Sir Plum Warner, selection, and how it’s changing

Unleashing the potential of Artificial Intelligence in national selection: A vision Plum Warner and Dusty Miller would have embraced!

It's challenging to find selectorial royalty; let's be honest, at all levels, piecing together cricket teams is a thankless pastime littered with casualties and told-you-so's.

First, I thought it prudent to look back to a time when selection committees were less critical and not as scrutinised as they are today. And a time when the thought of machine learning was associated with the advent of motor cars, not super-computers telling us who to play and why!

---

SpeakingNick is supported by your subscriptions. I thank you sincerely and if you are still a free subscriber, please, if you can, consider subscribing. Thank you.

---

Sir Pelham Francis ‘Plum’ Warner (1873–1963) was an English Test cricketer; he played in 15 Tests, captaining 10. Right-handed batting was his thing, and he was twice honoured as Wisden Cricketer of the Year in 1904 and 1921, perhaps more for his diplomacy than cricket.

The career highlight was his part in regaining the 1903–04 Ashes. After retiring as a player, Warner served as England's chairman of selectors and tour manager for several years; the controversial 1932-33 "Bodyline" series would have been particularly challenging. Another point to report is that Warner founded The Cricketer magazine—a personal favourite in my formative years.

Subjectively, I offer Plum Warner as the matriarch of national cricket selectors.

Moving to a more contemporary period, I second, Geoff 'Dusty’ Miller.

A mildly unorthodox figure with a singular approach, Dusty Miller had the potential to become as influential a voice in team selection as the aforementioned Plum Warner himself. As England's Chairman of Selectors from 2008 to 2013, Miller brought a refreshingly original perspective, routinely challenging conventions and defying expectations in his decision-making. Above anything, Miller actually watched county games, and knew a decent percentage of players. This change in thinking in itself was a revelation.

Cricket writer Colin Bateman placed Miller best, 

"Geoff Miller concedes that he probably enjoyed cricket too much. He did not take it as seriously as some, and when it became a rigorous, grim-faced business, he was happy to leave behind an eight-year Test career that was largely unfulfilled." 

Fellow Substacker Simon 'Yozzer' Hughes also had this to say about Dusty Miller, 

"He was the last remaining player who unfailingly visited the opposing team's dressing room after play to thank them for the game ... [and] the last man to field at slip with a whoopee cushion up his jumper." Come on, Yoz.

Miller's best performance as a player came during the 1978-79 Ashes when he took 23 wickets at 15.04 and was England's fifth-highest run-scorer—imagine that. Miller played county cricket for Derbyshire and Essex, captaining Derbyshire from 1979 to 1981.

So, what does all this have to do with artificial intelligence (AI)?

Well, artificial intelligence is the technology story of the year; it may prove to be the defining narrative of this decade and many more.

In a sentence, roughly 10-15 years ago, selecting cricket teams pre-analytics was entirely subjective and corruptible.

It followed a blinkered territorial approach in many parts of the cricket world.

Australia fought internally within the five and, latterly, six states. The notion was widespread that New South Welshmen enjoyed preferential grading regarding national selection. The New South Wales cap was a prelude to the 'Baggy Green. Tasmanians and players from Western Australia had to knock the proverbial door down.

The West Indies' geography fragments its cricket. A Jamaican president of the board was ostensibly good news for aspiring Jamaican cricketers. Then, a change at the board level would bring a different bias; I believe this was only sometimes conscious. Logistics made it difficult for committees to see everything. You could go on verbatim—Middlesex and Surrey are far more fashionable than Lancashire and Yorkshire. Auckland over Wellington. Karachi rather than Lahore. Etc.

I'm not telling you anything new when I say an extroverted personality isn't a common trait amongst cricket selectors. More a dour, uncompromising—and essential to the AI cause—binary approach is required. "We need someone to blame when it goes 'tits up', and if and when it works, we need you to drift to the shadows quietly—unlike the current Australian practice.

Been forwarded this newsletter? Sign up here, free and paid subs are available.

For the past month, I've been playing around with a private GPT—these AI agents run around in the background performing tasks that normal humans find either tedious, too complicated or, in many cases, all of the above! You can initiate the training with a single rudimentary prompt. From there, you can iterate and task up the bot with as much supplementary information as you like. I gathered team information from the last T20 World Cup for the six primary teams—India, Pakistan, England, Australia, New Zealand, and South Africa. They were my subjective selections! The data showed how team selection changed depending on injuries, player form, team form, opposition, ground dimensions, and team balance. For the first time, I used relational databases and queried this. All with the help of, you guessed right, AI.

You may think I'm losing my mind. I totally understand.

To avoid this let me share my initial prompt/description. I have iterated repeatedly on this and fed back to the model final team selections for the upcoming 2024 T20 World Cup.

Latterly, I have used the 2024 IPL as a data source to test the GPT.

The stats were basic to start, but the outputs improved as the GPT became more accepting of the direction I was pushing it in. Every second day, I asked the same question—this on IPL 2024—"Pick the best team possible from all IPL teams, with reasons for and against." I included individual players stats.

When I asked the GPT to pick India's T20 World Cup squad, it did a pretty good job.

Here is India's squad for the T20 World Cup:

Rohit Sharma (c), Hardik Pandya (vc), Yashasvi Jaiswal, Virat Kohli, Suryakumar Yadav, Rishabh Pant (wk), Sanju Samson (wk), Shivam Dube, Ravindra Jadeja, Axar Patel, Kuldeep Yadav, Yuzvendra Chahal, Arshdeep Singh, Jasprit Bumrah, Mohd. Siraj

Here is the Plum Miller (original naming—how bad!) Selector GPT.

Rohit Sharma (c), Hardik Pandya (vc), Yashasvi Jaiswal, Virat Kohli, Suryakumar Yadav, Rishabh Pant (wk), KL Rahul (wk), Shubman Gill, Ravindra Jadeja, Axar Patel, Kuldeep Yadav, Rinku Singh, Arshdeep Singh, Jasprit Bumrah, Mohd. Siraj

India opted for four spinners; Plum Miller selected three and included an extra batter, Rinku Singh. It picked KL Rahul instead of Sanju Samson and Shubman Gill over Shivam Dube—I believe this was the biggest mistake. I was surprised when it picked Rishabh Pant because of his lack of data over the last 12 months. *I did let Plum Miller know that MS Dhoni had retired from International cricket*

So, have I convinced you otherwise? Are messers Warner and Miller and the like irreplaceable? These all present interesting questions to me.

For now, I plan to continue watering the Plum Miller GPT—the great thing about this technology is they are emotionless and never tire of work; they keep returning for more! Unlike the players, you have to drop. 

More updates to follow.