Cumulative count of names from two separate columns - r

Question

I have a data set in chronological order which I have imported to R using:
mydata <- read.csv(file="test.csv",stringsAsFactors=FALSE)
Two of the columns in the data set are 'winner' and loser'. Each row in the data is a tennis match.
What I am looking to do is to add two columns which give me a cumulative count of the total matches the player in the 'winner' column has played up to and including the match on that row. And the same count for the 'loser' in that row.
So for example it would look like this:
winner loser winner_matches loser_matches
tom andy 1 1
andy greg 2 1
greg tom 2 2
I hope that makes sense.
I have tried using the following code but can't get it to work across both columns:
ave(mydata$winner_name==mydata$winner_name, mydata$winner_name, FUN=cumsum)
So the data below is the first 10 rows of around 20,000.

Answer 1

1) base Define a function which counts matches up to the ith row for the indicated player and then apply it for the winner and loser matches separately. No packages are used:
count_matches <- function(i, player) {
with(DF[1:i, ], sum(winner == player | loser == player))
}
n <- nrow(DF)
transform(DF, winner_matches = mapply(count_matches, 1:n, winner),
loser_matches = mapply(count_matches, 1:n, loser))
giving:
winner loser winner_matches loser_matches
1 tom andy 1 1
2 andy greg 2 1
3 greg tom 2 2
2) sqldf A different solution can be obtained using sqldf upon realizing that this problem can be solved with a self-join on a complex condition like this:
library(sqldf)
sqldf("select a.winner,
a.loser,
sum(a.winner = b.winner or a.winner = b.loser) winner_matches,
sum(a.loser = b.winner or a.loser = b.loser) loser_matches
from DF a join DF b on a.rowid >= b.rowid
group by a.rowid")
giving:
winner loser winner_matches loser_matches
1 tom andy 1 1
2 andy greg 2 1
3 greg tom 2 2
Note: The input used, in reproducible form, is:
Lines <- "winner loser
tom andy
andy greg
greg tom"
DF <- read.table(text = Lines, header = TRUE, as.is = TRUE)

Answer 2

We can get number of times that each player won or lost by data.table package:
library(data.table)
setDT(dat)[, winner_matches_won := seq_len(.N), by=(winner)]
setDT(dat)[, loser_matches_lost := seq_len(.N), by=(loser)]
dat
# winner loser winner_matches_won loser_matches_lost
# 1: tom andy 1 1
# 2: andy greg 1 1
# 3: greg tom 1 1
# 4: greg tom 2 2
# 5: tom greg 2 2
Data:
dat <- structure(list(winner = structure(c(3L, 1L, 2L, 2L, 3L), .Label = c("andy",
"greg", "tom"), class = "factor"), loser = structure(c(1L, 2L,
3L, 3L, 2L), .Label = c("andy", "greg", "tom"), class = "factor")), .Names = c("winner",
"loser"), class = "data.frame", row.names = c(NA, -5L))

Answer 3

You're really close to getting ave to work. The cumsum function doesn't know how to handle text so I created a dummy column that's equal to 1 for each row. That gives cumsum something to count.
Here's a sample dataframe.
mydata <-
data.frame(
winner = c("tom", "andy", "greg", "tom", "gary"),
loser = c("andy", "greg", "tom", "gary", "tom"),
stringsAsFactors = FALSE
)
And here's the code to add the two new columns.
library(tidyverse)
mydata <- mutate(mydata, one = 1) # Add dummy column
# Use ave() to calculate both the wins and losses
mydata$winner_matches <- ave(x = mydata$one, mydata$winner, FUN = cumsum)
mydata$loser_matches <- ave(x = mydata$one, mydata$loser, FUN = cumsum)
mydata <- select(mydata, -one) # Remove dummy column