Over 10 years ago I wrote a piece based on a paper by Brendon Brewer from the University of New South Wales that provided a way to model batter’s final scores in Test innings.

In that paper he assumes that, for most batters, each Test innings plays out over two phases: an initial phase, before the batsman is said to be "set" and during which he is more susceptible to dismissal, and a second and final phase during which his probability of dismissal slowly declines as each run is scored and asymptotes towards some fixed value.

Back in 2020 I wrote a piece here about the comparative performance of UK and Australian Chasers, and I would refer you back to that piece if you want to know more about the data set I’m using for this new article. This time, I’ll only be providing an update on the relative performances of the UK Chasers alone, using data from the first ever UK Chase show back on 29 June 2009, all the way through to 24 October 2025.

Back in late 2024 I wrote a blog post because I had recently started watching repeats of a British game show called Winning Combination that aired in 2020 and 2021 and in which nine contestants, numbered 1 to 9, answer questions in an attempt to become a part of a “final combination”.

That combination is simply the four numbers associated with the successful contestants and becomes the prize pool to be divided evenly amongst them if they complete a final challenge.

In the previous blog post, I mentioned in passing that the new UK Chase data source I’d kindly been given included occupation, but said no more about it there.

That was largely because I knew that occupations were sufficiently diversely described that nothing meaningful was likely to be obtained by analysing using raw occupations alone - we’d have maybe thousands of unique occupations. What I knew I needed was some way of clustering those occupations, or grouping them based on some agreed schema.

In the most recent blog post here, we analysed the performance of contestants on the UK Chase based on inferring their gender from their first names.

It turns out that the incredibly diligent website owner of OneQuestionShootout has the actual gender data for over 9,000 contestants, as well as their age and occupation. He’s been kind enough to give me access to that treasure trove of data, and today we’ll take our first look at it to investigate two main questions:

1. How does contestant performance and behaviour vary based on a contestant’s age and gender? We’ll look at performance based on the average Cash Builder amount accrued, the likelihood of getting home, the average amount contributed to the Prize Fund, and the average Contestant Winnings. We’ll also look at the propensity to take the Low, Middle, or High offer, which obviously has a big influence on how Cash Builder amounts are converted into team and individual winnings

2. What is the optimal mix of ages across the four people in a team?

Some time ago, I looked at the performance of Australian contestants based on the gender implied by their first names, but it turns our that I never got around to doing the same for UK contestants.

Today, we close that gap.

Firstly, a quick reminder that we are again using Derek Howard’s Gender by Name database to attach implied genders to names, and we are using the player-by-player data on OneQuestionShootout for our raw data.

Derek’s database attaches a probability of being male to a large number of first names, and we deem a contestant name to have been appropriately name-gendered if that probability is under 25% or over 75%. For the UK data that leaves only about 6% of the 9,260 contestants with an unknown name gender.

Back in 2020 I wrote a piece here about the comparative performance of UK and Australian Chasers, and I would refer you back to that piece if you want to know more about the data set I’m using for this new article. This time, I’ll only be providing an update on the relative performances of the UK Chasers alone, using data from the first ever UK Chase show back on 29 June 2009, all the way through to 24 October 2025.

The adequacy of any decision should never be assessed purely on its outcome, but instead on how well it reflected the information available at the time. In wagering, that means profitability is often a poor reflection of the ability of the wagerer, although it becomes a fairer measure the longer the time span of wagers we are reviewing.

But there’s another, perhaps better, metric that we can use, and that is the frequency with which the wagerer creates what called Closing Line Value (CLV). Put simply, a bet has positive CLV if, at the line or price at which the market closes, the bet has a positive expected return assuming that the bookmaker’s final line or price provides an accurate estimate of the team’s true outcome probabilities.

Over the last four blog posts we’ve been slowly creating an optimal set of strategies for playing the version of Shut The Box that we introduced in the first blog.

This is the fourth in a series of posts that investigates the creation of an optimal strategy for the game Shut the Box, and I’d recommend you read the first three blogs before continuing, starting with Part 1.

This is the third in a series of posts that investigates the creation of an optimal strategy for the game Shut the Box, and I’d recommend you read the first two blogs before continuing, starting with Part 1.

This is the second in a series of posts that investigates the creation of an optimal strategy for the game Shut the Box, the first of which I’d recommend you read before continuing.

I’ve long been fascinated about how computers can be used to determine optimal - or even just good - strategies for playing games with well-defined rules, and in this blog I’ll be working through the creation of such a strategy for a game sometimes known as Shut The Box.

Today we’ll be analysing the latest player-by-player data from the fabulous OneQuestionShootout website, and investigating a single issue: how well can we predict a contestant’s chances of getting back to the desk based solely on the Cash Builder amount that he or she recorded?

There are two fundamentally different ways of simulating the remainder of a sporting competition, which are loosely referred to as ‘cold’ and ‘hot’ simulations. The difference between them comes down to the assumptions made about the dynamics of the abilities of the teams involved. In cold simulations, teams’ assumed abilities are unrelated to the simulated results of earlier games, while in hot simulations they are related to earlier results.

(I’ve found surprisingly scant writing about this topic in academic journals or elsewhere. Please let me know if you come across anything.)

In an earlier blog, we built models to predict the Low and High Offers made to contestants in The Chase UK and found that we were able to explain roughly half the variability in the Offers made - a little more in the case of models built for the Low Offers by Seat, and generally a little less in the case of models build for the High Offers by Seat.

Since that blog was posted, I’ve had some excellent suggestions for things to try that might improve the models’ fit, and I’ve also had some thoughts of my own.

After you’ve watched The Chase for a while, you become familiar with the patterns of the contest and feel as though you can make reasonable guesses about the Low and High Offers that will be made to a contestant at the start of his or her head-to-head, based on how well they did in their Cash Builder, which seat they came from, and the overall state of the contest.

Darragh Ennis became the sixth UK Chaser when he appeared in an episode that aired on the 19th of November 2020 where he won chasing a target of 16 against a team of two finalists, offering five pushbacks and answering the last question with 7 seconds still on the clock.

That started a remarkable run of 14 consecutive wins, which was the best start of any UK Chaser to date.

In late 2021, I wrote a piece describing a series of models designed to estimate the progressive chances of a team’s success on an episode of The Chase Australia as that episodes unfolds.

For today’s blog we’re returning to analyse this Google document from James Spencer, which covers the entire history of Andrew O’Keefe’s reign as host from late 2015 to mid-2021, with the intention of understanding what, if anything the name gender composition of a team tells us about their likely performance.