MLS Analysis

mls_analysis.R

# ANALYSIS OF MAJOR LEAGUE SOCCER 
# MOTIVATION: http://thelogcabin.wordpress.com/2010/08/16/goals-per-game-in-mls/
#             http://thelogcabin.wordpress.com/2010/08/23/a-rule-change-in-major-league-soccer/



# LOAD ALL REQUIRED LIBRARIES

library(XML);
library(reshape);
library(plyr);
library(ggplot2);


# SCRAPE THE DATA FROM WEB AND SAVE TO FILE

url.base = 'http://www.mlsoccer.com/stats' ;
years = 1996:2010;

# save_data grabs data for each year and extracts table with team statistics 

save_data = function(y){
               url  = paste(url.base,as.character(y),'reg',sep = '/');
	       tab  = readHTMLTable(url, header = F, stringsAsFactors = F);
	       pos  = max(grep("United", tab));
	       tab  = tab[[pos]]; tab$year   = y;
	       tab	          
	     }

# extract team statistics across years and save as data frame
	
team.list  = llply(years, save_data);
mls        = merge_recurse(team.list);
names(mls) =  c("Name", "GP", "G", "A", "SHT", "SOG", "PG", "PA", "FC", 
                "FS", "YC", "RC", "SO", "SV", "SV%",   "GAA","Year");
# convert team names to factor and other stats to numeric	
		
mls[,-1] = sapply(mls[,-1], as.numeric);
mls[,1]  = as.factor(mls[,1]);

# save data frame as local csv file

write.csv(mls, 'mls.csv', row.names = F);

# EXPLORATORY ANALYSIS THROUGH PLOTS

# Plot1: Heatmap of Goals per Game across Years

plot1 = ggplot(mls, aes(x = Year, y = Name)) +
	geom_tile(aes(fill = G/GP), colour = 'white') +
	scale_fill_gradient(low = 'white', high = 'red') +
	theme_bw() +
	labs(x = "", y = "") +
	scale_x_continuous(expand = c(0,0)) +
	scale_y_discrete(expand = c(0,0)) +
	opts(legend.position = 'none',
	     axis.ticks  = theme_blank());

plot2 = qplot(Year, G/GP, data = mls, geom = 'line', facets = ~Name);

# Plot3: Average Statistics across Years

mls.m = melt(mls, id = c("Name", "Year"), variable_name = "metric");
mls.1 = ddply(mls.m, .(Year, metric), summarize, avg_val = mean(value));

plot3 = ggplot(mls.1, aes(Year, avg_val)) + 
        geom_point() + 
        geom_smooth(method = 'lm', se = F) + 
        facet_wrap(~metric, scales = 'free_y', ncol = 5) + 
        scale_x_continuous(breaks = c(1996, 2003, 2010));