---
title: "Which pros win the Mirrorball \u2014 and who lifts their celebrity?"
subtitle: "Data with the Stars · An Artometrics project"
date: 2026-11-03
format:
  html:
    toc: true
    embed-resources: true
    theme: cosmo
  pdf:
    documentclass: article
    geometry: margin=1in
execute:
  echo: false
  warning: false
  message: false
---

**Finding.** Derek Hough leads with 6 Mirrorball wins and +1.34 peer lift; Julianne Hough has the highest win rate among pros with 3+ seasons.

```{r}
#| label: setup
#| include: false
source("R/_setup.R")
```

## Charts

```{r}
#| label: 02_pro_lift
#| fig-cap: "02_pro_lift"
#| echo: false
# Rebuild chart with ggplot2 from bundled CSVs / story data.csv
story_dir <- file.path(root, "public/downloads/stories", "02_pro_win_rates")
df <- read_csv(file.path(story_dir, "data.csv"), show_col_types = FALSE)
# Prefer a pre-defined plot expression file if present; else smart default
plot_r <- file.path(root, "reports/qmd/R/plots", "02_pro_lift.R")
if (file.exists(plot_r)) {
  source(plot_r, local = TRUE)
  # expects object `p`
} else {
  # Default: if typical named mean columns, bar them
  num_cols <- names(df)[sapply(df, is.numeric)]
  chr_cols <- names(df)[sapply(df, function(z) is.character(z) || is.factor(z))]
  ycol <- intersect(c("mean","wins","delta","boost","composite","accuracy","n","perfects","share","lift","pro_mean","resid","streak","guest_mean","field_mean","pts"), num_cols)
  ycol <- if (length(ycol)) ycol[[1]] else num_cols[[1]]
  xcol <- intersect(c("dance_style","pro","celebrity","judge","bucket","theme_key","hometown_state","checkpoint","label","song","professional","role","era","bin","metric"), chr_cols)
  xcol <- if (length(xcol)) xcol[[1]] else if (length(chr_cols)) chr_cols[[1]] else NULL
  if (!is.null(xcol)) {
    dd <- df
    if (nrow(dd) > 20) dd <- dd[1:20,]
    dd[[xcol]] <- as.character(dd[[xcol]])
    dd <- dd[!duplicated(dd[[xcol]]), , drop = FALSE]
    dd[[xcol]] <- factor(dd[[xcol]], levels = rev(unique(dd[[xcol]])))
    p <- ggplot(dd, aes(.data[[ycol]], .data[[xcol]], fill = .data[[ycol]])) +
      geom_col() + scale_fill_gradient(low = "#A8DADC", high = "#6B4CE6") +
      labs(title = "02_pro_lift", x = ycol, y = NULL) + theme_dwts() + theme(legend.position = "none")
  } else {
    p <- ggplot(df, aes(.data[[num_cols[[1]]]])) + geom_histogram(bins = 30, fill = "#6B4CE6", color = "white") +
      labs(title = "02_pro_lift") + theme_dwts()
  }
}
save_both(p, "02_pro_lift")
p
```

[PNG](charts/02_pro_lift.png){download="02_pro_lift.png"} · [SVG](charts/02_pro_lift.svg){download="02_pro_lift.svg"}


```{r}
#| label: 02_pro_scatter
#| fig-cap: "02_pro_scatter"
#| echo: false
# Rebuild chart with ggplot2 from bundled CSVs / story data.csv
story_dir <- file.path(root, "public/downloads/stories", "02_pro_win_rates")
df <- read_csv(file.path(story_dir, "data.csv"), show_col_types = FALSE)
# Prefer a pre-defined plot expression file if present; else smart default
plot_r <- file.path(root, "reports/qmd/R/plots", "02_pro_scatter.R")
if (file.exists(plot_r)) {
  source(plot_r, local = TRUE)
  # expects object `p`
} else {
  # Default: if typical named mean columns, bar them
  num_cols <- names(df)[sapply(df, is.numeric)]
  chr_cols <- names(df)[sapply(df, function(z) is.character(z) || is.factor(z))]
  ycol <- intersect(c("mean","wins","delta","boost","composite","accuracy","n","perfects","share","lift","pro_mean","resid","streak","guest_mean","field_mean","pts"), num_cols)
  ycol <- if (length(ycol)) ycol[[1]] else num_cols[[1]]
  xcol <- intersect(c("dance_style","pro","celebrity","judge","bucket","theme_key","hometown_state","checkpoint","label","song","professional","role","era","bin","metric"), chr_cols)
  xcol <- if (length(xcol)) xcol[[1]] else if (length(chr_cols)) chr_cols[[1]] else NULL
  if (!is.null(xcol)) {
    dd <- df
    if (nrow(dd) > 20) dd <- dd[1:20,]
    dd[[xcol]] <- as.character(dd[[xcol]])
    dd <- dd[!duplicated(dd[[xcol]]), , drop = FALSE]
    dd[[xcol]] <- factor(dd[[xcol]], levels = rev(unique(dd[[xcol]])))
    p <- ggplot(dd, aes(.data[[ycol]], .data[[xcol]], fill = .data[[ycol]])) +
      geom_col() + scale_fill_gradient(low = "#A8DADC", high = "#6B4CE6") +
      labs(title = "02_pro_scatter", x = ycol, y = NULL) + theme_dwts() + theme(legend.position = "none")
  } else {
    p <- ggplot(df, aes(.data[[num_cols[[1]]]])) + geom_histogram(bins = 30, fill = "#6B4CE6", color = "white") +
      labs(title = "02_pro_scatter") + theme_dwts()
  }
}
save_both(p, "02_pro_scatter")
p
```

[PNG](charts/02_pro_scatter.png){download="02_pro_scatter.png"} · [SVG](charts/02_pro_scatter.svg){download="02_pro_scatter.svg"}


```{r}
#| label: 02_pro_wins
#| fig-cap: "02_pro_wins"
#| echo: false
# Rebuild chart with ggplot2 from bundled CSVs / story data.csv
story_dir <- file.path(root, "public/downloads/stories", "02_pro_win_rates")
df <- read_csv(file.path(story_dir, "data.csv"), show_col_types = FALSE)
# Prefer a pre-defined plot expression file if present; else smart default
plot_r <- file.path(root, "reports/qmd/R/plots", "02_pro_wins.R")
if (file.exists(plot_r)) {
  source(plot_r, local = TRUE)
  # expects object `p`
} else {
  # Default: if typical named mean columns, bar them
  num_cols <- names(df)[sapply(df, is.numeric)]
  chr_cols <- names(df)[sapply(df, function(z) is.character(z) || is.factor(z))]
  ycol <- intersect(c("mean","wins","delta","boost","composite","accuracy","n","perfects","share","lift","pro_mean","resid","streak","guest_mean","field_mean","pts"), num_cols)
  ycol <- if (length(ycol)) ycol[[1]] else num_cols[[1]]
  xcol <- intersect(c("dance_style","pro","celebrity","judge","bucket","theme_key","hometown_state","checkpoint","label","song","professional","role","era","bin","metric"), chr_cols)
  xcol <- if (length(xcol)) xcol[[1]] else if (length(chr_cols)) chr_cols[[1]] else NULL
  if (!is.null(xcol)) {
    dd <- df
    if (nrow(dd) > 20) dd <- dd[1:20,]
    dd[[xcol]] <- as.character(dd[[xcol]])
    dd <- dd[!duplicated(dd[[xcol]]), , drop = FALSE]
    dd[[xcol]] <- factor(dd[[xcol]], levels = rev(unique(dd[[xcol]])))
    p <- ggplot(dd, aes(.data[[ycol]], .data[[xcol]], fill = .data[[ycol]])) +
      geom_col() + scale_fill_gradient(low = "#A8DADC", high = "#6B4CE6") +
      labs(title = "02_pro_wins", x = ycol, y = NULL) + theme_dwts() + theme(legend.position = "none")
  } else {
    p <- ggplot(df, aes(.data[[num_cols[[1]]]])) + geom_histogram(bins = 30, fill = "#6B4CE6", color = "white") +
      labs(title = "02_pro_wins") + theme_dwts()
  }
}
save_both(p, "02_pro_wins")
p
```

[PNG](charts/02_pro_wins.png){download="02_pro_wins.png"} · [SVG](charts/02_pro_wins.svg){download="02_pro_wins.svg"}


## Data

Analysis CSV: [data.csv](../../public/downloads/stories/02_pro_win_rates/data.csv){download="02_pro_win_rates-data.csv"}

Source tables: `performances.csv`, `scores.csv`, `contestants.csv`, `seasons.csv` under `/downloads/`.