Nate Silver, the editor-in-chief of FiveThirtyEight, said a lot of people get data wrong, but there are ways to avoid such an issue.
Silver, who made headlines last year for giving President Donald Trump the best odds of becoming president compared to other news organizations, spoke at Domo's Domopalooza conference in Salt Lake City as the keynote speaker.
He said he’s fascinated by Utah since the Beehive State shifted more than any other state in the 2016 election.
When making predictions, he said experts can get major events wrong, even when they're predicting using data. For example, few people predicted Trump, Brexit or the 2007 financial collapse. He said big data can create big problems.
But he’s worried about how much data influences our country right now. Silver said companies are at a point where data has become heavily popular.
“I’m worried a little bit about where we are with how people view data and analytics,” he said. He added, “I’m worried where we are in this hype cycle."
This fear is mostly because people manipulate data to fit their opinions or worldview.
For example, on Election Day, people showed Trump with a 30 percent chance of winning and others had a 0.5 percent chance. People can manipulate and read data in ways to fit their ideals or what they want.
“We don’t really have real strategies with dealing with data sets this large,” he said. “It’s almost like philosophically we aren’t used to being bombarded with so much information.”
Silver offered some suggestions on how to read data accurately:
Think probabilistically
Silver said it’s important to account for uncertainty. Plan on what could possibly happen, even if it seems unrealistic. Margins of error are important, he said.
Know where you’re coming from
Silver said you should be aware of your background and where you’re coming from. To be successful, develop a diverse group of people, express your own opinions and ask others for suggestions.
By understanding your culture and your own background, you’ll find ways to hire the right people, he said. You'll also account for human error and reactions.
Try and err
Silver said people shouldn’t be afraid to make improvements. Test your data on real-world customers to see what works and what sticks and then embrace those products, he said.
It’s good to learn in real time, Silver said.
He finished his talk by saying that people should shouldn't rely simple on their gut, either.
“If people are going for their gut for that, that’s really dangerous,” he said.
Notes from Silver post-speech interview:
On his poker career: Silver played a lot of poker when he was younger, adding he learned a lot about outcomes and skills from playing the card game. “I kind of remember it fondly-ish,” he said.
On sports vs. politics: Silver works with ESPN to create data analysis about various sports. Sports, he said, is easier to predict than politics. Politics “doesn’t follow well-defined rules,” he said. Politics is constantly changing, so it’s harder to follow trends all of the time. “There are precedents for Trump, maybe not in American history, but certainly in world history,” Silver said.
What happened in the 2016 election: Silver said FiveThirtyEight became “more right” by giving Trump higher chances. He said the biggest problem was that a lot of forecasters saw states as independent. States like Ohio, Michigan and Wisconsin have similar demographics, which make them really one voting group. People don’t know how to translate polls to probabilities, either. Sure, Clinton had a three-point lead, but it still limited her probability.