AI, Gambling, and LeBron vs MJ
We'll soon be able to bet on an AI-generated 1-on-1 game between LeBron and Jordan
Recently I went to a gaming conference and was mesmerized by this one company, Inspired Entertainment.
Inspired creates simulated NFL games for people to bet on. Check out an example (skip to 0:30).
Isn’t that mind-blowingly realistic?
Inspired uses a RNG (random number generator) system to determine the outcome of each game; CGI to create the footage; and a production layer to make it feel like live sports on TV. They have official deals with the NFL, NBA and NHL, and they also make non-licensed products like virtual horse racing.
In 2024, Inspired drove $45 million in revenue from this business segment, called Virtual Sports.
A few weeks after this gaming conference, I was mesmerized again while scrolling X.
Somebody used Seedance – a new AI video model that’s been blowing minds lately – to generate this video of NBA legends and current stars playing 1-on-1.
It’s not perfect. You can tell it’s AI. But again, isn’t that mind-blowingly realistic?
The Inspired Entertainment and Seedance videos got me thinking. As a sports fan, I would love to watch and bet on LeBron’s 2016 Cavs vs Jordan’s 1996 Bulls, or Mahomes’ 2020 Chiefs vs Brady’s 2002 Pats, or Jeter’s 2001 Yankees vs Babe Ruth’s 1927 Yankees. Name your favorite ahistorical matchup or GOAT debate.
Now here’s a bold prediction: This product will be available to fans sooner than you think.
The Inspired and Seedance videos use totally different tech. Inspired’s Virtual Sports products are made with a 3D physics engine — similar to a video game like Madden or NBA2K. Meanwhile, the Seedance videos are made with an AI model that’s generating a net-new video.
But a combination of these technologies could soon make possible a realistic game between LeBron and Jordan, or Mahomes and Brady. The workflow would look roughly like this:
Step 1: A probabilistic simulation engine ingests historical stats, player tendencies, and more gameplay data to simulate a play-by-play script, and a statistically grounded outcome of the game.
Step 2: An AI video model would generate a video of each play.
Step 3: A production layer would stitch the plays together and add elements like the scoreboard, commentary and crowd noise.
Step 1 (the simulation engine) and Step 3 (the broadcast engine) are already technologically possible. Step 2 (the rendering engine) is a work in progress. AI video models can’t yet generate a 100%-realistic NBA or NFL play. There are too many characters, and too much contact between those characters — two things video models struggle with. But as the Seedance video above shows, models are getting closer.
I’d also argue other major sports could work tomorrow. MLB games feature fewer players on screen at once, completing more predictable movements. With the right resources, you could simulate a realistic broadcast of Shohei Ohtani’s Dodgers playing against Babe Ruth’s Yankees. Golf’s another interesting sport because the broadcasts show one player at a time.
So, let’s assume AI video models keep improving quickly, and a company emerges that can create these fully-realistic games. What does that mean for the sports industry?
The major leagues are all searching for new revenue streams — they’re seeing their domestic TV deals hitting a peak, they’re looking for more international growth, and they’re now feeling institutional investors breathing down their necks. Leagues are increasingly looking to build out their non-live content; when I worked at the NBA, their DTC platform League Pass was the largest company-wide initiative. And betting is an increasingly core part of the fan experience; 80% of American bettors say betting increases the entertainment value of live sports.
The major leagues could drive meaningful incremental revenues by monetizing both media rights and betting handle for these AI broadcasts. The betting handle could be significant because the margins could be closer to online casino (30-50% EBITDA margins) than sports betting (5-20% margins).
Here are examples of valuable media that could be created with today’s AI video models:
NBA All-Star Weekend has been flopping with fans. What if the NBA added to the broadcast a Three-Point Contest between all the greatest three-point shooters of all time?
MLB has a really long summer season. What if the MLB captured fans’ attention in late summer by letting fans vote on the historical matchup they wanted to see, and then created a Tyson-Paul style one-night primetime game on Netflix?
Golf has an awesome YouTube culture. What if the PGA & LIV collaborated with Bryson DeChambeau to “livestream” a round between Bryson and the five greatest golfers ever?
Each piece of content done right would drive real media dollars. I know leagues would have concerns around player likenesses; e.g. if Steph Curry loses to Larry Bird he won’t be happy! Player unions and retired players/estates are also barriers. But if branded as a casino-style simulation rather than a true testament to skill, then I think the leagues could overcome these barriers, and let the money do the talking.
I’ll leave you with one final Seedance video I enjoyed: Jordan and LeBron playing 1-on-1.
If this excites you as much as it excites me — and if you’re really excited and want to go build this company — then please reach out!
If only Mahomes and Brady had played each other.
Just another way to separate a fool from his money