How ESPN data predicted the Warriors record season, made Big Data cool

By Noreyana Fernando

Columbia University

New York City, New York, USA


Two days ago, fans were thrilled by the Golden State Warriors who broke a 72-win record by winning their 73rd game against the Memphis Grizzlies. Who could have seen that coming?

Benjamin Alamar’s ESPN data team did.

Alamar, a PhD and director of ESPN’s sports analytics team, explained how the team used data to predict this record-breaking game well before it happened.

The Warriors record-breaking 73 winning games all comes down to numbers, Dr. Benjamin Alamar explains.
The Warriors record-breaking 73 winning games all comes down to numbers, Dr. Benjamin Alamar explains.

His team started by doing power indexes, which are essentially rankings, Alamar explained the second day of fhe Big Data Media Conference, a joint venture of World Newsmedia Network (WNMN) and INMA. The power indexes were interesting — but were soon taken to the next level when they were used to predict games.

It all started when Alamar wrote an article predicting a win — and article that drew many comments, along with the attention of ESPN’s programming team. The programming department was soon using Alamar’s daily data e-mails to determine which programmes would be broadcast, with the more competitive games taking precedence.

The next step happened when someone from the digital content group just by chance saw one of these automated e-mails and transformed it into an infographic that soon became one of the most successful pieces of content ESPN had produced.

“It was the kind of information we had never provided fans with before,” Alamar said.

Alamar took a step back in time to look at the evolution of sports data. In the beginning, there was only box score data. A mere summary of the game and that was it. Then came play-by-play data, whose every line was a tidbit of information.

Today, there is player, ball, and referee tracking. For example, in the NBA games, six cameras monitor each player and ball to record their position 25 times a second.

“That’s a massive new dataset,” he said. “This is happening in every major sport there is. It’s constant at the pro-level and it is starting to push down to lower levels too.”

In baseball, too, optical and radar tracking are used to track a player’s speed and distance. These are factual, cool but basic details. Data by itself does not suffice, Alamar said. It needs context, such as going back to a key moment of a game and predicting the likelihood of someone making that shot.

Equipped with a team of not just data experts, but machine learning experts and data visualisation experts, Alamar’s biggest take away was just start collecting data.

“The most valuable use for these data, we didn’t know when we were collecting it,” he said. “So just get going and trust that there will be a useful and valuable application to it in the end.”

About Noreyana Fernando

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.