2012 Olympics Swimming - 100m Butterfly Men Finals prediction Author: Matt Malin Inspired by mages’ blog with predictions for 100m running times, I’ve decided to perform some basic modelling (loess and linear modelling) on previous Olympic results for the 100m Butterfly Men’s medal winning results. Code setup

<span>library</span>(XML)
<span>library</span>(ggplot2)

swimming_path <- <span>"http://www.databasesports.com/olympics/sport/sportevent.htm?sp=SWI&enum=200"</span>

swimming_data <- <span>readHTMLTable</span>(
  <span>readLines</span>(swimming_path), 
  <span>which =</span> <span>3</span>, 
  <span>stringsAsFactors =</span> <span>FALSE</span>)

<span># due to some potential errors in passing header = TRUE:</span>
<span>names</span>(swimming_data) <- swimming_data[<span>1</span>, ]
swimming_data <- swimming_data[-<span>1</span>, ]

swimming_data[[<span>"Result"</span>]] <- <span>as.numeric</span>(swimming_data[[<span>"Result"</span>]])
swimming_data[[<span>"Year"</span>]]   <- <span>as.numeric</span>(swimming_data[[<span>"Year"</span>]])
swimming_data             <- <span>na.omit</span>(swimming_data)

loess_prediction <- function(
  <span>medal_type =</span> <span>"GOLD"</span>, 
  <span>prediction_year =</span> <span>2012</span>) 
{
  medal_type <- <span>toupper</span>(medal_type)
 
 swimming_loess <- <span>loess</span>(
    Result ~ Year, 
    <span>subset</span>(swimming_data, Medal == medal_type),
    <span>control =</span> <span>loess.control</span>(<span>surface =</span> <span>"direct"</span>))
  
  swimming_prediction <- <span>predict</span>(
    swimming_loess, 
    <span>data.frame</span>(<span>Year =</span> prediction_year), 
    <span>se =</span> <span>FALSE</span>)

  <span>return</span>(swimming_prediction)
}

log_lm_prediction <- function(
  <span>medal_type =</span> <span>"GOLD"</span>, 
  <span>prediction_year =</span> <span>2012</span>) 
{
  medal_type <- <span>toupper</span>(medal_type)
  swimming_log_lm <- <span>lm</span>(
    <span>log</span>(Result) ~ Year, 
    <span>subset</span>(swimming_data, Medal == medal_type))
  
  swimming_prediction <- <span>exp</span>(<span>predict</span>(
    swimming_log_lm, 
    <span>data.frame</span>(<span>Year =</span> prediction_year), 
    <span>se =</span> <span>FALSE</span>))

  <span>return</span>(swimming_prediction)
}

swimming_data <- <span>rbind</span>(
  <span>data.frame</span>(
    swimming_data[<span>c</span>(<span>"Year"</span>, <span>"Medal"</span>, <span>"Result"</span>)], 
    <span>type =</span> <span>"actual"</span>),
  <span>data.frame</span>(
    <span>Year =</span> <span>rep</span>(<span>2012</span>, <span>3</span>),
    <span>Medal =</span> <span>c</span>(<span>"GOLD"</span>, <span>"SILVER"</span>, <span>"BRONZE"</span>),
    <span>Result =</span> <span>c</span>(
      <span>loess_prediction</span>(<span>"gold"</span>), 
      <span>loess_prediction</span>(<span>"silver"</span>),
      <span>loess_prediction</span>(<span>"bronze"</span>)),
    <span>type =</span> <span>rep</span>(<span>"loess_prediction"</span>, <span>3</span>)))

medal_colours <- <span>c</span>(
  <span>GOLD   =</span> <span>rgb</span>(<span>201</span>, <span>137</span>, <span>16</span>, <span>maxColorValue =</span> <span>255</span>),
  <span>SILVER =</span> <span>rgb</span>(<span>168</span>, <span>168</span>, <span>168</span>, <span>maxColorValue =</span> <span>255</span>),
  <span>BRONZE =</span> <span>rgb</span>(<span>150</span>, <span>90</span>, <span>56</span>, <span>maxColorValue =</span> <span>255</span>))
        
swimming_plot <- <span>ggplot</span>(
  swimming_data,
  <span>aes</span>(
    <span>x =</span> Year, 
    <span>y =</span> Result, 
    <span>colour =</span> Medal, 
    <span>group =</span> Medal)) + 
  <span>scale_x_continuous</span>(<span>limits =</span> <span>c</span>(<span>1968</span>, <span>2012</span>)) +
  <span>geom_point</span>() + 
  <span>stat_smooth</span>(
    <span>aes</span>(<span>fill =</span> Medal), 
    <span>alpha =</span> <span>0.25</span>, 
    <span>data =</span> <span>subset</span>(swimming_data, <span>type =</span> <span>"actual"</span>), 
    <span>fullrange =</span> <span>FALSE</span>, 
    <span>method =</span> loess)
    
swimming_plot <- swimming_plot + 
  <span>scale_fill_manual</span>(<span>values =</span> medal_colours) + 
  <span>scale_colour_manual</span>(<span>values =</span> medal_colours) + <span>theme_bw</span>()

Predictions ... [Read more...]

[This article was first published on Actuarially (Matt Malin), and kindly contributed to R-bloggers]. (You can report issue about the content on this page here) Want to share [Read more...]

The Actuary Puzzle 508 - Square numbers Author: Matt Malin From the puzzle pages of The Actuary June 2012, I attempt to solve the following, making use of R: This square contains exactly 21 smaller squares. Each of these smaller squares has sides of integer length, with no two smaller squares having sides ... [Read more...]