAAW also has the highest turnover rate (15.9 per 40 minutes), suggesting the high steal numbers observed in Figure 15 are partly a product of more frequent turnovers rather than purely superior defensive activity. The NBA has by far the lowest turnover rate (10.0 per 40 minutes), reflecting the superior ball handling and decision making of professional male players. The gap between steals and turnovers is largest in NCAAW, indicating that while steals are high, many turnovers result from other causes such as out of bounds violations, shot clock violations, and unforced errors.

2.5.8 Home Advantage

Figure 17 displays the home win percentage by competition over time from 2016 to 2025. Peak seasons for each competition are highlighted.

Home advantage is present across all four competitions, with college basketball showing a significantly stronger effect (NCAAM 66.6%, NCAAW 62.1%) compared to professional leagues (NBA 56.8%, WNBA 55.5%). The COVID-19 bubble seasons of 2020 and 2021 provide a natural experiment — all competitions show a clear drop in home win percentage when games were played without crowds, confirming that crowd atmosphere is a meaningful driver of home advantage. The NBA home advantage has gradually declined from approximately 59% pre-COVID to around 55% post-COVID and has not fully recovered, possibly reflecting changes in how teams approach travel and game preparation in the modern era.

2.5.9 NBA Specific Statistics

Three statistics — fastbreak points, points in the paint, and turnover points — are only available for NBA games and are explored in this section. Fastbreak points refer to points scored in transition, where a team pushes the ball up the court quickly before the opposing defence can set up. Teams that score heavily in transition tend to play at a faster pace, forcing turnovers and converting them into easy scoring opportunities at the other end.

Figure 18 displays the average fastbreak points per game by NBA team across all seasons from 2016 to 2025, broken down by conference.

The Toronto Raptors and Golden State Warriors stand out as the leading fastbreak teams over this period, reflecting their identity as high energy, transition-heavy teams. Recently Oklahoma City Thunder’s young core has generated fastbreak opportunities through aggressive defence and forced turnovers. At the other end of the spectrum, teams like the Utah Jazz (who have been rebuilding for 10 years) and Miami Heat historically favour a more controlled, half court style of play which naturally limits fastbreak opportunities. Figures 19 to 24 break down fastbreak points by division over time, using team logos and colours to identify individual franchises.

Figure 19 displays the average fastbreak points per season for Atlantic Division teams from 2016 to 2025.

Figure 20 displays the average fastbreak points per season for Central Division teams from 2016 to 2025.

Figure 21 displays the average fastbreak points per season for Southeast Division teams from 2016 to 2025.

Figure 22 displays the average fastbreak points per season for Northwest Division teams from 2016 to 2025.

Figure 23 displays the average fastbreak points per season for Pacific Division teams from 2016 to 2025.

Figure 24 displays the average fastbreak points per season for Southwest Division teams from 2016 to 2025.

Across all six divisions, fastbreak scoring has fluctuated considerably from season to season, reflecting changes in roster composition, coaching philosophy and team identity over the decade. The Golden State Warriors in the Pacific Division and the Memphis Grizzlies in the Southwest stand out as consistently high fastbreak teams, while teams such as the San Antonio Spurs and New York Knicks have consistently ranked among the lowest. Notably, the Oklahoma City Thunder in the Northwest Division show a sharp increase in fastbreak points in recent seasons (2024-2025), consistent with their emergence as one of the NBA’s most dynamic and athletically gifted young teams — a trend that culminated in their 2025 NBA Championship.

3 Data Use Cases

The following section proposes two data use cases that leverage the basketball game-by-game dataset spanning the NBA, WNBA, NCAAM and NCAAW from 2016 to 2025. Each use case identifies a practical application of the data for teams, coaches, analysts, and front office staff, supported by visualisations and statistics. For each use case, a written description, hypothesis, and discussion of limitations are provided.

3.1 Use Case 1: Dean Oliver’s Four Factors — What Makes a Champion?

3.1.1 Description

Dean Oliver’s Four Factors of Basketball Success, introduced in his 2004 book Basketball on Paper, propose that winning in basketball is primarily determined by four key statistical categories: shooting efficiency (Effective Field Goal Percentage), ball security (Turnover Rate), offensive rebounding (Offensive Rebound Rate), and getting to the free throw line (Free Throw Rate). Oliver argued that these four factors, when combined, explain the majority of variation in team winning percentage and provide a framework for evaluating team quality beyond simple scoring metrics.

This use case applies Oliver’s Four Factors framework to game-by-game data across all four competitions from 2016 to 2025, with the aim of identifying which factors most consistently separate championship winning teams from the rest of the competition. A particular focus is placed on the 2025 season, where the four factors of each competition’s champion are compared directly against the competition average, providing a practical scouting tool for coaches and analysts looking to understand what it takes to build a championship calibre team.

The four factors were calculated as follows:

  • Effective FG% (eFG%) = (Field Goals Made + 0.5 × Three Point Field Goals Made) / Field Goals Attempted
  • Turnover Rate (TOV%) = Turnovers / (Field Goals Attempted + 0.44 × Free Throws Attempted + Turnovers)
  • Offensive Rebound Rate (ORB%) = Offensive Rebounds / (Offensive Rebounds + Defensive Rebounds)
  • Free Throw Rate (FTR) = Free Throws Attempted / Field Goals Attempted

3.1.2 Visualisations

Table 3 presents the average four factor values for champion and non-champion teams across all four competitions from 2016 to 2025.

Table 3: Four Factors — Champions vs Non-Champions (2016-2025)
Competition Status eFG% TOV Rate ORB Rate FT Rate
NBA Champion 0.555 0.106 0.221 0.250
NBA Non-Champion 0.529 0.109 0.232 0.261
WNBA Champion 0.526 0.137 0.225 0.265
WNBA Non-Champion 0.490 0.147 0.247 0.274
NCAAM Champion 0.555 0.139 0.302 0.324
NCAAM Non-Champion 0.505 0.162 0.285 0.334
NCAAW Champion 0.535 0.153 0.317 0.290
NCAAW Non-Champion 0.454 0.194 0.313 0.290

Across all four competitions, champion teams consistently show higher eFG% and lower TOV rates than non-champions, suggesting that shooting efficiency and ball security are the most universal traits of championship teams. The differences in ORB rate and FT rate are less consistent, with champions in professional leagues actually showing lower ORB rates than non-champions — a finding explored further in the radar charts below.

Figure 25 displays a radar chart of the four factors for champion and non-champion teams in the NBA from 2016 to 2025. Note that for Turnover Rate, a smaller value is better — champion teams appear smaller on this axis because they commit fewer turnovers.

NBA champions show a clear advantage in eFG% (0.555 vs 0.529), confirming that shooting efficiency is the primary differentiator in the NBA. The turnover rate is slightly lower for champions (0.106 vs 0.109), while ORB rate is actually lower for champions (0.221 vs 0.232) — suggesting NBA champions prioritise transition defence over crashing the offensive glass. FTR is also slightly lower for champions (0.250 vs 0.261), challenging the popular narrative that drawing fouls is a key championship trait.

Figure 26 displays a radar chart of the four factors for champion and non-champion teams in the WNBA from 2016 to 2025.

WNBA champions show a similar pattern to the NBA, with a notable eFG% advantage (0.526 vs 0.490) and lower turnover rate (0.137 vs 0.147). The ORB rate gap is larger in the WNBA than the NBA, with champions again showing lower offensive rebounding rates (0.225 vs 0.247), reinforcing the finding that champion teams in professional leagues prioritise defensive positioning over second chance opportunities.

Figure 27 displays a radar chart of the four factors for champion and non-champion teams in NCAAM from 2016 to 2025.

NCAAM champions show the largest eFG% advantage of all four competitions (0.555 vs 0.505) and a substantially lower turnover rate (0.139 vs 0.162). Interestingly, NCAAM champions show a higher ORB rate than non-champions (0.302 vs 0.285), in contrast to the professional leagues — suggesting that offensive rebounding plays a more important role in college championship success, possibly reflecting the greater emphasis on physical play and second chance points in the college game.

Figure 28 displays a radar chart of the four factors for champion and non-champion teams in NCAAW from 2016 to 2025.

NCAAW champions show the most dramatic separation from non-champions of all four competitions, particularly in eFG% (0.535 vs 0.454) and TOV rate (0.153 vs 0.194). The eFG% gap of 0.081 is more than double that of the NBA, suggesting that in women’s college basketball, shooting efficiency is an exceptionally strong predictor of championship success. FTR is almost identical between champions and non-champions (0.290 vs 0.290), making it the least differentiating factor in NCAAW.

Figure 29 displays a heatmap of the correlation between each of the four factors and winning across all four competitions from 2016 to 2025.

The heatmap confirms that eFG% is by far the strongest predictor of winning across all four competitions (0.458-0.519), shown by the deep blue shading. Turnover Rate is negatively correlated with winning across all competitions (-0.069 to -0.247), meaning teams that turn the ball over more tend to win less. Offensive Rebound Rate is also negatively correlated with winning (-0.115 to -0.174), which is a counterintuitive finding suggesting that teams which chase offensive rebounds may be leaving themselves vulnerable in transition defence. FTR has the weakest and most variable relationship with winning across competitions (0.096 to 0.229), being least important in the NBA and most important in college basketball.

Figure 30 displays a scatter plot of Effective FG% versus win rate for all team seasons across the four competitions from 2016 to 2025, with a linear trend line for each competition.

The scatter plot confirms a clear positive relationship between eFG% and win rate across all four competitions — teams that shoot more efficiently win more games. The relationship is consistent across competitions, with dots clustering in the upper right (high eFG%, high win rate) and lower left (low eFG%, low win rate). The college competitions show a wider spread of data points, reflecting the greater range of team quality across Division I programs. The NBA and WNBA show tighter clusters, consistent with the more even distribution of talent in professional leagues.

Figure 31 displays the four factors for the 2025 men’s champions (Oklahoma City Thunder and Florida Gators) compared to the rest of their respective competitions.

The Oklahoma City Thunder’s 2025 championship profile is particularly noteworthy. Their turnover rate (0.099) was substantially lower than the rest of the NBA (0.122), making elite ball security their defining championship trait. This directly contradicts the popular narrative that OKC relies heavily on drawing fouls — their FTR (0.237) was actually lower than the league average (0.250), suggesting they won the championship through disciplined, efficient basketball rather than foul baiting. The Florida Gators showed advantages in all four factors, with a particularly notable eFG% advantage (0.550 vs 0.510) and lower turnover rate (0.133 vs 0.152), consistent with the broader finding that eFG% and TOV rate are the key differentiators in college basketball.

Figure 32 displays the four factors for the 2025 women’s champions (Las Vegas Aces and UConn Huskies) compared to the rest of their respective competitions.

The UConn Huskies present the most dominant championship profile of all four 2025 champions, with a massive eFG% advantage (0.581 vs 0.460) — the largest gap observed across any champion in any competition. Their TOV rate (0.135) was also substantially lower than the NCAAW average (0.199), and their FTR (0.193) was significantly below the competition average (0.284), suggesting they won the championship by scoring efficiently without needing to draw fouls. The Las Vegas Aces showed more modest advantages across all four factors, consistent with a closely contested WNBA season.

Figure 33 displays the average season statistics for the 2025 men’s champions compared to the rest of their respective competitions.

The Oklahoma City Thunder led the rest of the NBA in scoring, assists, steals and blocks while recording fewer turnovers, presenting the profile of a complete, well-rounded championship team. The Florida Gators similarly outperformed the NCAAM average across most categories, with notably higher blocks and steals suggesting a strong defensive identity alongside their offensive efficiency.

Figure 34 displays the average season statistics for the 2025 women’s champions compared to the rest of their respective competitions.

The UConn Huskies dominated across every statistical category compared to the NCAAW average, scoring significantly more points, generating more assists, and committing fewer turnovers. This comprehensive statistical dominance is consistent with their historic reputation as one of the most dominant programs in women’s college basketball. The Las Vegas Aces showed more modest advantages, leading in scoring and assists while recording slightly fewer turnovers than the WNBA average.

3.1.3 Hypothesis

It was hypothesised that all four of Dean Oliver’s factors would consistently separate champions from non-champions across all four competitions, with eFG% expected to be the strongest differentiator given the fundamental importance of shooting efficiency in basketball.

The data strongly supports this hypothesis. eFG% is the most consistent and strongest predictor of winning across all four competitions, with a correlation ranging from 0.458 (NBA) to 0.519 (NCAAW). Champions outperform non-champions in eFG% in every competition and every season analysed, making shooting efficiency the single most reliable indicator of championship quality.

Turnover rate also consistently separates champions from non-champions, with champions recording lower turnover rates in every competition. This confirms that protecting the ball is a fundamental championship trait across all levels of basketball.

However, two findings challenge conventional basketball wisdom. First, offensive rebound rate is actually lower for champions in the NBA and WNBA compared to non-champions, suggesting that professional championship teams prioritise transition defence and court positioning over offensive rebounding. Second, free throw rate shows minimal separation between champions and non-champions, particularly in the NBA (0.250 vs 0.261), directly contradicting the narrative that drawing fouls is a key component of championship basketball.

The 2025 champions provide a compelling real world application of these findings. The OKC Thunder’s historically low turnover rate and the UConn Huskies’ dominant eFG% advantage both reflect teams that excelled in the factors most strongly associated with winning, providing coaches and analysts with a clear blueprint for what championship calibre basketball looks like across different competitions and levels of play.

3.1.4 Limitations

Several limitations should be considered when interpreting the findings of this use case.

First, the four factors are calculated from game-by-game team totals rather than possession-level data, meaning the turnover rate and offensive rebound rate calculations are approximations based on Oliver’s original formulas rather than exact possession counts. More precise calculations would require play-by-play data with explicit possession tracking.

Second, the champion flag is based on the overall season champion in each competition, meaning the four factor values represent a team’s average performance across the entire season including games before and after their championship run. A more granular analysis could examine four factor performance specifically during the playoffs or tournament to determine whether championship form differs from regular season performance.

Third, the four factors framework was developed by Oliver primarily in the context of NBA basketball. While it has been widely applied to other competitions, its relative weighting across the four factors may differ across the NBA, WNBA, NCAAM and NCAAW, meaning direct comparisons of absolute values across competitions should be interpreted with caution.

Finally, the four factors framework does not account for opponent quality. A team with a high eFG% against weak opponents may not perform the same way against elite defences in a championship series, meaning the framework provides a useful but incomplete picture of championship readiness.

3.2 Use Case 2: Pace & Space vs Defence Wins Championships — NBA Playing Style Analysis

3.2.1 Description

One of the most debated questions in modern basketball is whether offensive or defensive philosophy is the more reliable path to winning. The “pace and space” offensive philosophy — characterised by high three point volume, efficient three point shooting, and fast paced, spread out offences — has dominated the NBA conversation since the Golden State Warriors’ dynasty of the mid-2010s. The counterargument, captured in the classic adage “defence wins championships”, holds that elite defensive teams are more consistent winners, particularly in the playoffs where games slow down, schemes become more targeted, and physicality increases over a seven game series.

This use case classifies NBA teams from 2016 to 2025 into three playing style categories — Pace & Space, Defensive, and Balanced — using season level averages of three point attempts, three point percentage, steals, blocks, and opponent score. A fourth category, Elite, is introduced for teams that qualify as both Pace & Space and Defensive, representing the most complete teams in the league. Win rates are then compared across styles to determine which approach is most effective in the regular season and playoffs, and the 2025 NBA champion Oklahoma City Thunder are examined as a case study.

Teams were classified as follows:

  • Pace & Space — top third in three point attempts AND top third in three point percentage in their season
  • Defensive — top third in combined steals and blocks AND bottom third in opponent score in their season
  • Elite — qualifying for both Pace & Space and Defensive criteria simultaneously
  • Balanced — all remaining teams

Note: The 2018 NBA season was excluded from this analysis due to the data quality issues identified in the missing values and outliers section, which significantly affected the steals and blocks columns used in the style classification.

3.2.2 Visualisations

Table 4 presents the average season statistics for each playing style classification from 2016 to 2025, excluding 2018.

Table 4: Average Season Stats by Playing Style - NBA (2016-2025, excl. 2018)
Style 3P Attempts 3P% Steals Blocks Opp. Score Win Rate
Pace & Space 37.4 37.6 7.3 4.8 110.7 0.616
Defensive 31.6 36.1 8.3 5.4 106.7 0.609
Elite 36.0 37.1 8.3 5.4 105.9 0.681
Balanced 32.0 35.5 7.6 4.8 111.6 0.456

The table confirms that the classification system captures meaningful statistical differences between styles. Pace & Space teams lead in three point attempts (37.4) and three point percentage (37.6%), while Defensive teams lead in steals (8.3) and blocks (5.4) and concede the fewest points (107.0). Elite teams show the best of both worlds, combining high three point attempts (36.0) with strong defensive numbers and the lowest opponent score (106.0). Notably, both Pace & Space (0.625) and Defensive (0.609) teams have significantly higher win rates than Balanced teams (0.456), while Elite teams lead all categories (0.681).

Figure 35 displays the average win rate by playing style for NBA teams from 2016 to 2025, excluding 2018.

Elite teams have the highest win rate (68.1%), followed closely by Pace & Space (62.5%) and Defensive (60.9%) teams. Balanced teams trail significantly at 45.6%, suggesting that teams without a clear stylistic identity struggle to compete with those that have a defined philosophy. The near-identical win rates of Pace & Space and Defensive teams is a compelling finding — both styles win at roughly equal rates, directly challenging the notion that one approach is universally superior to the other.

Figure 36 displays the number of Pace & Space and Defensive teams per season from 2016 to 2025, excluding 2018, showing how the distribution of playing styles has changed over time.

The distribution of Pace & Space teams has remained relatively stable across seasons, while the number of Defensive teams shows more variation from year to year. Elite teams are rare across the entire period, appearing in only five team seasons across nine years, reflecting the difficulty of simultaneously excelling at both offensive and defensive dimensions of the game. The overall trend does not suggest a dramatic shift toward one style over another at the team level, though the NBA as a whole has clearly moved toward higher three point volume as shown in the data description section.

Figure 37 displays the Pace & Space classified teams and their win rates over time from 2016 to 2025, excluding 2018. Team logos identify individual franchises.

The Golden State Warriors appear as a Pace & Space team in multiple seasons (2016, 2017, 2019, 2022, 2023, 2024), reflecting their identity as the team most associated with the pace and space revolution. The 2016 Warriors (89% win rate) — the greatest regular season team in NBA history at 73-9 — are the highest point in the chart, providing compelling evidence for the effectiveness of the pace and space philosophy at its peak. Another notable Pace & Space teams includes the Boston Celtics in recent seasons have been among the more successful teams in their eras.

Figure 38 displays the Defensive classified teams and their win rates over time from 2016 to 2025, excluding 2018. Team logos identify individual franchises.

The San Antonio Spurs appear as a Defensive team in 2016 and 2017, consistent with their long-standing reputation as one of the NBA’s premier defensive organisations. The Oklahoma City Thunder appear as a Defensive team in 2025 with a remarkable 81.6% win rate — the highest of any Defensive team in the dataset — confirming that their 2025 championship was built on a defensive foundation. This is seen in their roster with multiple All-Defensive candidates in Chet Holmgren, Alex Caruso, Lu Dort, and Cason Wallace. The Philadelphia 76ers, Toronto Raptors, and Milwaukee Bucks also feature prominently as Defensive teams across multiple seasons, reflecting the success of defence-first philosophies in the modern NBA.

Figure 39 displays the five Elite team seasons identified across the dataset, showing team logos and win rates.

Only five team seasons qualify as Elite across nine seasons of data — the LA Clippers (2016, 64.6%), Toronto Raptors (2020, 73.6%), Golden State Warriors (2022, 64.6%), Boston Celtics (2019, 59.8%) and Boston Celtics (2024, 78.0%). The Boston Celtics 2024 season is particularly notable as they went on to win the NBA Championship that year, and their 78.0% win rate was the highest of any Elite team. The Golden State Warriors 2022 also resulted in an NBA Championship, meaning two of the five Elite team seasons ended in championships — a remarkable conversion rate that strongly supports the idea that teams capable of excelling at both offensive and defensive dimensions simultaneously are the most likely to win titles.

Figure 40 displays the NBA champions from 2016 to 2025 and their playing style classification, excluding 2018.

The champions by style breakdown reveals a remarkable balance — three Pace & Space champions (Cleveland Cavaliers 2016, Golden State Warriors 2017, Milwaukee Bucks 2021), three Defensive champions (Toronto Raptors 2019, Los Angeles Lakers 2020, Oklahoma City Thunder 2025), two Elite champions (Golden State Warriors 2022, Boston Celtics 2024), and one Balanced champion (Denver Nuggets 2023). This near-perfect split directly challenges any claim that one style is superior — both pace and space and defensive philosophies have proven equally capable of producing champions in the modern NBA.

Figure 41 displays the win rate by playing style for regular season versus playoffs from 2016 to 2025, excluding 2018.

The playoff versus regular season comparison yields one of the most surprising findings in this analysis. Elite teams are the only style that performs better in the playoffs (68.7%) than in the regular season (68.0%), suggesting that complete two-way teams are uniquely well-suited to the demands of postseason basketball. Defensive teams show the largest drop-off from regular season (60.8%) to playoffs (50.4%), challenging the conventional wisdom that defence wins championships and suggesting that elite defence alone is insufficient in the postseason. Pace & Space teams show a more modest decline (61.6% to 58.0%), holding up relatively well in the playoffs, possibly because offensive firepower remains valuable regardless of game context. Balanced teams perform almost identically in both contexts (45.7% vs 45.1%), suggesting they lack the defining strengths needed to elevate their performance in the postseason.

Figure 42 displays the average season statistics by playing style for NBA teams from 2016 to 2025, excluding 2018.

The style profiles are clearly distinguishable in the statistical breakdown. Pace & Space teams lead in three point attempts (37.4) and three point percentage (37.6%), while Defensive teams lead in steals (8.3) and blocks (5.4) and allow the fewest opponent points (107.0). Elite teams show the most balanced profile across all five statistics, combining high three point activity with strong defensive metrics. Balanced teams sit in the middle across most categories, reinforcing their identity as teams without a defining statistical strength.

3.2.3 Hypothesis

It was hypothesised that Pace & Space teams would have higher win rates in the regular season given the modern NBA’s emphasis on three point shooting, but that Defensive teams would outperform in the playoffs as games slow down and defensive schemes become more impactful over a seven game series.

The data only partially supports this hypothesis. In the regular season, Pace & Space teams (61.6%) do slightly outperform Defensive teams (60.8%), consistent with the hypothesis. However, in the playoffs, the opposite of what was expected occurred — Defensive teams show a dramatic drop-off to 50.4%, while Pace & Space teams hold up better at 58.0%. This finding challenges the conventional wisdom that defence wins championships and suggests that in the modern NBA, offensive capability may actually be more durable in the postseason than defensive ability alone.

The Elite category provides the most compelling finding — teams that combine both Pace & Space and Defensive qualities are the only style to improve in the playoffs, achieving a 68.7% win rate compared to 68.0% in the regular season. This suggests that the most championship-ready teams in the modern NBA are those that cannot be neutralised by a single strategic adjustment — they can win through both offensive efficiency and defensive dominance depending on the game context.

The champions by style breakdown further reinforces the finding that no single style dominates — three Pace & Space champions, three Defensive champions, two Elite champions, and one Balanced champion across nine seasons demonstrates that multiple paths to a championship exist in the modern NBA.

3.2.4 Limitations

Several limitations should be considered when interpreting the findings of this use case.

First, the playing style classification is based on season level averages rather than game-by-game or possession-level data, meaning a team classified as Pace & Space based on their season average may have played very differently in individual games or during the playoffs. A more granular analysis using playoff-specific data would provide a more accurate picture of how teams actually perform during the postseason.

Second, the classification thresholds are defined using within-season percentile rankings (top third), meaning the absolute statistical values required to qualify as Pace & Space or Defensive vary from season to season. A team that was Pace & Space in 2016 may not qualify in 2025 as the entire league has shifted toward higher three point volume, potentially creating inconsistencies in how teams are classified across different eras.

Third, the analysis is limited to the NBA and does not include the WNBA, NCAAM or NCAAW, meaning the findings cannot be generalised to other competitions where the pace and space revolution has developed differently and where the defensive landscape may reward different approaches.

Fourth, the 2018 season was excluded due to data quality issues, reducing the sample from ten to nine seasons. While this is unlikely to fundamentally change the findings, it does slightly reduce the robustness of the analysis.

Finally, coaching, player quality, and roster construction are not accounted for in this analysis. A team with an elite superstar may win championships regardless of their stylistic classification, meaning the four-factor style approach provides a useful but necessarily simplified framework for understanding team success.

4 Conclusion

This report has presented a comprehensive exploratory data analysis of game-by-game team performance data across four basketball competitions — the NBA, WNBA, NCAAM and NCAAW — spanning ten seasons from 2016 to 2025. The dataset of 246,748 observations provided a rich foundation for examining scoring patterns, shooting trends, rebounding, assists, defensive activity, and home advantage across professional and collegiate basketball.

The data description section revealed several compelling findings about the nature of modern basketball. The rise of three point shooting is well documented across all four competitions, with NBA three point attempts increasing 54% from 2016 to 2025. However, the data challenges the simplistic narrative that more threes automatically leads to better shooting — college basketball shows declining three point efficiency despite increasing volume, suggesting that shot selection is deteriorating as the three pointer becomes more fashionable at lower levels. Similarly, the popular “foul baiting” narrative is directly contradicted by the data, with free throw attempts declining across all four competitions over the decade. The normalised per 40 minute analysis also revealed that the WNBA leads all competitions in assists per 40 minutes, suggesting a more team oriented style of play than is commonly appreciated, while NCAAW’s high steals numbers were contextualised by their equally high turnover rates.

The NBA specific fastbreak analysis provided an interesting team level view of transition scoring, with the Toronto Raptors and Oklahoma City Thunder identified as the leading fastbreak teams over the period. The Thunder’s emergence as a dominant fastbreak team in recent seasons culminated in their 2025 NBA Championship, providing a compelling narrative thread between the data description and use case sections.

Use Case 1 applied Dean Oliver’s Four Factors framework across all four competitions and confirmed that effective field goal percentage is the single most consistent predictor of winning, with correlations ranging from 0.458 in the NBA to 0.519 in NCAAW. Turnover rate is the second most important factor, while offensive rebound rate showed a counterintuitive negative correlation with winning in professional leagues, suggesting that chasing offensive rebounds may leave teams vulnerable in transition defence. The 2025 champion analysis provided a practical application of the framework — the Oklahoma City Thunder’s historically low turnover rate and the UConn Huskies’ dominant shooting efficiency both reflect teams that excelled in the factors most strongly associated with winning, offering coaches and analysts a clear statistical blueprint for championship calibre basketball.

Use Case 2 addressed the long-standing debate between pace and space and defensive basketball philosophies. The analysis revealed that both styles win at near-identical regular season rates (Pace & Space 62.5%, Defensive 60.9%), directly challenging any claim that one approach is universally superior. The most surprising finding was that Defensive teams show the largest drop-off in the playoffs (60.8% to 50.4%), contrary to the popular belief that defence wins championships. Elite teams — those capable of excelling at both offensive and defensive dimensions simultaneously — are the only style to improve in the playoffs, and produced two of the nine NBA champions in the analysis period. The finding that three Pace & Space, three Defensive, two Elite, and one Balanced team won the NBA Championship across nine seasons demonstrates that multiple valid paths to a championship exist in the modern NBA.

Taken together, the findings of this report suggest that the most effective teams in modern basketball are those that shoot efficiently, protect the ball, and are capable of competing in multiple stylistic dimensions. The data challenges several popular narratives — including the supremacy of defence in the playoffs, the value of offensive rebounding for champion teams, and the effectiveness of foul drawing as a winning strategy — and provides evidence based alternatives that coaches, analysts, and front office staff can use to inform roster construction and strategic decision making.

Future research could extend this analysis in several directions. Play-by-play data would enable more precise possession-level calculations of the four factors and more accurate playoff-specific style classifications. Incorporating player-level data would allow for analysis of how individual player profiles contribute to team style classifications and championship outcomes. Finally, extending the playing style analysis to the WNBA, NCAAM and NCAAW would provide a more complete picture of whether the pace and space versus defence debate plays out differently across competitions and levels of play.