Skip to content

Latest commit

 

History

History
69 lines (46 loc) · 2.56 KB

index.md

File metadata and controls

69 lines (46 loc) · 2.56 KB
Raw

title : Race Horse Prediction subtitle : Using linear regression to predict horse race outcomes author : Ian McCarthy job : SAP Consultant framework : io2012 # {io2012, html5slides, shower, dzslides, ...} highlighter : highlight.js # {highlight.js, prettify, highlight} hitheme : tomorrow # widgets : [] # {mathjax, quiz, bootstrap} mode : selfcontained # {standalone, draft} knit : slidify::knit2slides ---.class #id

Horse racing prediction

I would like to take this opportunity to demonstrate, using Slidify, one simple model for making race predictions, which is an area rich in data!

The method used

Successful horse racing betters are able to 'read the form' of diferent horses and make predictions on how they will run against each another.

I have taken a more automated approach that uses a simple linear regression fitted to a data set of all 2013 Australian horse racing events.

It makes a prediction of how long it will take each horse to run a future race using its last race performance and the characteristics of the future race . Comparing the times of all horses in the race allows their finishing position to be predicted.

--- .class #id

The model's input variables

The input variables (delta of last race and future race) used in the linear regression model are as follows:

  • Race distance - of course the farther the race the longer it takes to run.
  • The Going - how hard (faster) or soft (slower) a track is.
  • Weight - each horse carries a jockey and often an additional weight handicap

This is the general form of the R code for fitting the model and then predicting a horse's finish time:

prediction_model <- lm(finish_time ~ distance + weight + going, data = aust_races_2013)

predicted_time <- (prediction_model, data = horses_last_race)

The outcome of this is not shown.

--- .class #id

The effect of distance on average winning times

dist <- seq(1000, 2000, by=100)
time <- 0.070*dist
plot(dist, time, main="Average winning race times by distance", 
     xlab="Race Distance (m)", ylab="Time (s)", pch=19)
abline(lm(time~dist), col="red")

plot of chunk unnamed-chunk-2

--- .class #id

The web app

I have created a web app based on the described model which can be viewed here: https://ianmccarthy.shinyapps.io/HorseRace/ (it may be necessary to zoom out to see all the input boxes).

The web app can make predictions for up to 5 horses, the race prediction changes as data is changed.

I hope you enjoy the app. Thank you.