nsaidi_Data101 Final
Nassor Saidi
2026-06-30
load the library
library(tidyverse)load the data
setwd("~/Nassor/MC/Data101")
boxing <- read_csv("boxing_pay_data.csv")## Rows: 4670 Columns: 27
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (3): Boxer, Date, Venue
## dbl (24): Purse, lnRPurse, weight, Age, Wins, Losses, KO, W-Title, PPV, ESPN...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Part A: Introduction
This project seeks to answer the research question - Is a boxers wins related to their earnings?
The money involved in professional entertainment gets a lot of
attention. Professional athletes tend to be some of the highest paid
individuals globally. This is no different in the combat sports world
where boxer pay is scrutinized. Names like Floyd Money
Mayweather” and “Saul Canelo Alvarez” have made a name for
the money they generate as athletes in the sport of boxing.
Professional boxers or “prize fighters” and legally classified as independent contractors and must negotiate contract agreements to get paid for their work. A purse is the total amount of “prize” money a boxer will be paid - including up front and back end money. However, the average boxer is not in a position to earn much from their fights and according to the Latino Policy and Politics Institute, most professional boxers need to work multiple jobs. Within a changing landscape of professional combat sports, the athletes themselves tend to be under-informed on the business-side of their profession. This exploration will seek insights to inform larger conversations about fighter pay as professional athletes.
This project will begin by looking into wins as a predictor for increased earnings, but will also explore other variables that may also contribute to higher pay.
First Let’s Take a Look at the data
dim(boxing)## [1] 4670 27head(boxing)## # A tibble: 6 × 27
## Boxer Date Venue Purse lnRPurse weight Age Wins Losses KO `W-Title`
## <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 Gamboa, … 2009… Buff… 37998 10.7 126 27.1 12 0 10 0
## 2 Gonzalez… 2009… Buff… 20000 10.0 127 30.7 27 2 18 0
## 3 Lara, Er… 2009… Buff… 3000 8.14 154 25.8 2 0 1 0
## 4 Bogere, … 2009… Buff… 1500 7.45 135 20.2 4 0 2 0
## 5 Frazier,… 2009… Buff… 1000 7.04 139 30.9 1 0 1 0
## 6 Gonzalez… 2009… Buff… 1500 7.45 146 31.5 2 1 1 0
## # ℹ 16 more variables: PPV <dbl>, ESPN <dbl>, HBO <dbl>, FOX <dbl>,
## # TopRank <dbl>, GoldenBoy <dbl>, RDS <dbl>, Y2009 <dbl>, Y2010 <dbl>,
## # Y2011 <dbl>, Y2012 <dbl>, Y2013 <dbl>, Y2014 <dbl>, Y2015 <dbl>,
## # Y2016 <dbl>, Y2017 <dbl>Key Variables
- Wins
- LnRPurse
- KO
- PPV
- Venue
Part B: Clean the Dataset and Conduct an Exploratory Data Analysis (EDA)
Exploring the class of key variables
summary(boxing$KO)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.000 1.000 4.000 6.881 10.000 51.000class(boxing$KO)## [1] "numeric"summary(boxing$Wins)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 3.00 8.00 11.39 17.00 65.00class(boxing$Wins)## [1] "numeric"max(boxing$Wins)## [1] 65#Max number of wins is 65
summary(boxing$lnRPurse)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 6.587 7.366 8.195 8.567 9.210 18.455unique(boxing$PPV)## [1] 0 1class(boxing$PPV)## [1] "numeric"boxing_venue <- boxing |>
group_by(Venue) |>
count() |>
arrange(desc(n))|>
head(10)
boxing_venue## # A tibble: 10 × 2
## # Groups: Venue [10]
## Venue n
## <chr> <int>
## 1 MGM Grand 313
## 2 Doubletree Hotel 272
## 3 Fantasy Springs 268
## 4 Hard Rock 261
## 5 Mandalay Bay 185
## 6 StubHub 176
## 7 MGM Arena 161
## 8 CosmoLV 160
## 9 Texas Station 122
## 10 Belasco Theater 112options(scipen = 999)
boxer_count <- boxing |>
group_by(Boxer) |>
summarize(count = n(), mean_purse = mean(Purse), max_purse = max(Purse), mean_win = mean(Wins), mean_KO = mean(KO)) |>
arrange(desc(count)) |>
filter(count >= 10)
boxer_count## # A tibble: 57 × 6
## Boxer count mean_purse max_purse mean_win mean_KO
## <chr> <int> <dbl> <dbl> <dbl> <dbl>
## 1 Magdaleno, Jessie 21 17576. 155000 11.5 8.52
## 2 Chavez, Joaquin R. 18 3000 6000 4.39 1.78
## 3 Vargas, Jessie 18 227028. 2800000 16.6 7.39
## 4 Magdaleno, Diego 17 11788. 25000 16.2 5.24
## 5 Diaz, Joseph Jnr. 16 32250 200000 11.1 6.75
## 6 Bogere, Sharif 15 11647. 75000 16.1 10.3
## 7 De La Hoya, Diego 15 13967. 80000 7.4 4.8
## 8 Gutierrez, Jesus A. 15 3343. 10000 7.2 2.27
## 9 Breazeale, Dominic 14 10464. 50000 7 6.64
## 10 Gavril, Ronald 14 21143. 125000 10.9 8.29
## # ℹ 47 more rowsfiltering for top 10 venues
boxing1 <- boxing |>
filter(Venue %in% c("MGM Grand", "Doubletree Hotel", "Fantasy Springs", "Hard Rock", "Mandalay Bay", "StubHub", "MGM Arena", "CosmoLV", "Texas Station", "Belasco Theater"))Check the distribution for number of wins
There are zeros so we have to use square root as opposed to the log transformation.
summary(boxing1$Wins)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 5.00 12.00 14.15 21.00 65.00ggplot(boxing1, aes(x=Wins))+
geom_density()
The distribution of wins is strongly skewed right.
Perform square root transformation for Wins
ggplot(boxing1, aes(x=sqrt(Wins)))+
geom_density()
The square root function was able to make the distribution of wins more bell shaped
Check distribution for lnRPurse
ggplot(boxing1, aes(x=lnRPurse)) +
geom_density()
Unable to transform this distribution because the numeric values for purse were already log transformed
Creating a vizualization
ggplot(boxing1, aes(x = Wins, y = lnRPurse)) +
geom_point(alpha = 0.2) +
geom_smooth(method = "lm", color = "red", se = FALSE) +
labs(title = "Boxer Wins vs Purse",
x = "Wins", y = "Fight Purse") +
theme_minimal()
Scatterplot was selected to visualize the relationship of these two variables, which appears to be linear, meaning as wins increase, so does log purse. However, most of the wins were in 0 - 30 range and most of the purses were were around or below log 10 ($22026.5). Furthermore, we see much more variability on the upper end of our regression line.
Statistical Analysis: Regression Analysis
Compute a correlation coefficient between key variables
cor(boxing$Wins, boxing$lnRPurse, use = "complete.obs")## [1] 0.7635325cor(boxing$PPV, boxing$lnRPurse, use = "complete.obs")## [1] 0.5261753cor(boxing$KO, boxing$lnRPurse, use = "complete.obs")## [1] 0.7394403Perform Simple Regression
Predictor Variable(s): Wins Response Variable: Purse
options(scipen = 0)
model <- lm(lnRPurse ~ Wins, data = boxing1)
summary(model)##
## Call:
## lm(formula = lnRPurse ~ Wins, data = boxing1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.9618 -0.5350 -0.1228 0.4463 5.8516
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.419804 0.042128 176.1 <2e-16 ***
## Wins 0.120267 0.002313 52.0 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.195 on 2028 degrees of freedom
## Multiple R-squared: 0.5715, Adjusted R-squared: 0.5712
## F-statistic: 2704 on 1 and 2028 DF, p-value: < 2.2e-16y = 7.42 + 0.12x - For every Win, the predicted lnRPurse goes up by 0.12 points. - If a boxer has 0 Wins, the predicted lnRPurse is 7.42 Adjusted R-squared: More than half of the variation are explained by this model (57.15%). The p value is close to zero: Having a p value < alpha of 0.05, indicates that wins are a meaningful predictor of log purse.
Predictions for wins in table format
winput <- data.frame(Wins = c(5, 15, 25, 35, 45, 55, 65))
purse_estimate <- exp(predict(model, winput))
cbind(winput, purse_estimate)## Wins purse_estimate
## 1 5 3044.636
## 2 15 10135.528
## 3 25 33740.963
## 4 35 112322.963
## 5 45 373920.804
## 6 55 1244774.570
## 7 65 4143828.624plot(model, which = 1)
plot(model, which = 2)
The residuals are not randomly scattered and the QQ plot does not
fit. - Residuals vs Fitted Plot shows that most residuals are
clumped at the left-most side of the line of fit, demonstrating the
funnel shape. - The QQ Residuals Plot shows a non-normal distribution of
the residuals at the tails ends of the line.
Run RMSE
rmse <- sqrt(mean(residuals(model)^2))
rmse## [1] 1.194695On average, my predicted fight purse is off by about 1.195
rmse points.”
Multiple Regression
Predictor Variable(s): Wins(#), KO(#), PPV(0 or 1), Venue (limited to the 10 most prevalent venues in the dataset) Response Variable: Purse
multi_model1 <- lm(lnRPurse ~ Wins + KO + Venue + PPV, data = boxing1)
summary(multi_model1)##
## Call:
## lm(formula = lnRPurse ~ Wins + KO + Venue + PPV, data = boxing1)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.3575 -0.4606 -0.0662 0.3916 5.3464
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.511124 0.098788 76.032 < 2e-16 ***
## Wins 0.075174 0.005664 13.272 < 2e-16 ***
## KO 0.023850 0.007553 3.158 0.001614 **
## VenueCosmoLV 0.432796 0.127400 3.397 0.000694 ***
## VenueDoubletree Hotel -0.441198 0.115152 -3.831 0.000131 ***
## VenueFantasy Springs -0.055508 0.115413 -0.481 0.630604
## VenueHard Rock 0.113849 0.116315 0.979 0.327799
## VenueMandalay Bay 0.299973 0.124036 2.418 0.015676 *
## VenueMGM Arena 0.250859 0.132793 1.889 0.059023 .
## VenueMGM Grand 0.662178 0.116842 5.667 1.66e-08 ***
## VenueStubHub 0.545236 0.125930 4.330 1.57e-05 ***
## VenueTexas Station -0.106896 0.134190 -0.797 0.425778
## PPV 1.840537 0.097644 18.850 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.025 on 2017 degrees of freedom
## Multiple R-squared: 0.6865, Adjusted R-squared: 0.6846
## F-statistic: 368.1 on 12 and 2017 DF, p-value: < 2.2e-16Adjusted R-squared: Nearly 70% of the variation are explained by this model (0.6846). The p value is close to zero for Wins and PPV, very small for some Venues such as “MGM Grand” and “StubHub”. KO and a couple other Venues (“Fantasy Springs”, “Doubletree Hotel”, and “Mandalay Bay”) had a p value < alpha 0.05, while other Venues were not relevant to this prediction model.
plot(multi_model1, which = 1)
plot(multi_model1, which = 2)
The residuals are not randomly scattered and the QQ plot does not
fit.
rmse <- sqrt(mean(residuals(multi_model1)^2))
rmse## [1] 1.021824Part D: Conclusion and Future Directions
I began this exploration by asking - Is a boxers wins related to their earnings? 1. I looked over my distributions and plotted the variables. 2. I created a prediction model using a two-part regression analysis. 3. I made predictions and checked how well my model made predictions.
The results indicated a linear relationship between wins and log of
the purse. However, the model was off by 1.195 rmse points
and did not demonstrate homoscedasticity or normalcy of the residuals.
In this case, the initial predictor variable was significant, however
the accuracy of the model increased after adding other predictor
variables and the model became more effective at explaining the overall
variance in the dependent variable. In particular, the PPV demonstrated
a strong influence on the response variable. I also ran a check for the
Root Mean Square Error (RMSE) to measure the average difference between
the predicted values and the actual observed values. I found that on
average the multiple regression model was off by 1.02 rmse
points, a 0.173 point improvement from the simple linear
regression.However, the diagnostic plots did not improve much,
indicating that this linear approach is not the best model for observing
the relationship of these variables. Moving forward with this dataset, I
would explore how well Pay Per View (PPV) and venue explain for
earnings, given that it had the strongest magnitude in my multiple
regression. In professional boxing, some broadcasting platforms have
transformed to subscription based and no longer host PPV events. I
wonder how this might affect boxer’s ability to negotiate for pay.
Additionally, I would look for a dataset to compare with a MMA fighter
earnings to further inform the discussion on fighter pay.
Part E: References
Anderson, Peter (2019), “Boxing Data”, Mendeley Data, V1, doi: 10.17632/vpbsd5bryy.1
Mondragón, R., Valenzuela, A., Jr., & Hernández, J. M. (2024, May 16). Down but not out: Labor struggles for professional boxers in California’s ring. UCLA Latino Policy and Politics Institute. https://latino.ucla.edu/wp-content/uploads/2024/05/Down-but-Not-Out-Labor-Struggles-for-Professional-Boxers-in-Californias-Ring.pdf