Problem Statement

Create an R markdown file that converts a structured text file into a .csv, containing the following:

      Player’s Name
      Player’s State
      Total Number of Points
      Player’s Pre-Rating
      Average Pre-Chess Rating of Opponents

For the first player, the line should read:

      Gary Hua, ON, 6.0, 1794, 1605

Download


Download the raw text file from Github.

# Download the tournament .txt file from Github
txtfile <- getURL("https://raw.githubusercontent.com/mmippolito/cuny/main/data607/project1/tournamentinfo.txt")

Extract


Extract each player (2 lines). The raw data will look like this:

    1 | GARY HUA                        |6.0  |W  39|W  21|W  18|W  14|W   7|D  12|D   4|
   ON | 15445895 / R: 1794   ->1817     |N:2  |W    |B    |W    |B    |W    |B    |W    |

# Create a regex to get each player line by line. The lines will look like this:
#
#    1 | GARY HUA                        |6.0  |W  39|W  21|W  18|W  14|W   7|D  12|D   4|
#   ON | 15445895 / R: 1794   ->1817     |N:2  |W    |B    |W    |B    |W    |B    |W    |
#
# Sometimes there could be a bye or a half-bye instead of a win/lose/draw, like this: |B   |
# And sometimes there is a provisional rating after a player's pre-performance rating, 
# e.g. 12531685 / R: 1291P12->1259P17
#
re <- " +(\\d+) \\| ([A-Z\\- ]+) +\\|([\\d\\.]+) +\\|(([A-Z] +(\\d+)?\\|){7}) *[\\r\\n]+"
re <- str_c(re, " +([A-Z]{2}) +\\| +\\d+ \\/ R: +(\\d+)[P ].+[\\r\\n]+")
matches <- str_match_all(string = txtfile, pattern = re)

Walk Matches


Walk through each match, which will be an array of values, like this:

      [,2] [,3]                              [,4]  [,5]                                         [,6]     [,7] [,8] [,9]  
 [1,] “1”  “GARY HUA                       ” “6.0” “W  39|W  21|W  18|W  14|W   7|D  12|D   4|” “D   4|” “4”  “ON” “1794”
 [2,] “2”  “DAKSHESH DARURI                ” “6.0” “W  63|W  58|L   4|W  17|W  16|W  20|W   7|” “W   7|” “7”  “MI” “1553”
 [3,] “3”  “ADITYA BAJAJ                   ” “6.0” “L   8|W  61|W  25|W  21|W  11|W  13|W  12|” “W  12|” “12” “MI” “1384”

Further parsing is needed on the “opponent” field. Each line will look like this:

W  39|W  21|W  18|W  14|W   7|D  12|D   4|

and this code will make it look like a vector:

[1] “39” “21” “18” “14” “7”  “12” “4” 

# Player IDs
ids <- as.integer(matches[[1]][,2])

# Strip whitespace from player name and change to title case
players <- str_to_title(str_trim(string = matches[[1]][,3]))

# Points
pts <- as.numeric(matches[[1]][,4])

# States
states <- matches[[1]][,8]

# Pre-performance rating
preperf <- as.integer(matches[[1]][,9])

# Further parse match results; each line will look like this:
# W  39|W  21|W  18|W  14|W   7|D  12|D   4|
tmp_opp <- str_match_all(string = matches[[1]][,5], pattern = "([A-Z]) +(\\d+)?\\|")

# Create an array of 7 lists of opponent IDs and 7 opponent results (Win/Lose/Draw/Bye/Half-bye).
# For the opp_result, choose the [i + 7]th element, 
# since we'll choose the second column of 7 (elements 8 to 14).
# For the opponent ids, choose the [i + 14]th element, 
# since we'll choose the third column of 7 (elements 15 to 21).
results <- list(7)
opponents <- list(7)
scores <- list(7)
for(i in 1:7) {
  results[[i]] <- (sapply(tmp_opp, "[[", i + 7))
  opponents[[i]] <- as.integer(sapply(tmp_opp, "[[", i + 14))
  scores[[i]] <- case_when(
    results[[i]] == "W" ~ 1,
    results[[i]] == "D" ~ 0.5,
    results[[i]] == "L" ~ 0,
    results[[i]] == "B" ~ 1,
    results[[i]] == "H" ~ 0.5,
    results[[i]] == "U" ~ 0,
    results[[i]] == "X" ~ 1,
    results[[i]] == "F" ~ 0
  )
}

Create Dataframe


Now that all the variables are populated, create a dataframe out of them.

# Create main dataframe from variables
df <- data.frame(id = ids, player = players, state = states, pts = pts, prerating = preperf, 
                 opp = opponents, opp_result = results, score = scores)

# Rename opponent columns (opponent's player id, result against that opponent, 
# and score against that opponent)
for(i in 1:7) {
  names(df)[i + 5] <- str_c("opp", i)
  names(df)[i + 12] <- str_c("result", i)
  names(df)[i + 19] <- str_c("score", i)
}

# Show the first few lines of the dataframe
df[1:10]

Calculate Ratings


With the dataframe created and populated with variables, now comes the real excitement. That’s right: calculating the mean of each player’s opponents’ pre-tournament rating.

# Create a slimmed-down data frame with only player ID and prerating
dfslim <- select(df, id, prerating)

# Make a copy of the original data frame, only selecting player ID and the 
# opponents he/she played against
dftmp <- df %>% select(id, starts_with("opp"))

# Join the slim data frame against each opponent to populate that player's pre-tournament rating
dftmp <- dftmp %>% merge(y = dfslim, by.x = "opp1", by.y = "id", all.x = TRUE) %>% rename(prerating_opp1 = prerating) %>%
  merge(y = dfslim, by.x = "opp2", by.y = "id", all.x = TRUE) %>% rename(prerating_opp2 = prerating) %>%
  merge(y = dfslim, by.x = "opp3", by.y = "id", all.x = TRUE) %>% rename(prerating_opp3 = prerating) %>%
  merge(y = dfslim, by.x = "opp4", by.y = "id", all.x = TRUE) %>% rename(prerating_opp4 = prerating) %>%
  merge(y = dfslim, by.x = "opp5", by.y = "id", all.x = TRUE) %>% rename(prerating_opp5 = prerating) %>%
  merge(y = dfslim, by.x = "opp6", by.y = "id", all.x = TRUE) %>% rename(prerating_opp6 = prerating) %>%
  merge(y = dfslim, by.x = "opp7", by.y = "id", all.x = TRUE) %>% rename(prerating_opp7 = prerating) %>%
  arrange(id) %>% select(id, starts_with("prerating_"))

# Compute pre-tournament average
dftmp <- dftmp %>% mutate(opp_prerating_avg = as.integer(rowMeans(subset(dftmp, select = 
  c(prerating_opp1, prerating_opp2, prerating_opp3, prerating_opp4, 
    prerating_opp5, prerating_opp6, prerating_opp7)), na.rm = TRUE))) %>%
  select(id, opp_prerating_avg)

# Join the new dataframe that includes the prerating average back to the original data frame
dfnew <- df %>% merge(y = dftmp, by = "id")

# Show the first few rows and columns of the new dataframe
dfnew %>% select(player, state, pts, prerating, opp_prerating_avg)

Export


Export the transformed data frame to CSV.

# Export to CSV
dfnew %>% select(player, state, pts, prerating, opp_prerating_avg) %>%
  write.csv("project1-out.csv", quote = FALSE, row.names = FALSE)

ELO


Calculate the player’s expected score \(E_1\) as follows:

\[\begin{aligned} E_1 &= 10^{r_1/400}\ /\ (10^{r_1/400} + 10^{r_2/400}) \\ \end{aligned}\]
where \(E_1\) is Player 1’s expected score, \(r_1\) is Player 1’s pretournament rating, and \(r_2\) is the average of his opponents’ pretournament ratings.

Then we’ll calculate each player’s actual score as follows:

\[\begin{aligned} W &= win &= 1\ pt \\ D &= draw &= 0.5\ pt \\ L &= loss &= 0\ pt \\ B &= full-point\ bye &= 1\ pt \\ H &= half-point\ bye &= 0.5\ pt \\ U &= unplayed\ (zero-point\ bye) &= 0\ pt \\ X &= win\ by\ forfeit &= 1\ pt \\ F &= loss\ by\ forfeit &= 0\ pt \\ \end{aligned}\] 


# Calculate expected score = 7 * expected score using avg pretournament ratings of opponents
dfnew <- dfnew %>% mutate(exp_score = 7 * (10^(prerating / 400)) / 
                            (10^(prerating / 400) + 10^(opp_prerating_avg / 400)))

# Calculate actual score based on actual matches - this should exactly match the "pts" column
dfnew <- dfnew %>% mutate(actual_score = score1 + score2 + score3 + 
                            score4 + score5 + score6 + score7)
dfnew %>% select(player, state, pts, exp_score, actual_score)

Analysis


Evaluate which players did the best in the tournament, comparing actual score against expected score.

# Calculate difference between actual and expected scores
dfnew <- dfnew %>% mutate(score_diff = (pts - exp_score))

# Display the top ten performers
top_performers <- dfnew %>% select(player, state, pts, exp_score, score_diff) %>%
  arrange(desc(score_diff))
top_performers[1:10,]


As shown, Aditya Bajaj of Michigan was the player who most exceeded expectations.

…