I asked AI to help me predict the winner of the Cricket World Cup

CRICKET teams hire the best minds to help them win matches, but what if they used artificial intelligence? Like, could a team of robot minds help us work out cricket?

We've heard from plenty of analysts, experts, pundits, former cricketers and your uncle on the couch about who is favourite to win the 50 over cricket World Cup in 2023. For instance, Mitchell Marsh thinks Australia will meet Pakistan in the final. But is their way to sift through all that nonsense and find some truth?

Can artificial intelligence foretell which team will lift that winner's trophy on 19 November, 2023 in Ahmedabad?

Cricket World Cup winner odds

Before we let AI predict the winner, lets see the current odds for the winning country of the Cricket World Cup. Odds are subject to change.

The best cricket bookmakers

Get ready for the Cricket World Cup with the best offers and odds from our top expert reviewed bookmakers for cricket:

Can ChatGPT predict the World Cup winner?

So I asked. Starting with ChatGPT, but because it ends in 2021 it was not going to give any great information. So I moved over to Perplexity. I started with the most important question, who is most likely to win the Cricket World Cup in 2023? 

The AI explains how it can't predict the future, and while it can not provide an opinion on this, it can list down all relevant information for the World Cup, which it did. Number of teams, format, host nation, prize money, so much pointless information. 

We took another stab at the same question, asking whether it could predict each team's percentage chance of winning the World Cup. What followed was another apology from AI, citing its inability to predict due to several variables such as team form, player injuries, and weather conditions. 

This back-and-forth continued for a bit. I almost gave up on the exercise itself, but AI folded first, and gave me the most likely winner.

Which team does AI believe the be the most likely winner?

It listed out India as odds-on favourites. Along with England and Australia they made up a trio of teams most likely to win the World Cup.

Furthermore, Sri Lanka were dark horses, Pakistan unpredictable, New Zealand the team to look out for, South Africa were termed as chokers, Bangladesh a surprise package, whereas the West Indies were described as a unit that could be a threat to other teams.

AI was not aware that the West Indies had failed to qualify for the marquee tournament and also listed Zimbabwe and Ireland as other teams of the World Cup.

AI and misinformation

It is staggering how AI actually went on to provide misinformation. When informed that the West Indies, Ireland and Zimbabwe had not qualified, it went on to re-list the teams that had qualified, once again mentioning the West Indies. Then upon correction, replacing the West Indies with Zimbabwe. A hattrick of mistakes by the supposed encyclopaedia of tomorrow. 

After this rather futile correction exercise, we asked "Did Netherlands not qualify?" to which AI responded by doubling down on its error, claiming that the Dutch had indeed not qualified. Man 4, Artificial Intelligence 0. 

We corrected the AI that Netherlands had qualified, after which it finally spat out the correct list of teams.

Getting odds with AI

I asked "Who will win the tournament according to experts and betting odds?", which finally resulted in some plausible information. India were singled as the favourites as per experts, with betting odds of 3.20, which were the most favourable. England, Australia and Pakistan were also listed as other teams which are favourites.

When prompted to provide the percentage chance of each team lifting the cup, AI mentioned the use of probability calculations to estimate the likelihood of each team winning. When given the green signal to do so, the bot, after a lot of frustrating back and forth between man and machine, went on to explain the science behind its calculations, giving a repeated disclaimer of betting odds being subject to change, and provided us with the following percentages:

• 1. India 28.6%
• 2. England 21.06%
• 3. Australia 14.29%
• 4. Pakistan 11.76%
• 5. New Zealand 11.11%
• 6. South Africa 9.09%
• 7. Afghanistan 2.78%
• 8. Bangladesh 2.44%
• 9. Sri Lanka 1.96%
• 10. Netherlands 92.31%

When we thought we were on to something, the Netherland's number arrives. It turned out that that was the odds of them making the next round. So perhaps if we are being friendly, we can guess the AI got confused with the qualifying event. 

What is more hilarious is that AI was quick to mention South Africa's barren run at the World Cup after presenting its percentage calculation, citing that their actual probability of winning the tournament may be lower than the estimated probability, due to their history of underperforming in major tournaments. That's a lot of sass for a robot. 

The revised percentage for the Netherlands, upon request, was 4.76%, still on the high side, and more than all three of Afghanistan, Bangladesh, and Sri Lanka, which goes on to show you the severe limitations of AI. Unless you think Netherlandis a legit chance to win one tournament for every 21 played. 

It was clear that we wouldn't get a great answer from AI about the winner. What could it do better? Maybe work out who would take the most wickets in the World Cup?

Who does AI believe will get the most wickets?

The results were underwhelming, as an incorrect list of top wicket takers at the World Cup Qualifiers was displayed initially, followed by some horrendous picks after being prompted again, which included:

• 1. Wanindu Hasaranga (injured)
• 2. Glenn McGrath (retired)
• 3. Lockie Ferguson
• 4. Muttiah Muralitharan (retired)
• 5. Jofra Archer (injured)

As it currently stands, Lockie Ferguson is the only person on the list with a chance of taking the most wickets, and he's in horrendous knick. According to AI, form is temporary, but class is permanent. Even when you're retired. 

We did ask the AI to remove the retired players, and it came back with this:

• 1. Wanindu Hasaranga (injured)
• 2. Pat Cummins
• 3. Jasprit Bumrah
• 4. Mitchell Starc
• 5. Lockie Ferguson

Not a mind-blowing performance by AI, but acceptable, which led us to ask a similar question, this time requesting for the top run scorers of the World Cup. The results were somewhat worse, with the initial cohort of batters looking in absolute shambles:

• 1. Sean Williams (did not qualify for World Cup)
• 2. Pathum Nissanka
• 3. Dimuth Karunaratne
• 4. Laura Wolvaardt (female South African cricketer)
• 5. Pat Cummins 

If there was a notion that machines were to take over the world someday, it was delayed the minute Laura Wolvaardt's name was suggested as a top run scorer at the men's World Cup. The incompetence is telling, so much so that it makes the Pat Cummins prediction look far less ridiculous than it actually is. AI also stated he was Australia's leading run scorer in the 2019 World Cup, so it lost the fact check battle as well.

A revised list replaced Laura and Pat with Rohan Mustafa of the UAE, and Aaron Jones of the USA, as AI re-dug its crystal ball grave by suggesting names of cricketers whose teams did not qualify. 

The final list was finally the class quintet of Virat Kohli, Babar Azam, Kane Williamson, Rohit Sharma, and Joe Root. Third time's the charm. 

So is AI useful for finding the Cricket World Cup winner?

AI was right, there is no way to predict the future. We set it up to fail by just asking. But it was pretty surprising how many simple bits of information it got wrong.

No one can predict the future, we can be pretty sure Netherlands wouldn't win one in 21 editions of the World Cup, Glenn McGrath will not take the most wickets, and sadly Laura Woolvardt can't save South Africa this World Cup. 

Remember to gamble responsibly

A responsible gambler is someone who:

• Establishes time and monetary limits before playing
• Only gambles with money they can afford to lose
• Never chase their losses
• Doesn’t gamble if they’re upset, angry, or depressed

For help with a gambling problem, call the National Gambling Helpline on 0808 8020 133 or go to www.gamstop.co.uk to be excluded from all UK-regulated gambling websites.