Adventures with R - Cricket Analysis (Predicting performance of Players based on previous ODI performances)

This is a continuing series on R and work on Cricket Analysis

Previous Series

Basic Setup

Getting all players in a single file

Clustering players based on ODI performance

This blogpost is focused on trying to predict individual player's performance on Runs the player will score based on performance in previous matches the player has played

** This code took longer to set up. The initial work was testing out different regression techniques, I kept coming back to the basic logistic regression as the model performance was not that improved between the runs. I believe the data is thin at an individual player level and therefore most regression techniques are having an issue with respecting to predicting runs scored. The model itself is not that great, it is basically predicting runs the player will score based on the balls faced, the Venue Type (Home or Away) and the Opposition Type (Strong or Weak)

Files are available at Code

Full code is reproduced for reference

library(lubridate)
library(GGally)
library(ggplot2)
library(scatterplot3d)
library(sqldf)


setwd('C:/Training/R/CricketAnalysis/')

df <- read.csv("ODIData.csv")
names(df)[names(df) == 'VenueType.y'] <- 'VenueType'


Player1 <- sqldf("Select Player, Runs, Mins, BF, Fours, Sixes, SR, Pos, Inns, 
                 case when VenueType='Away' then 0 
                         when VenueType='Home' then 1
                         else 0 end as VenueType,
                         case 
                         when Opposition = 'v South Africa' then 1
                         when Opposition = 'v Australia' then 1
                         when Opposition = 'v New Zealand' then 1
                         when Opposition = 'v England' then 1
                         when Opposition = 'v Pakistan' then 1
                         else 0 end as OppositionType,
                         case when Result = 'Won' then 1 
                         else 0 end as ResultType,
                         case
                         when Dismissal = 'not out' then 1
                         when Dismissal = 'retired notout' then 1
                         else 0 end as DismissalType
                         
                         from df")

Player1$Boundaries <- (Player1$Fours + Player1$Sixes)
Player1$BoundariesPct <- ifelse(!Player1$Boundaries, 0, Player1$Boundaries / Player1$Runs)
Player1$SR <- ifelse(!Player1$BF, 0, (Player1$Runs * 100)/(Player1$BF))


# Players in database
PlayerTable <- read.csv("PlayerTable.csv")
player_results <- list()

for (i in 1:nrow(PlayerTable)) {
  
  player_results[[i]] <- data.frame(playerid = PlayerTable$PlayerID[i], stringsAsFactors = FALSE)
  
}

Player <- data.table::rbindlist(player_results, fill = TRUE)
Player <- unique(Player[,1])

rm(list= ls()[!(ls() %in% c('Player','df','Player1' ))])


## Creating loop variables for Balls faced, Opposition Type & Venue Type
BF <- data.frame(rep(1:200, each=1))
names(BF) <- c("BF")

OppType <- data.frame(rep(0:1, each=1))
names(OppType) <- c("OppType")

VenueType <- data.frame(rep(0:1, each=1))
names(VenueType) <- c("VenueType")

extraction_list <- list()
m <- 0

for (h in 1:nrow(Player)) {
  
  for (i in 1:nrow(BF)) {
    
    for (j in 0:nrow(OppType)) {
      
      for (k in 0:nrow(VenueType)) {
        
  
  m <- m+1
  
  tryCatch ({
  
  ds <- Player1[ which(Player1$Player==Player$playerid[h]), ]
  
  fit <- lm(Runs ~ BF + OppositionType + VenueType,
            data=ds)
  
  post <- predict(fit, data.frame(Player = Player$playerid[h],
                                  BF = BF$BF[i], 
                                  OppositionType=OppType$OppType[j], 
                                  VenueType=VenueType$VenueType[k])
                  )
  
  extraction_list[[m]] <- data.frame(Player = Player$playerid[h],
                                     Runs = post, 
                                     BF = BF$BF[i],
                                     OppositionType = OppType$OppType[j],
                                     VenueType = VenueType$VenueType[k],
                                     stringsAsFactors = FALSE)
  }, error = function(e) e)
  
      }
    }
  }
}

PlayerForecast <- data.table::rbindlist(extraction_list[!extraction_list=='NA'], fill = TRUE)



PlayerDS <- sqldf("Select distinct Player, PlayerName from df")

finaldf <- merge(PlayerForecast, PlayerDS, by="Player", all.x = TRUE)

write.csv(finaldf, "PlayerForecast.csv")