NEW YORK—Day-after conversations around the water cooler about sporting events inevitably come around to “did you see that play” or “call”—and when it comes to the Super Bowl, “commercial.”
However, no one ever thinks to ask “when did you see that” –no one except for real-time streaming specialist Phenix. The company, which has created its own spin on WebRTC to power live video streaming at scale with about 500 milliseconds of latency, is keenly aware of the when, especially when it comes to Super Bowl LIII.
“On average, there was a streaming delay ranging from 30 to 47 seconds depending on the platform,” says Jed Corenthal, CMO of Phenix.
To arrive at that finding, Phenix compared the time plays actually occurred on the field at Mercedes-Benz Stadium in Atlanta to when the same play began on eight platforms used to stream the game, including the CBS Sports App, NFL.com, Xfinity, CBSSports.com, YouTube TV, the NFL App, CBS All Access and the Yahoo Sports App.
The company released its findings Feb. 5 in the form of two online graphs, one laying out the averages for measured lag behind real-time for Super Bowl 2019 and the other presenting the measured audience drift –or ranges of lag for viewers on the same platform—for the game.
More than 800 data points were collected for the eight streams on multiple devices, such as laptop computers, smartphones and tablets. Each was compared by measuring the delay behind the over-the-air, cable or satellite TV signal.
To establish the start times for each sampled play, a Phenix researcher in the stadium reported on a cellphone when measured plays actually began, explains Corenthal, adding that using the cellphone created about 40 to 50 milliseconds of latency.
Some 20 people around the country using different digital devices and streaming platforms then recorded when the same play began on the streaming devices. Measurements were taken throughout the game.
Phenix found that the CBS Sports App performed best among the streaming platforms with an average of 28.2 seconds of latency compared to the on-field game. On the other end of the continuum was the Yahoo Sports App with an average of 46.7 seconds of latency.
The study also looked at audience drift, or the range of lag viewers on the same platform experienced. YouTube TV recorded the greatest drift with 107.8 seconds, while the NFL App with 69.3 seconds recorded the least drift.
The culprit is HLS, or HTTP Live Streaming, which “wasn’t designed for real time,” says Corenthal.
“They [HLS-based platforms] may be able to do things to it that alter the latency of their particular stream, but when you add up the encoding, the content delivery and all of the points along the way, what happens in HLS is you are kind of relegated to a certain amount of latency because that is the way the technology is built,” he says.
Latency may or may not be a problem for viewers, but any Super Bowl fan who was watching the game on traditional TV and also turning to a smartphone or tablet for the game likely saw the effects of streaming latency firsthand.
“Look, for some people that’s OK. It’s not a big deal. But what we are seeing is the general fan is getting tired of that; tired of streams that are behind; streams that are buffering,” says Corenthal.
“The good news is most of the streams had pretty good quality, and quality and scale have been the first two pegs of the triangle that have been somewhat solved. It’s when latency is added –all three pegs of quality, scale and latency—that’s where these platforms come up a bit short in our estimation.”
More information is available on the Phenix website.