Opinions of former champions and appointed selectors have their place, but for Australian cricket to truly prosper artificial intelligence must be embraced.
Sophisticated algorithms dominate our lives, they determine which shares are bought and sold for our retirement funds, how long we have to wait for a bus, train, or at a set of traffic lights and what we see on our social media feeds.
Yet for over 100 years and counting, Australian men’s cricket has relied upon middle-aged men of similar experience and background to decide who represents the national team, a job now clearly beyond such a group and perhaps any group.
Questions as to the value and relevance of an all-rounder in a Test team; the most desirable combination of left and right handed players in a batting line-up; the trade-off between a bowlers strike rate and runs per over are extremely complicated; they require the concurrent assessment of thousands of variables and are best suited to machines.
For today’s selectors, their job is complicated further with a requirement to compare performances in a range of different formats, competitions and conditions. How does one objectively compare a Joe Burns century in a home Test against an inexperienced Sri Lankan bowling line-up, to Matt Renshaw hundreds in county cricket or a Marcus Stoinis match winning 44 in the Big Bash?
In addition to answering tough questions on known issues, artificial intelligence provides the best opportunity to bring greater innovation to the game.
In 2018, Major League baseball’s Tampa Bay Rays pitching ranks were dismantled by injuries and trades, but off the back of a creative data-driven strategy they put forward a plan to better utilise their pitches and in doing so, defied critics to surge to a 90-win regular season.
Tampa Bay, noting how team’s score the most runs in the first innings, started their best pitcher in more games than any other team, which included a record run of nine consecutive scoreless starts, but importantly only kept their starting pitcher in the game for 1-2 innings, keeping him fresh allowing him to back up more frequently.
They called this new role the ‘opener’ and it has since been mimicked by several other major league teams.
In contrast to this approach, the Australian selectors were given seven months to bring together a viable batting line up post ‘sandpaper-gate’, they came up with their own version of the ‘opener’ strategy, promoting a batsman not used by his state in that position to lead off for the national team.
By installing Aaron Finch to open the innings in five critical Tests, the selectors seemingly were relying upon the unquantified good bloke theory last seen being used by Bob Simpson in the ’80s.
(Photo by Daniel Kalisz – CA/Cricket Australia/Getty Images)
The further advantage provided by artificial intelligence is that it never has to explain its decisions to a fanbase or at a post-match press conference, and thus doesn’t succumb to the pressure of making safe but stupid decisions.
In American football we see teams continue to struggle with this pressure, here rules require offensive teams to make ten yards forward progress within a given set of four plays known as a down. If a team cannot make ten yards the opposition receives the ball.
If ten yards hasn’t been completed prior to the fourth down, teams have three options with their final play: attempt a field goal, punt the ball giving up possession but gaining metres or the final, most risky option, which is to run or to pass the ball attempting to win back possession but knowing that if the required yards are not gained the opposition wins the ball back with no additional field position lost.
History shows that in the NFL teams take the third option, the risky option, too little and in the process cost themselves between 0.50 and 0.66 wins a season.
In short form cricket, especially T20, captains are often faced with similar types of dilemmas. A question often faced by the fielding team is whether to continue with a ‘part-time’ bowler, who to that point in the game, may have exceeded expectations in either wickets taken or runs conceded, or revert to the safer more conventional choice of a specialist bowler.
Just like the fourth down in the NFL, this decision requires the simultaneous weighing of multiple factors, in a T20 World Cup final, who would you want to make that type of decision, a captain or coach with the weight of a nation on their shoulders or a machine with perfect objective recall?
Diverse opinions on cricket are wonderful, and matters of selection are always going to be contentious in a sport where so few get to the main stage of representing their country, but if Australia cannot quickly embrace a future with artificial intelligence at the centre of selection and strategy in our national cricket teams, we will be destined for many more years of mediocrity.
Jeremy Wang
Guest
I agree with you jack, a whole team of terminators or matrix players would easily defeat any country
Spanner
Roar Rookie
Dreadful bias - each state should have a selector - Head and Wade not picked and have no one to push their barrow - ridiculous !
Jack Dyer
Roar Rookie
7 Western Australians in the ODI squad for India . . No bias there . . ??!! Bring in the bots already!
Jack Dyer
Roar Rookie
Somehow those driverless cars are getting about though aren’t they . On Cummings You use other factors - for example how every other bowler in the history of the game has gone the first time they played in England . . Everyone had a first some time . . What can we reasonably expect of a bowler in that circumstance, what does that mean for the bowlers we put around Cummings.
Matt H
Roar Guru
I believe the old saying is "S#$% in S#$% out". Any system is only as good as its assumptions. There is a place for metrics and algorithms to inform the selectors' decision making. At the moment it is too skewed towards gut feel and the "glory pick". This is the Greg Chappell influence. But you also need the human element. Sometimes you just have to have a feel for these things. Any AI or metric in the world would have dropped Steve Waugh well before he made his first century. And there is the experience factor. Pat Cummins has never bowled in a test match in England and I'm not sure he has played much County either. How can you reliably make a solely data-based assessment on that basis?
Rellum
Roar Guru
I hate to break it to you but A.I. right now is not going to solve metrics faster than the coaches and data guys do and not for a while yet. Your Tampa pitching example is all metrics. My point was to do with batting orders which have changed little over time. Scoring rates have changed but I think that is little to do with major individual innovation, rather gradual changes. Small steps have happened but nothing to really shake up how the games is played that have lasted the test of time. One day cricket and T20 cricket are the major changes and they I guess were in response the cricket lack of change. And to another point, if we are letting AI make like in game tactical choices then I suggest we stop playing the games. It is meant to be a contest between humans to see who is the best.
Paul
Roar Guru
Just a few additional thoughts Jack. I don't understand your second point and how this relates to cricket. The flaws in any computing system, especially one like you're suggesting, is the quality of the programming and the quality of the inputs. The programming to set up the interaction for all the variables would be seriously complex. The historical data which we currently don't have (because there's been no need to keep it), would take years to gather and input before anything statistically meaningful was produced. I don't doubt cricket AI will be included in the future of cricket Jack, but probably not in my life time. In the meanwhile, I guess we're stuck with a captain's knowledge and "gut feel".
Jack Dyer
Roar Rookie
Thanks James and agree with your points - always has to be a combination and in the early years of AI we have seen a few terrible cases of bias. In a more competitive cricket world we just have to look to the next frontier to again be world leaders . . Cricket Academy no longer going to cut it . .
JamesH
Roar Guru
Really interesting read, Jack. I'm not against looking at a more scientific approach to selection and match situations via the use of algorithms but just bear in mind that the use of an algorithm doesn't necessarily eliminate bias. Humans still have to put the algorithm together and determine the appropriate weightings to give to the various factors processed by the algorithm, which is always going to involve a degree of subjectivity. The effect of this would lessen over time, of course, as the algorithm builds on the actual outcomes of its previous results.
Jack Dyer
Roar Rookie
AI is a lot more than metrics . . its the reason we don't have stockbrokers, and soon won't drive cars . . . It's not because someone has a stats sheet. Question - if we always come back to the Top 6 bats why do we continually flirt with all-rounders . . when and how do all-rounders get used, is it better to use an all rounder, or rest a fast bowler, a robot will answer this question better than any human. No innovation in Test Cricket huh? What game are you watching, scoring rates, the role of the wicket keeper, defensive bowling from earlier in an innings, have all changes a lot and that is just in the last 20 years . .
Jack Dyer
Roar Rookie
Sorry JRR I am backing a robot over that team . .
Jack Dyer
Roar Rookie
You are going to have the same issue Maurice - everyone brings their biases , Langer has them, Warney, Punter the same . . they are not objective, read every article on selection on this site, they are all based off their history and biases, a robot can objectively recall all required information that is going to maximise our performance.
Jack Dyer
Roar Rookie
Cheers for taking the time to read the article Paul. Few points: (1) there would always have to be a balance between AI and human input. (2) you have to make the distinction between live 'in-play' scenarios and set piece or play. AI more suited to the former, and will beat human every time in the long run. (3) On some of your specific questions there are certainly situations where AI will be well in advance of humans - (i) How does AI decide how much a pitch has changed over the course of a 5 day Test - by automatically assessing the history of every first class match on the pitch, understanding the make up of batting & bowling sides in those matches - left handed bowlers, and there relative strength, weather inputs, etc, etc, a lot better than the old gut and feel - The same way a driver less car adjusts to changed driving conditions. (ii) Finally, how do algorithms tell a captain when a bowler is “on”, or whether, in a certain circumstance, the batting order should be changed - both examples where again perfect objective recall will beat a captain in the long run. Needless to say we won't go into the 'hot hand' is a myth debate but again historical performance of bowlers post wicket being taken, batsmen next in - current form, history s bowler, performance in different match situations, ground, weather, humidity, mode of that innings previous dismal, most common dismal type of batting line up, bowling performance against left or rig hetc, etc, all variable better computed by AI. (iii) Batting order to be changed - again this requires the weighting of lots and lots of variables and more suited to robot than human, to name a few - overs to go, target to chase/set, current form of batsmen, pitch today, pitch history, bowling line up, mode of that innings previous dismal, most common dismal type of batting line up, bowling performance against left or right, versus batsmen with high / low strike rate, versus batsmen with high 4/6 ratios, batsmen with high 1/2 run ratios, bowling most common dismissal type, batsmen preference for leg side, off-side, straight, ground size, wind direction and speed . . . etc, etc, i If I have to be my house on it I don't want Aaron Finch making those calls. . . how about you?
qwetzen
Roar Rookie
Am all for using algorithms/stats. It'd eliminate bias.
Rellum
Roar Guru
The selectors only really have a small pool of players to pick from. The real issue is the system to breeds the FC players, and that system is broken right now. We are not producing quality players, especially batsmen. The national setup already uses metrics with everything they do. We could show creativity and flexibility in ODI's and T20 games but Test cricket is an old game and most things have been tried and eventually everyone comes back to the set top six bats because it works.
Jay Rodda Rams
Roar Rookie
As long as Khawaja opens at the World Cup we will be competitive. It has to be Darcy Short, Usman Khawaja(vc), David Warner, Steve Smith, Shaun Marsh, Glen Maxwell, Tim Paine(c), Billy Stanlake, Pat Cummins, Ashton Agar, Mitch Starc, 12th man Jhye Richardson
Maurice W Moyle
Guest
Sack the dinosaur selectors and replace with recently retired players who know the players qualities and are up todate with the modern game.
Paul
Roar Guru
Tell me, how do algorithms account for pitch conditions, Jack? How does AI decide how much movement a bowler will get when it doesn't have all the variables like exactly how raised the seam is, how hard is the ball, etc? How does AI tell a captain how much impact rain will have on the condition of both the pitch and the ball, the bowlers footing in the run up, etc? How does AI decide how much a pitch has changed over the course of a 5 day Test? Finally, how do algorithms tell a captain when a bowler is "on", or whether, in a certain circumstance, the batting order should be changed? CA have been keeping all kinds of stats and numbers about players ever since the Buchanan days at least and I'm yet to see a noticeable trend in our ability to win games. I'm sure they have a place in cricket - somewhere - but the fact that the ball hits a piece of grass in a game that can last days, means AI can only provide some answers and at the moment, I'd suggest these are very few.