Six Nations: The stats superbrains fine-tuning England’s tactics
Where does the slender difference between winning and losing hide?
On the training pitch? On the tactical whiteboard? In your star player’s pre-match imagination as they visualise a key passage of play?
As Eddie Jones’ England run out against Wales at Twickenham on Saturday, it may be in a line of code running on a laptop 10 miles across town in Balham.
Gordon Hamilton-Fairley and James Tozer, 28 and 26 respectively, will be the men staring at the screen.
Both played rugby at university, but neither to a particularly high level. Hamilton-Fairley rates himself as “incredibly average”, Tozer trumps him, but only just, at “very average”.
Jones’ opinion of the pair however is considerably higher.
In his autobiography published in November, the Australian lauded their work as “amazing”.
Hamilton-Fairley and Tozer are employed to tease out tactical truths from the avalanche of data that accompanies every top-level Test match, picking out the key statistics amid the static.
Every week they feed data from the previous round of matches into a series of coding scripts and machine-learning models.
Out of the other end comes the success rate for various tactical choices, potential vulnerabilities for each team and the tactical tendencies they tend to fall back on.
These, along with six closely guarded metrics the pair have identified as key to Test success, go into two reports that land on Jones’ desk at Pennyhill Park shortly after the final whistle – one reviewing England’s performance on the weekend just gone and another assessing the opposition to come.
The algorithms that drive Netflix plotlines,Amazon suggestions,Liverpool’s Premier League title charge and new coaching orthodoxies in the NBA, have arrived in rugby.
And according to Jones, the “data has been worth every penny”.
“We believe we’re right at the cutting edge of what is being done with data science and statistics in rugby, but also that there is still a lot more we can do,” Tozer told BBC Sport.
Their work may inform the gameplan that England players are trying to impose, but Tozer and Hamilton-Fairley are unknown to Owen Farrell and his team-mates.
They work remotely and are never in camp. Their matchday experience is the same as other armchair fans around the country.
“Occasionally we get tickets to games, but mostly we’re sitting on the sofa, laptops in hand, waiting to see how the tactics play out,” Tozer tells BBC Sport of his matchday experience.
“As soon as the data comes in after the final whistle, we start working on the match review.”
While they do so, the pundits pour over their own stats – territory, possession, tackles, turnovers and the like – on the television screen in the background.
If those are the ABCs, Hamilton-Fairley and Tozer are dealing in advanced algebra.
Take England’s win over Ireland at Twickenham a fortnight ago.
Ireland enjoyed 61% of possession and 59% of territory. They made 235 metres with ball in hand, 21 metres more than the hosts, beating 17 defenders to England’s 10. They conceded seven penalties while England strayed into the supposedly fatal double figures, offering up 13.
But, on the pitch, Ireland were a distant second, fortunate to keep the losing margin to just 12 points.
“Rugby sometimes gives the illusion of being sophisticated in this respect, because of the number of stats that you might see displayed on TV,” explains Tozer.
“But when you ask most pundits or coaches which metrics determine the outcome of a match, they won’t be able to give you an accurate answer.
- Wilson and Watson return for England
- Wales’ Williams and Biggar fit for Twickenham
- Moriarty vows to get stuck in
- ‘Brutal’ England clash and Warburton on WhatsApp
“Generally, they cherry-pick numbers that support their perception of a game, and claim that they matter.”
Perhaps that very human tendency to tailor the facts was behind a slump in form during which England lost six of a run of seven matches into the summer of 2018.
What is certain is that Hamilton-Fairley and Tozer becoming more closely involved with match preparation coincided with a marked and immediate improvement in performances.
“The problem Eddie had was he and his coaches were being given a significant amount of charts and numbers, but could not answer simple questions like: “What are the most important metrics to look at? What explains why teams win or lose?” explains Hamilton-Fairley.
“Following the South Africa tour in 2018, after a run of poor results, we sat down with Eddie and explained how England’s playing style had changed, and why this was causing poor results.
“Eddie had always been very receptive to our outside perspective, but it was this meeting that I think finally pricked his attention.”
There had been plenty of background work behind that breakthrough pitch.
Hamilton-Fairley had done his first project for England way back in 2010, while on a gap year between school and university, assessing how much experience World Cup winning squads tend to contain.
He continued part-time for the Rugby Football Union (RFU) through university and while working in his first job, before a partnership with Tozer sprung from a disagreement.
Tozer, who studied English at Oxford before spending five years learning to code, use statistics and to implement machine learning, works as a data journalist at the Economist.
He writes about the numbers behind everything from air pollution in India to the rise and fall of different music formats.
And in October 2015, he wrote about rugby, specifically Chris Robshaw’s ill-starred decision to kick to the corner and chase victory, rather than take a kick that could have drawn England’s World Cup pool match against Wales.
Scraping together the sparse data publicly available, Tozer concluded Robshaw’s call was a reckless gamble.
By contrast, Hamilton-Fairley, with access to the RFU’s more in-depth in-house data, believed Robshaw played the percentages correctly.
However, he had read enough to recognise someone he could work with.
“I knew as soon as I met James that he could add another level of analytics through data science and machine learning that could really make a significant difference,” he remembers.
This weekend, if Owen Farrell faces a similarly tough call to the one that confronted Robshaw five years ago, he, Tozer and, most importantly, the models they have created, will be in agreement on the best way to beat Wales.
Source: Read Full Article