EDA on my New York Times Crossword data

nyt_xword.R

library(data.table)

# output from https://github.com/mattdodge/nyt-crossword-stats
#   python fetch_puzzle_stats.py -u "..." -p "..." -s 1993-11-21

DT = fread("data.csv")
# Sunday is a bit of its own thing, so put it last
DT[, day := factor(day, levels = c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun"))]
DT[, iso_week := format(date, "%G-%V")]

DT[, by = day, {
  # Don't get a min like '0' which is a clear outlier/bug. My min time is ~2 minutes.
  lower_bound = 0.5 * quantile(elapsed_seconds, 0.05)
  `:=`(
    rolling_median = frollapply(elapsed_seconds, 26L, median)
    rolling_min = frollapply(elapsed_seconds, 26L, \(x) min(x[x > lower_bound]))
  )
}]

# TIME SERIES OF SOLVE TIMES BY WEEKDAY

# Could mess around a bit more with the order of the days / centering of the grid,
#   but this gets close without much fuss.
par(mfrow = c(3, 3))
for (DAY in levels(DT$day)) {
  DT[day == DAY][
    # global lower bound, but day-specific upper bound
    elapsed_seconds < quantile(elapsed_seconds, 0.95) * 1.5 & elapsed_seconds > 120,
    matplot(
      date, cbind(elapsed_seconds, rolling_median, rolling_min)/60,
      type = "l", lty = 1L, las = 1L, lwd = c(0.5, 2L, 2L), col = c(1, 2, 4),
      xlab = "", ylab = "Solve Time (m)", main = DAY
    )
  ]
}

# DISTRIBUTION OF SOLVE TIMES BY WEEKDAY

par(mfrow = 1:2)
DT[, {
  x = boxplot(
    elapsed_seconds/60 ~ day,
    notch = TRUE, ylim = c(2, 30), las = 1L, col = 2:8,
    xlab = "", ylab = "Solve Time (m)", main = "Distribution by Day"
  )
  bounds <- x$stats[2:4, ]
  bounds_time <- sprintf("%02d:%02d", floor(bounds), round(60 * (bounds %% 1)))
  writeLines(sprintf(
    "%s: %s",
    levels(day),
    apply(matrix(bounds_time, nrow = 3L), 2L, paste, collapse = " / ")
  ))
  text(rep(1:7, each = 3L), bounds, bounds_time)
}]

# STACKED ROLLING MEDIANS
DT[, {
  data <- dcast(.SD, iso_week ~ day, value.var = "rolling_median")
  y <- as.matrix(data[, !"iso_week"]) / 60
  matplot(
    date[match(data$iso_week, iso_week)], y,
    lty = 1L, type = "l", lwd = 2L, col = 2:8, las = 1L,
    xlab = "", ylab = "Solve Time (m)", main = "Rolling Median by Day"
  )
}]