README File for the supplementary materials of the publication ``Predicting matches in international football tournaments with random forests’’


- Folder Data contains 3 different data files:
	a) odds.RData contains the odds and the resulting 3-way-probabilities from the bookmakers
	b) odds_payouts_sorted.RData contains similar information, but here rows are sorted equally to the original data set in wc.data.02.14.rda
	c) wc.data.02.14.rda contains the main data sets (match results, covariate information, etc.) 


- Folder Methods contains 6 files:
	a) gamboost_fun.R: implementation of the Gamboost method (for Poisson response) 
	b) gamboost_nb.R: implementation of the Gamboost method (for neg. binomial response; this method is only shortly mentioned in Section 4) 
	c) grp_lasso.R: implementation of the Group Lasso method
	d) help_funs.R: implementation of some additional functions (prediction of goals, prediction of probabilities, loss functions, etc.)
	e) RF_Result_ordfor.R: implementation of the RF Result (ordinalForest) method
	f) RF_Result_party.R: implementation of the RF Result (party) method


- 01_compare_methods.R is the main analysis file. It applies all methods repeatedly to the World Cup data and predicts the respective test data
 
- 02_evaluate_results.R evaluates the results produced by 01_compare_methods.R and yields the results presented in Tables 4-6

- compare_methods.RData contains results produced by 01_compare_methods.R (in case 01_compare_methods.R is not executed)

- RCode_Data_Schauberger_Groll.Rproj loads an R-project in RStudio and sets the path to the unzipped folder for convenience
