A lot can happen over a walk in New York City's iconic Central Park. On a serene summer morning, for a few fleeting moments, Central Park feels wonderfully detached from the city it anchors. Perhaps that is why the park has long occupied a special place in New York's imagination, serving as the backdrop for countless epiphanies and life-changing conversations. The park has an uncanny way of making improbable ideas seem possible.

A couple of years ago, on one such June morning on the eve of the memorable India-Pakistan T20 World Cup clash in New York, two lifelong cricket tragics set off on what appeared to be an ordinary stroll. In reality, it may well prove to be one of the most consequential walks in cricket's technological evolution.

Those "cricket tragics" were anything but ordinary. One of the men was Anand Rajaraman, Silicon Valley entrepreneur, venture capitalist, co-founder of Rocketship VC and co-owner of the San Francisco Unicorns. The other was Vishal Misra, Vice Dean of Computing and AI at Columbia University, pioneering computer scientist and one of the members of the founding team of Cricinfo in the 1990s.

By the time the walk ended, the seeds had been planted for what would eventually become one of cricket's most ambitious artificial intelligence projects: SFU AI.

"Getting a chance to get involved with the operations of a professional team seemed very intriguing and exciting to me," Misra recalled.

Misra was offered the opportunity not only to become a minority owner in the Unicorns but also to lead their Modelling and Data (MAD) team, the initiative that would ultimately build the SFU AI.

"Anand and Venky (Hariharan) are computer science PhDs. They are venture capitalists. Their VC fund is actually completely data driven, they invest based on looking at data, and so they wanted to run this team based on data. That was the foremost idea behind getting a professional cricket team," Misra said.

Misra, who has also served Dream11 in an advisory capacity, had previously authored a research paper on a predictive technique that today forms the nucleus of SFU AI. Rajaraman had already assembled the majority of the MAD team, a group of cricket fanatics who were all Stanford PhDs working in big tech in Silicon Valley. Misra joined in to provide vision and direction. The team quickly programmed the methodology outlined in that paper into a workable tool. The results, according to those involved, were startling.

To understand why SFU AI's creators believe they may have built something fundamentally different, it is worth revisiting one of the most dramatic contests in recent cricketing history.

Rajaraman and Misra were both present at the India-Pakistan thriller in New York during the 2024 T20 World Cup. For all practical purposes, India appeared out of the contest at the halfway stage. "I was at the ground and it didn't feel like India were out of the contest," said Misra, one of the few among the 25,000 spectators and billions watching worldwide who felt the game remained evenly poised.

As it turned out, SFU AI agreed with him. Popular win predictors such as WinWiz and Cricinfo showed Pakistan having more than 95% chance of victory during the chase. SFU AI, however, viewed the contest very differently as it had India and Pakistan neck-to-neck during the chase barring a couple of overs when Fakhar Zaman threatened momentarily.

Even after Mohammad Rizwan's dismissal, with Pakistan requiring roughly 40 runs from six overs and still possessing six wickets in hand, with hitters such as Naseem Shah still to come at No. 9, SFU AI believed India held the upper hand. Most conventional models, meanwhile, continued to view Pakistan as overwhelming favourites.

The difference lay in Misra's philosophy. "That technique, what it loosely does is it looks at previous games that proceeded similarly. So, we have a historical database of all the games that have happened, and we map and create a sort of simulated version of the current game based on past games. Create a digital twin basically in AI parlance," Misra explained.

The concept of the digital twin lies at the heart of modern AI-driven predictive analysis. In cricketing terms, a digital twin is a data-driven virtual alter ego of a player, team or match that can be used to predict, simulate and optimize decision-making. Questions such as: How is Virat Kohli likely to perform if the opposition opens with a left-arm quick bowling around the wicket? Or should Pat Cummins bowl the 17th over or save himself for the 19th? can all theoretically be answered through such simulations.

The win-loss predictor may be the most visible manifestation of SFU AI, but it is merely the tip of the iceberg. Every probability generated by the model is, in reality, the cumulative outcome of thousands of microscopic calculations and simulations. Those capabilities can broadly be categorized into three areas: draft strategy, pre-game strategy and in-game strategy.

Perhaps SFU AI's most groundbreaking work lies in player acquisition. The platform can identify deficiencies within a squad and recommend precisely the type of player required to address those shortcomings. For instance, it can tell a franchise that what it lacks is a left-handed top-order batter or a death-overs specialist.

According to Misra, there is currently no cricket tool capable of comprehensively identifying squad deficiencies, evaluating available talent pools and quantifying the impact of potential acquisitions. The system can identify undrafted players, free agents or transfer targets and project how much a particular acquisition could improve a team's win percentage.

In what is believed to be a first in cricket analytics, SFU AI can also translate performances across competitions. A player's performances in domestic cricket can be projected onto leagues such as the IPL, Major League Cricket or even international cricket.

The system accounts for variables such as the quality of opposition, playing conditions and the presence of elite players in order to estimate how a player's numbers might translate at a higher level.

It also functions as a dynamic draft assistant. If a team identifies a left-handed middle-order batter as its preferred target and another franchise selects him first, SFU AI immediately recalibrates and produces the next-best option. Simultaneously, it analyses the alternatives available to rival franchises and even factors in salary-cap constraints.

Player identification extends well beyond conventional metrics such as averages and strike rates. One of SFU AI's proprietary metrics is Average Delta Win Probability, particularly for batters chasing targets where win-probability swings are most measurable.

The "delta" represents the difference between a player's peak win probability. That is the difference between the highest point of win expectancy reached while he was at the crease and the minimum win probability experienced during his innings. Put simply, it measures how effectively a player drags his side out of trouble and places it in a winning position before departing.

By that measure, Vaibhav Sooryavanshi emerged as an extraordinary outlier during IPL 2026. According to SFU AI's analysis, Sooryavanshi had an average delta-win percentage of 22 percent per game. Among players who featured in at least ten games, the second-best was Prabhsimran Singh at 11 percent.

What could prove even more transformative is SFU AI's long-term ambition to tackle something as dynamic and uncertain as player auctions. "That's a roadmap item. We have all of ingredients ready, we need to put them together," Misra said with child-like enthusiasm.

SFU AI has already transformed the way coaches prepare for games, particularly through the introduction of Cricket Lens, a natural language interface built on top of SFU AI. Think of SFU AI as the engine and Cricket Lens as the conversational layer, much like ChatGPT. Coaches can directly interrogate the system in plain English. They can ask what strategies should be devised against specific opponents, generate graphics at the click of a button or create visualisations ranging from wagon wheels and pitch maps to comparative performance charts.

They can even generate bespoke video playlists through simple prompts. A coach might type: "Show me every time Rachin Ravindra was beaten off the back foot by a left-arm pacer" or "Show me every inswinger of Mitchell Starc that resulted in either a bowled or LBW dismissal against right-handers." Within seconds, the system can produce precisely those clips, potentially revealing subtle clues such as whether Starc delivered those balls from close to the stumps or from wide of the crease.

Field placements provide another compelling example. Ahead of IPL 2026, SFU AI identified several unconventional catching positions for some of the world's best batters. For Rohit Sharma, it highlighted short fine leg. For Nicholas Pooran and Hardik Pandya, it identified short third man. For Shubman Gill, it pointed towards short cover.

Each of those positions was identified before the season began and subsequently validated by actual dismissals during IPL 2026. The implications may be game changing. Instead of relying on standard, template-based field settings, teams may increasingly deploy mathematically optimized traps specifically designed to dismiss individual batters.

The platform's in-game capabilities are equally sophisticated. SFU AI can recommend the optimal bowler for the next over, identify the most suitable batter to send in next and advise whether a side should adopt an aggressive or conservative approach over a given phase.

It can determine what constitutes an acceptable return over the next five overs while balancing wicket preservation and scoring rate. It can even suggest who should bowl the penultimate over of an innings based on opposition match-ups and prevailing conditions.

"We want to automate as many of these things as possible, and have AI do a lot of the planning for us. Ultimately, we want the AI to be the co-pilot of a coach, in the sense the AI will keep telling you, okay, do this, do that, and coach can decide yes or no, but the AI will give sort of constant suggestions and options to the coach in real time," Misra said.

Yet, for all its sophistication, SFU AI still confronts significant challenges. The availability of comprehensive data remains the biggest limitation. Predicting who should bowl the next over, for instance, may require accounting for shorter boundary dimensions, wind direction, dew, pitch deterioration and a host of other contextual variables.

While factors such as wind, dew and boundary asymmetry can be incorporated into models, emotional intelligence remains far more elusive. "Someone is going through a personal problem, like their mother is very sick back in Pakistan or Australia or if someone's career is on the line. Data doesn't know that," Misra said.

Similarly, if a player has picked up a niggle in the previous game, is battling fatigue, or is dealing with dressing-room issues, those variables may not yet form part of the modelling process.

Yet, those limitations may not remain limitations forever. One of SFU AI's greatest advantages today is proximity. Misra no longer operates from the confines of his study room. As part of the San Francisco Unicorns support staff, Misra now has access to some of the game's intangibles. He has a front row seat to dugout discussions, the opportunity to better understand the emotional quotient of players and, perhaps most importantly, direct insight into the minds of elite cricketers in real time as a game unfolds. Few minds in modern cricket are as coveted in that regard as Ravichandran Ashwin's.

Following an extended interaction with the veteran off-spinner during one such game for the Unicorns, Misra took to social media to express his amazement.

"His insights are incredible. Spending three hours in the dugout with Ashwin was pound for pound the most intense learning experience for me - how the mind of a top-tier professional cricketer works and what kind of data and insights he is looking for!"

For SFU AI, there can scarcely be a richer source of intelligence. After all, Ashwin is now widely regarded as one of cricket's foremost analytical minds. Misra, meanwhile, is no ordinary computer scientist. Having helped usher cricket into the internet age, he now finds himself, alongside the might of SFU AI and Ashwin, trying to shepherd it into the age of artificial intelligence.