Op-Ed: AI Is Making It Cheaper To Build. Sports Needs To Make It Cheaper To Learn

Testing what works is an important part of developing successful fan engagement. Speeding the experimentation process can increase the odds of launching a hit

AI is rapidly reducing the cost of building digital experiences. In sports, the obvious opportunity is to create fan engagement faster and cheaper. But the more interesting consequence is what happens when being wrong gets cheaper, too.

I've spent much of my career making games, including at some of the largest game companies in the world. We used to talk about games as roughly a 10%-hit-rate business. It wasn't a formal statistic so much as an accepted reality: even great teams with significant resources got it wrong far more often than they got it right.

If decades of game development haven't made us reliably better at predicting hits, then, instead of trying to predict our way out of a hit-driven business, we can change the economics of finding them. Make it cheaper to be wrong, and you can take more shots on goal.

But taking advantage of that requires more than making experiences cheaper to build. It requires rethinking how the sports industry tests ideas, gets them in front of fans, and learns what works.

Mobile Changed the Cost of Being Wrong

Mobile games pushed this idea further than perhaps any other part of the games industry. Developers tested concepts before committing significant resources, soft-launched games into smaller markets, measured how real players behaved, killed weak ideas, and concentrated investment behind the ones that showed promise. Studios such as Supercell became known not just for the games they launched but also for the number they killed.

That created a portfolio of bets instead of concentrating risk in a handful of expensive ones. More important, every additional shot created another opportunity to learn. Mobile didn't make hits more predictable. It created a better system for finding them.

However, as user acquisition became more expensive and changes to mobile privacy made measurement harder, the economics of testing deteriorated. Soft launches became longer, not shorter. In 2023, Playtika went as far as temporarily suspending its new-game-development pipeline, saying that rising acquisition costs had made new titles difficult to scale profitably.

The lesson is easy to miss. Making the product cheaper or faster to build is valuable only if you can still afford to learn whether it works. Once distribution and measurement became more expensive, some of the advantage created by cheaper production disappeared. The important number wasn't simply the cost of production. It was the cost of learning.

Lower cost of learning more viable opportunities more shots on goalAI has the potential to push those economics much further. We may be able to build and test in days or weeks what once took months, giving us more shots and a much faster learning cycle.

Applying that model to sports introduces a different version of the problem. In mobile, the pressure came largely from acquisition and measurement. In sports, much of it sits in the operating model around the experience.

The sports industry wasn't built around rapid experimentation. Building is only one part of getting something in front of a fan. Rights, approvals, partners, budgets, and distribution sit between an idea and the moment you learn whether it works.

And, when the team building the experience is an external partner, the cost of learning is shaped by both sides. A partner can make experimentation dramatically faster and cheaper, only to have those gains disappear inside a process designed for much larger initiatives.

GreenPark Sports spent years running directly into that tension, offering a window into what happens when the cost of building falls faster than the cost of learning.

$55M and a Decade of Experiments in Sports

This year, JOA acquired GreenPark Sports' IP assets, bringing nearly a decade of work in interactive sports fandom into our company. Over that period, GreenPark raised more than $55 million and worked alongside the NBA, LaLiga, and MLS. What we inherited wasn't just technology. It was the accumulated learning from years of trying to figure out what sports fans actually wanted.

In 2019, GreenPark's idea of rapid experimentation was roughly 12 weeks with an external studio to produce rough playable prototypes. From there, between one and four concepts might be selected for another six months of development.

At the time, that was fast. But it also meant that every experiment carried significant weight. Twelve weeks of development, budget, partner involvement, and internal resources was effectively the minimum bet required to learn whether an idea had merit. Before there was much evidence that fans wanted something, the idea needed to look big enough to justify making that bet.

That shaped the kinds of ideas worth pursuing. One early GreenPark goal was to engage 1% of all fans, with ambitions measured in millions of daily active users. When the minimum bet is large, ideas need to look big before anyone has learned whether they are good. An experience that might deeply engage 20,000 fans around a particular club, player, or rivalry is difficult to justify if testing the idea requires the machinery of a product intended for millions.

Those assumptions didn't always survive contact with fans. GreenPark built experiences that looked compelling but behaved differently once fans encountered them. At one point, the realization was captured internally in four words: We had it inverted.

That is what experimentation is for. It isn't simply a mechanism for killing bad ideas. It exposes assumptions you didn't know were wrong.

GreenPark understood the problem. By 2021, Test Fast & Frequently was one of its company values. An assessment from the same period identified a

LINK:	https://www.sportsvideo.org/2026/10/20/ai-is-making-it-cheaper-to-buil...
	See more stories from svg

Europe Stories