Getty Boston Celtics coaches use analytics -- and gut feelings, too.
Mike Zarren has every reason to extol the virtues of basketball analytics, having accelerated his way up the Boston Celtics‘ organizational chart in part because of his expertise in that area. But he’s intelligent enough — and enough of a true hoop person — to be wary of absolutes.
So even while defending the digit mining that has become an increasing aspect of roster construction and game strategy, he believes in a holistic approach. That differs from some others in the NBA who preach a more rigid adherence to the stats.
“We’re never only making decisions based on numbers,” the Celts’ VP of basketball operations and team counsel told Heavy Sports.
Coming up on 20 years with the organization, Zarren, as described by the Celtics, “is widely recognized as one of the leaders in the field of advanced statistical analysis of basketball players and teams, and is an important part of the team’s strategic planning and player personnel evaluation processes.” In addition to the years of service, he was serving an apprenticeship of sorts as a local kid fan of the C’s.
Even with the “analytics guy” label now, he’s leaving no info unturned.
“You need every piece of reliable information you can get to make any decision that you need to make,” he said. “So whatever type of information proves to be reliable you should use. Sometimes it’s numbers. Sometimes it’s people’s eyes. Sometimes it’s something you heard from a buddy of a player. It could be any of those things.
“The scene in the movie Moneyball where there’s the old timey scouts sitting around the table smoking cigars, and the fat kid in the corner pokes his head up from behind the computer screen and says they’re all wrong is not the way things should work in the real world. And it’s created a dichotomy that probably I wish would die — analytics people versus other people. It shouldn’t be like that.”
‘Everybody Is Just Trying to Help Us Make Better Decisions’
Zarren came to the club with a certain amount of knowledge, having gone beyond his status as a supporter to study the game. His next lesson came quickly.
“I was going to every Celtics game growing up as a kid, and I thought I knew something about basketball,” he said. “Then you go to the first practice, it was a whole different language. So, you know, we’ve tried to make it so that the computers can speak the coach’s language. And the coaches have had to learn some of the computer’s language.
“Does anyone think the coaches drawing up the plays aren’t looking at some numerical information? Of course they are. So then the question is how reliable is it? Are they looking at the right stuff? Can we help them make any decision that they’re trying to make? So we just don’t differentiate between those groups. There’s certainly people on our staff who have machine learning and elaborate academic backgrounds and others who don’t. But everybody is just trying to help us make better decisions every day.”
Zarren pointed out that the field of basketball analytics today is giant steps beyond that which he brought in nearly two decades ago.
“You know, part of the problem when I showed up was we didn’t have the data that we have now,” he said. “So now I can actually teach a computer to recognize a flare screen. And so when coaches talk about the effectiveness of a flare screen, you know, back then we just had the box score, and so you really couldn’t have a conversation about the stuff the coaches were teaching the players using any useful numbers.”
Referencing the advent of Second Spectrum tracking, he added, “Now we actually can talk about the movements on the court. Since you have the cameras tracking players’ movements, it’s just been much, much easier to actually test a bunch of things. That’s been the biggest advancement the whole time I’ve been here. Now that we can track the players’ movements we can actually say what’s happening and have our analysts speak the coaches’ language.”
Zarren: Analytics Is About Finding Reliable Information
Zarren disputes, to a degree, the notion of some coaches expressed in the accompanying story that shooting analytics don’t hold rigidly true in all situations.
“This is classic statistical analysis,” he said. “We know the power of sample sizes. We know how much variance there is from night to night in shooting. And you can do a really good statistical analysis and say, here’s the band in which we think that guy’s actual shooting percentage falls.”
How the information Zarren and the Celtics’ crew gather is weighted versus the traditional eye test of scouts is another matter.
“The easy answer is you need to know how reliable your numbers are, just like you need to know how reliable your scouts are,” he said. “It’s not that weird. And it’s a little odd that sometimes, you know, there are different standards for how certain something has to be before you trust it. But really the question is, how reliable is any piece of information? And people make mistakes when watching things, and people make mistakes when coding statistical analyses, too.
“But every day you’re trying to get better and smarter about the decisions you’re making by looking at all the information you can get. The world is complicated, and so are the numbers going to be, too. So are the scouting reports going to be. The world’s complicated, and basketball’s a complicated game.
“We’re trying to take every piece of information we can get about every decision we make, and use it in the most reliable way. Some of those things are numbers, and some aren’t. And some aren’t clearly numbers or not. For example, a scout goes and watches a guy in a college game and gives him a rating. If you do something with that collection of ratings from all of your scouts, is that a statistical analysis or not?”
Zarren made a point to note that analytics is more than just taking every piece of information, throwing it into a blender and pouring out a conclusionary concoction.
“It’s important to know that we’re looking for reliable information,” he said. “There’s certainly a lot of unreliable information. And some of what statistical analysts are doing is not so much trying to answer question X, but instead trying to say, you know, what are we not good at answering? What should we not rely on? That’s important, too.”