The four factors are a useful way to break down basketball efficiency because they focus on the areas that usually decide games: shooting, turnovers, rebounding, and free throws. Instead of just saying a team was good or bad on offense or defense, they show where that performance actually came from.
For this analysis, I used the 2024 WNBA Advanced Stats table from Basketball-Reference. After copying the table into Excel and saving it as a CSV, I loaded the data into R to compare how well the four factors explained Offensive Rating and Defensive Rating across the league.
The table below includes the team ratings and the offensive and defensive four-factor numbers used in the two regression models. The logos are included just to make the table easier to read, and the color shading helps show where teams were stronger or weaker in each category.
| 2024 WNBA Advanced Stats Used in the Regression Models | |||||||||||
| Offensive and defensive four-factor data by team | |||||||||||
| TEAM |
RATINGS
|
OFFENSIVE FOUR FACTORS
|
DEFENSIVE FOUR FACTORS
|
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ORTG | DRTG | OFF EFG% | OFF TOV% | OFF ORB% | OFF FT/FGA | DEF EFG% | DEF TOV% | DEF DRB% | DEF FT/FGA | ||
| New York Liberty | 109.6 | 97.9 | 0.521 | 14.2 | 25.1 | 0.203 | 0.476 | 14.5 | 79.2 | 0.167 | |
| Connecticut Sun | 105.0 | 96.4 | 0.488 | 13.8 | 25.2 | 0.239 | 0.482 | 17.6 | 78.2 | 0.204 | |
| Minnesota Lynx | 104.6 | 96.5 | 0.518 | 15.3 | 22.3 | 0.182 | 0.460 | 16.5 | 74.2 | 0.181 | |
| Las Vegas Aces | 108.0 | 101.2 | 0.523 | 12.4 | 16.6 | 0.223 | 0.488 | 14.1 | 79.5 | 0.189 | |
| Seattle Storm | 104.2 | 98.6 | 0.478 | 13.5 | 24.2 | 0.211 | 0.477 | 16.5 | 74.6 | 0.210 | |
| Indiana Fever | 106.1 | 109.5 | 0.523 | 15.7 | 24.6 | 0.194 | 0.507 | 12.8 | 76.7 | 0.228 | |
| Atlanta Dream | 99.0 | 102.5 | 0.452 | 14.0 | 25.4 | 0.227 | 0.488 | 14.3 | 78.3 | 0.208 | |
| Washington Mystics | 99.7 | 103.4 | 0.506 | 17.0 | 20.6 | 0.172 | 0.502 | 16.7 | 75.2 | 0.241 | |
| Phoenix Mercury | 103.6 | 107.8 | 0.503 | 15.1 | 20.0 | 0.222 | 0.496 | 13.1 | 71.6 | 0.189 | |
| Chicago Sky | 99.1 | 105.6 | 0.457 | 14.3 | 29.5 | 0.187 | 0.499 | 14.2 | 76.8 | 0.228 | |
| Los Angeles Sparks | 98.6 | 107.7 | 0.478 | 16.6 | 22.0 | 0.226 | 0.522 | 14.2 | 76.1 | 0.211 | |
| Dallas Wings | 104.2 | 114.0 | 0.490 | 15.8 | 30.2 | 0.206 | 0.536 | 13.8 | 73.4 | 0.236 | |
The first model looks at offense. I regressed ORtg on the four offensive factors together: offensive effective field goal percentage, offensive turnover percentage, offensive rebound percentage, and offensive free throws per field goal attempt.
The idea is pretty straightforward. If the four factors really explain offensive efficiency, then they should explain a large portion of the differences in Offensive Rating from team to team.
| Offensive Regression Coefficients | ||||
| ORtg regressed on the offensive four factors | ||||
| TERM | ESTIMATE | STANDARD ERROR | T STATISTIC | P VALUE |
|---|---|---|---|---|
| Intercept | 25.4153 | 9.5264 | 2.6679 | 0.0321 |
| Offensive eFG% | 158.4509 | 11.8203 | 13.4050 | 0.0000 |
| Offensive TOV% | −1.2798 | 0.1946 | −6.5770 | 0.0003 |
| Offensive ORB% | 0.4288 | 0.0746 | 5.7459 | 0.0007 |
| Offensive FT/FGA | 40.4983 | 13.7468 | 2.9460 | 0.0215 |
The offensive regression had an R-squared of 0.972. That means the offensive four factors explained about 97.2% of the variation in Offensive Rating across WNBA teams.
The second model looks at defense. I regressed DRtg on the four defensive factors together: opponent effective field goal percentage, opponent turnover percentage, defensive rebound percentage, and opponent free throws per field goal attempt.
This is the same general idea as the offensive model, but flipped to the defensive side. A good defense should force worse shooting, create turnovers, finish possessions with rebounds, and avoid sending teams to the line too often.
| Defensive Regression Coefficients | ||||
| DRtg regressed on the defensive four factors | ||||
| TERM | ESTIMATE | STANDARD ERROR | T STATISTIC | P VALUE |
|---|---|---|---|---|
| Intercept | 92.9998 | 9.3435 | 9.9534 | 0.0000 |
| Defensive eFG% | 129.2923 | 14.2768 | 9.0561 | 0.0000 |
| Defensive TOV% | −1.6573 | 0.1383 | −11.9841 | 0.0000 |
| Defensive DRB% | −0.5172 | 0.0676 | −7.6497 | 0.0001 |
| Defensive FT/FGA | 50.6029 | 11.0244 | 4.5901 | 0.0025 |
The defensive regression had an R-squared of 0.994. So the defensive four factors explained about 99.4% of the variation in Defensive Rating across teams.
| R-Squared Comparison | |||
| Which side of the ball was explained better by the four factors? | |||
| SIDE | REGRESSION | R SQUARED | PERCENT EXPLAINED |
|---|---|---|---|
| Offense | ORtg on offensive four factors | 0.972 | 97.2% |
| Defense | DRtg on defensive four factors | 0.994 | 99.4% |
The offensive regression had an R-squared of 0.972, while the defensive regression had an R-squared of 0.994. I got these numbers by running two separate multiple linear regressions in R. The first regression used ORtg, or Offensive Rating, as the outcome variable and the four offensive factors as the explanatory variables. The second regression used DRtg, or Defensive Rating, as the outcome variable and the four defensive factors as the explanatory variables.
The R-squared value shows how much of the variation in each respective rating is explained by the model. In the offensive regression, the R-squared of 0.972 means the four offensive factors explained about 97.2% of the variation in Offensive Rating across teams. In the defensive regression, the R-squared of 0.994 means the four defensive factors explained about 99.4% of the variation in Defensive Rating.
Both numbers are extremely high, which shows that the four factors play a major role in explaining team efficiency. That makes sense because shooting efficiency, turnovers, rebounding, and free throw rate are all directly connected to how teams score points as well prevent points from being scored.
Since the defensive model had the higher R-squared, the answer to the main question is that defense was more easily explained by four-factor variation across teams. The difference between the two models is small, so offense was still explained very well, but defense had the slightly stronger fit in this dataset.