FRsmall

Alexis Drakopoulos

Analysis of Powerlifting Dataset

This is an analysis of a large scale dataset of over 3,000 meets and 300,000 attempts. This dataset only includes athletes that have attended and attempts lifts at these meets. This means the data may not be representative of some populations and may not apply to others. However there may be numbers or figures that are of interest to you, so I hope you enjoy it. The dataset and code is available and on each section you can toggle to see it. The data analysis/conclusions are all done by Alexis Drakopoulos.

Everything was done with R.

Packages Used: 

      library(tidyverse)
      library(data.table)
      install.packages("ggcorrplot")
      library(ggcorrplot)
    # Importing our files
          powerliftingdata <- fread('openpowerlifting.csv')
          meetsdata <- fread('meets.csv')

Contents

Data Cleaning

During import of the dataset there were various apparent problems with the data, ranging from missing or negative values, to ages that were not relevant such as a 5 year old. Here is a list of what I did to clean the data, you can skip this section if you are only interested in the results.

    > glimpse(powerliftingdata)
Observations: 386,414
Variables: 17
$ MeetID          0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,...
$ Name            "Angie Belk Terry", "Dawn Bogart", "Dawn Bogart", "Dawn Bogart", "Destiny Dula", "Courtney N...
$ Sex             "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F", "F...
$ Equipment       "Wraps", "Single-ply", "Single-ply", "Raw", "Raw", "Wraps", "Raw", "Raw", "Wraps", "Raw", "W...
$ Age             47, 42, 42, 42, 18, 28, 60, 60, 52, 52, 24, 56, 56, 54, 17, 32, 25, 18, 22, 56, 30, 44, 34, ...
$ Division        "Mst 45-49", "Mst 40-44", "Open Senior", "Open Senior", "Teen 18-19", "Open Senior", "Mst 60...
$ BodyweightKg    59.60, 58.51, 58.51, 58.51, 63.68, 62.41, 67.31, 67.31, 65.95, 65.95, 65.50, 71.21, 71.21, 7...
$ WeightClassKg   "60", "60", "60", "60", "67.5", "67.5", "67.5", "67.5", "67.5", "67.5", "67.5", "75", "75", ...
$ Squat4Kg        NA, NA, NA, NA, NA, -183.7, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, -183.7, ...
$ BestSquatKg     47.63, 142.88, 142.88, NA, NA, 170.10, 124.74, 124.74, 120.20, NA, 138.35, 120.20, 120.20, 1...
$ Bench4Kg        NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, ...
$ BestBenchKg     20.41, 95.25, 95.25, 95.25, 31.75, 77.11, 95.25, 95.25, 54.43, NA, 65.77, 43.09, 43.09, 74.8...
$ Deadlift4Kg     NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, 204.12, NA, ...
$ BestDeadliftKg  70.31, 163.29, 163.29, NA, 90.72, 145.15, 163.29, 163.29, 108.86, 108.86, 136.08, 129.27, 12...
$ TotalKg         138.35, 401.42, 401.42, 95.25, 122.47, 392.36, 383.28, 383.28, 283.49, 108.86, 340.20, 292.5...
$ Place           "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1...
$ Wilks           155.05, 456.38, 456.38, 108.29, 130.47, 424.40, 391.98, 391.98, 294.25, 112.99, 354.89, 287....
    > glimpse(meetsdata)
Observations: 8,482
Variables: 8
$ MeetID       0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 2...
$ MeetPath     "365strong/1601", "365strong/1602", "365strong/1603", "365strong/1604", "365strong/1605", "365s...
$ Federation   "365Strong", "365Strong", "365Strong", "365Strong", "365Strong", "365Strong", "365Strong", "365...
$ Date         "2016-10-29", "2016-11-19", "2016-07-09", "2016-06-11", "2016-04-10", "2017-04-22", "2017-01-21...
$ MeetCountry  "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA", "USA...
$ MeetState    "NC", "MO", "NC", "SC", "SC", "NC", "GA", "MO", "SC", "NC", "IN", "IN", "NC", "SC", "CA", "NV",...
$ MeetTown     "Charlotte", "Ozark", "Charlotte", "Rock Hill", "Rock Hill", "Charlotte", "Martinez", "Lake Oza...
$ MeetName     "2016 Junior & Senior National Powerlifting Championships", "Thanksgiving Powerlifting Classic"...

Missing and Negative Values

There were a total of 240,000 missing values, of which 2,402 missing bodyweights. There were also 3,500 people above the age of 80 and below the age of 15, which I removed. I also removed columns such as divisions due to there being over 3,000 of them, and recalculated wilks and totals myself.

The data had 136,687 distinct names, meaning the average person went to nearly 3 meets.

The data also had 2,919 negative values, these were most likely a typo in data entry, nevertheless we will remove them for data analysis since there are so few of them. 

    > powerliftingdata %>%
+     distinct(Name) %>%
+     count()
# A tibble: 1 x 1
       n
   
1 136687

> powerliftingdata %>%
+     select(BodyweightKg) %>%
+     is.na() %>%
+     sum()
[1] 2402

> powerliftingdata %>%
+     select(Age) %>%
+     summarize(missingval = sum(is.na(Age)),
+               min = min(Age, na.rm = T),
+               max = max(Age, na.rm = T),
+               mean = mean(Age, na.rm = T))
  missingval min max     mean
1     239267   5  95 31.66824

> powerliftingdata %>%
+     select(Division) %>%
+     distinct() %>%
+     count()
# A tibble: 1 x 1
      n
  
1  4247

> powerliftingdata %>%
+   filter( BodyweightKg <= 0 |
+             BestSquatKg <= 0 |
+             BestBenchKg <= 0 |
+             BestDeadliftKg <= 0) %>%
+   count()
# A tibble: 1 x 1
      n
  
1  2919
#Removing unecessary columns, extreme ages and missing body weights
powerliftingdataclean <- powerliftingdata %>%
  select(-Division, -Squat4Kg, -Bench4Kg, -Deadlift4Kg,
        -TotalKg, -Wilks, -WeightClassKg, -Place) %>%
  filter(Age >= 15 & Age <= 80) %>%
  na.omit(cols='BodyweightKg')

  # Also having gender as 'F' and 'M' makes analysis hard/worse, let's turn it into factors.
powerliftingdataclean$Sex <- powerliftingdataclean$Sex %>%
  as.factor()

# Let's also turn Equipment to factors since we only have 4 uniques
powerliftingdataclean$Equipment <- powerliftingdataclean$Equipment %>%
  as.factor()

# Also there seem to be about 1100 values that are negative (weights), let's remove those
powerliftingdataclean <- powerliftingdataclean %>%
  filter( BodyweightKg >= 0 &
            BestSquatKg >= 0 &
            BestBenchKg >= 0 &
            BestDeadliftKg >= 0)

# Let's order the Equipment so our graphs look nice, Raw -> Wraps -> Single-ply -> Multi-ply
powerliftingdataequipordered <- powerliftingdataclean
powerliftingdataequipordered$Equipment <- factor(powerliftingdataequipordered$Equipment,
                            levels = c('Raw','Wraps','Single-ply','Multi-ply'),
                            ordered = TRUE)

Future Buttons

Basic Statistics

Here is some basic data to gain an insight into our dataset and what kind of powerlifters attend these meets

Firstname Lastname Email
John Doe john@example.com
Mary Moe mary@example.com
July Dooley july@example.com 

> test2 <- powerliftingdataclean %>%
+   group_by(Equipment) %>%
+   summarize(AvgAge = mean(Age),
+             Equipchoice = paste0(round(100*n()/105252,1), '%'),
+             AvgWeight = mean(BodyweightKg),
+             AvgTotal = mean(TotalKg),
+             AvgBench = mean(BestBenchKg),
+             AvgSquat = mean(BestSquatKg),
+             AvgDeadlift = mean(BestDeadliftKg))
> test2
# A tibble: 4 x 8
  Equipment  AvgAge Equipchoice AvgWeight AvgTotal AvgBench AvgSquat AvgDeadlift
                                        
1 Multi-ply    36.5 1.7%             98.6      711      188      281         242
2 Raw          29.3 65.5%            83.4      468      108      165         195
3 Single-ply   31.9 12.2%            84.5      572      143      216         214
4 Wraps        30.9 20.6%            92.0      548      129      201         218
		 

> powerliftingdataclean %>%
+   group_by(Sex) %>%
+   summarize(AvgAge = mean(Age),
+             Count = length(Sex),
+             AvgWeight = mean(BodyweightKg),
+             AvgTotal = mean(TotalKg),
+             AvgBench = mean(BestBenchKg),
+             AvgSquat = mean(BestSquatKg),
+             AvgDeadlift = mean(BestDeadliftKg))
# A tibble: 2 x 8
  Sex   AvgAge Count AvgWeight AvgTotal AvgBench AvgSquat AvgDeadlift
                              
1 F       30.9 33901      69.3      322     66.7      117         138
2 M       29.7 71351      93.3      586    143        211         233

Introductory Data Analysis

Weight vs Age of Powerlifters

Weight vs Age

Strength vs Age of Powerlifters

Strength vs Age

Effect of Equipment on different lifts

Weight vs Age Weight vs Age Weight vs Age

Distribution of lift %

y-axis represents the % of total lift composition, and x-axis represents the tier of powerlifter, 0 being the weakest and 1 being the strongest of lifters.

Weight vs Age Weight vs Age

© Alexis Drakopoulos - 2017

+44 7413535033 - alexis.drakopoulos95@gmail.com