Ranking items scalably with the Bradley-Terry model

Introducing the {BradleyTerryScalable} package


Ella Kaye


July 6, 2017

slides (2017) materials (2017)

slides (2022) materials (2022)



I am very excited to be introducing the package BradleyTerryScalable at useR!2017. The package is available on GitHub.

BradleyTerryScalable is an R package for fitting the Bradley-Terry model to pair-comparison data, to enable statistically principled ranking of a potentially large number of objects.

Given a number of items for which we have pair-comparison data, the Bradley-Terry model assigns a ‘strength’ parameter to each item. These can be used to rank the items. Moreover, they can be used to determine the probability that any given item will ‘beat’ any other given item when they are compared. Further details of the mathematical model, and the algorithms used to fit it, are available in the package vignette.

Update: 2022

I reworked this presentation for a job interview. The 2022 slides and materials are for that version. It’s a little shorter, at 10 minutes rather than 15 minutes. The main difference, however, is in the much-improved slidecraft and style!

Event details

Event: useR!2017

Date: July 6th, 2017

Time: 11:36 AM

Location: Brussels, Belgium

Slides (2017)

Keyboard shortcuts for slideshow (once you’ve clicked inside it):

  • Use and to navigate through the slides.

  • Use o for an overview of all slides.

  • Use h to see a list of other shortcuts

    • f to toggle full screen not working, but the rest are fine.

    • For full screen slides, go to slides then press f.

Slides (2022)



BibTeX citation:
  author = {Ella Kaye},
  title = {Ranking Items Scalably with the {Bradley-Terry} Model},
  date = {2017-07-06},
  url = {https://ellakaye.co.uk/talks/2017-07-06_introducing-BradleyTerryScalable},
  langid = {en}
For attribution, please cite this work as:
Ella Kaye. 2017. “Ranking Items Scalably with the Bradley-Terry Model.” July 6, 2017. https://ellakaye.co.uk/talks/2017-07-06_introducing-BradleyTerryScalable.