CMJ Visualization Markdown

Familiarization and Cleaning

1. Load appropriate packages and libraries

2. Import CMJ data

3. Familiarize myself data structure, formatting, ranges, etc in Excel through exploratory sorting/filtering

4. Remove outliers based on weight (-1), team (NA), and metrics of +/- 3 standard deviations

Data Analysis

1. Correlation analysis to evaluate similarity when selecting CMJ metrics, trying to avoid redundancy

# Reorganize data long to wide for correlation analysis
cmj_wide <- cmj_data_clean %>%
  group_by(player_id, test_date, metric) %>%
  summarize(mean_value = mean(value), .groups = 'drop') %>%
  pivot_wider(
    id_cols = c(player_id, test_date),
    names_from = metric,
    values_from = mean_value
  )

# Calculate correlations for metrics
cor_matrix <- round(cor(cmj_wide[,3:ncol(cmj_wide)], use = "complete.obs"), 2)

2. Define selected CMJ metrics to be used for analysis

# Define the selected metrics
selected_metrics <- c("Concentric Mean Power / BM", "Eccentric Mean Power / BM", 
                      "Force at Zero Velocity", "RSI-modified", "Vertical Velocity at Takeoff")

2. Calculate percentiles for Player 3185 vs Team and vs League Outfielders

# Calculate percentiles for Player 3185 vs Team
Team_comparison <- cmj_data_clean %>%
  filter(parent_team_name == "Team" | player_id == "3185") %>%
  group_by(metric) %>%
  filter(metric %in% selected_metrics) %>%
  summarize(
    player_value = mean(value[player_id == "3185"]),
    team_percentile = ecdf(value[parent_team_name == "Team"])(mean(value[player_id == "3185"])) * 100
  ) %>%
  ungroup()

# Calculate percentiles for Player 3185 vs Position 3 (excluding Team)
position_comparison <- cmj_data_clean %>%
  filter((primary_position == "3" & parent_team_name != "Team") | player_id == "3185") %>%
  group_by(metric) %>%
  filter(metric %in% selected_metrics) %>%
  summarize(
    player_value = mean(value[player_id == "3185"]),
    position_percentile = ecdf(value[primary_position == "3" & parent_team_name != "Team"])(mean(value[player_id == "3185"])) * 100
  ) %>%
  ungroup()}

Plot Creation

1. Correlation plot creation

# Create correlation plot
png("correlation_plot.png", width = 1500, height = 1300, res = 150)

# Reformat plot for easier readability
corrplot(cor_matrix, method = "color", type = "upper", order = "alphabet", tl.col = "black", 
         tl.srt = 45, addCoef.col = "black", number.cex = 0.9, tl.cex = 0.9, mar = c(0,0,2,0))
dev.off()

2. Formatting data for each plot creation

# For Team comparison
team_spider_data <- data.frame(
  matrix(c(100, 100, 100, 100, 100,
           0, 0, 0, 0, 0,
           50, 50, 50, 50, 50,
           team_comparison$team_percentile),
         nrow = 4, byrow = TRUE)
)
colnames(team_spider_data) <- selected_metrics
rownames(team_spider_data) <- c("max", "min", "team_avg", "player")

# For Position 3 comparison
position_spider_data <- data.frame(
  matrix(c(100, 100, 100, 100, 100,
           0, 0, 0, 0, 0,
           50, 50, 50, 50, 50,
           position_comparison$position_percentile),
         nrow = 4, byrow = TRUE)
)
colnames(position_spider_data) <- selected_metrics
rownames(position_spider_data) <- c("max", "min", "position_avg", "player")

3. Creating each plot

# vs Team
png("player_vs_team_spider.png", width = 1000, height = 900, res = 100)
layout(matrix(c(1,2), nrow=2, ncol=1), heights=c(4,1))

# Spider plot
par(mar=c(3, 4, 2, 4))
radarchart(team_spider_data,
           pcol = c("blue", "red"),        
           pfcol = c(rgb(0, 0, 1, 0.2),   
                     rgb(1, 0, 0, 0.2)),   
           plwd = 2,
           cglcol = "grey",
           cglty = 1,
           axislabcol = "grey",
           caxislabels = seq(0, 100, 25),
           title = "Player 3185 vs Team (Percentiles)")
legend("topright", 
       legend = c("Team Average", "Player 3185"), 
       col = c("blue", "red"), 
       lwd = 2,
       pch = 20,
       pt.cex = 2,
       bty = "n")

# Percentiles text
par(mar=c(0,0,0,0))
plot.new()
text(0.5, 0.9, "Player 3185 Percentiles:", font=2)
for(i in 1:length(selected_metrics)) {
  text(0.5, 0.8-i*0.15,
       paste(selected_metrics[i], ": ", 
             round(team_comparison$team_percentile[i], 1), "%"))
}
dev.off()

# vs Position 3
png("player_vs_position3_spider.png", width = 1000, height = 900, res = 100)
layout(matrix(c(1,2), nrow=2, ncol=1), heights=c(4,1))

# Spider plot
par(mar=c(3, 4, 2, 4))
radarchart(position_spider_data,
           pcol = c("green", "red"),       
           pfcol = c(rgb(0, 1, 0, 0.2),   
                     rgb(1, 0, 0, 0.2)),   
           plwd = 2,
           cglcol = "grey",
           cglty = 1,
           axislabcol = "grey",
           caxislabels = seq(0, 100, 25),
           title = "Player 3185 vs Position 3 Players (Percentiles)")
legend("topright", 
       legend = c("Position 3 Average", "Player 3185"), 
       col = c("green", "red"), 
       lwd = 2,
       pch = 20,
       pt.cex = 2,
       bty = "n")

# Percentiles text
par(mar=c(0,0,0,0))
plot.new()
text(0.5, 0.9, "Player 3185 Percentiles:", font=2)
for(i in 1:length(selected_metrics)) {
  text(0.5, 0.8-i*0.15,
       paste(selected_metrics[i], ": ", 
             round(position_comparison$position_percentile[i], 1), "%"))
}
dev.off()

Visualizations

1. Correlation Matrix of CMJ Metrics
   

   Correlation Matrix of CMJ Metrics

2. Player 3185 Percentiles vs Team
   

   Player 3185 Percentiles vs Team

3. Player 3185 Percentiles vs League Outfield Average
   

   Player 3185 Percentiles vs League Outfield Average

Key Considerations

– Selected force plate metrics:

• Vertical Velocity at Takeoff: proxy for jump height, simple performance metric

• RSI-modified: ratio of jump effectiveness, output vs input (Time to Takeoff)

• Force at Zero Velocity: strength number, highest force created during jump

• Concentric Mean Power / BM: relative to BM for inter-athlete comparison, concentric number to compare to eccentric

• Eccentric Mean Power / BM: relative to BM for inter-athlete comparison, eccentric number to compare to concentric

– Logic and Thought Process

• With Excel’s ease of use for data viewing and sorting, I looked at the minimums and maximums of each column, finding -1 for some Weight entries and NA for Team entries
• As all CMJ metrics are “bounded” with relative physiological minimum and maximum thresholds, but with plenty of inter-athletes variation, I chose 3 standard deviations for outlier detection and removal as opposed to a slightly more stringent method like IQR
• A percentile would be an easier ranking number (as opposed to a Z-score, for example) for an athlete to understand

– Findings and Practical Applications

• My synopsis:
  • Player 3185 (randomly selected) is very strong in the +96th percentile for Force at Zero Velocity
    • I’m assuming Player 3185 is bouncy and forceful in shallow joints angles. A below average Vertical Velocity at Takeoff plus an above average RSI-modified means Time to Takeoff must be very quick
      • This is supported with above average Concentric Mean Power / BM and below average Eccentric Mean Power / BM
  • Player 3185 could benefit from some complementary full-range of motion strength training and deeper joint angle plyometrics

– Limitations and Future Considerations

• Force at Zero Velocity had the most outliers removed, I would look more into that metric’s reliability and see if there’s potentially a more consistent metric for strength
• Dive into intra- and inter-athlete reliability per metric
• Evaluate athlete’s spider graph in the first vs last month of training, see if any significant improvements have been made
• Find correlations of Team to League Average to potentially avoid redundancy. However, since the Team is obviously the best in the league at drafting and developing, I attempted to give the athlete some reassurance by also including league averages