sports analytics 101: team use cases (part 1)

Welcome back to Sports Analytics 101, a series of blog posts outlining the core concepts behind sports analytics in non-technical terms. In the first installment, I outlined at “The Case for Sports Analytics,” diving into the mental and situational constraints that analytics helps alleviate. In this installment, I begin introducing the use cases for sports analytics.

There are a number of sports industry stakeholders who can benefit from analytics, including but not limited to teams, leagues, and the media.

All of these stakeholders may share common metrics, methods, and strategies. In fact, it’s not uncommon for sports analytics pros to move between these stakeholder groups (e.g. a team analyst might leave to work for the league office). However, these three stakeholders each have slightly different use cases for analytics. This is the first of two posts that will cover use cases specific to teams.

Teams are probably the most widely-known sports analytics stakeholders. A team is, of course, the primary stakeholder in Moneyball. And within the domain of teams, there are, broadly speaking, three categories of sports analytics use cases:

  • In-game strategy
  • Player personnel strategy
  • Sports science: physical performance and medical decision-making

While much of the analytics work done within a team can be classified into at least one of these three categories, there is by no means a hard boundary between the three. A piece of analysis could fall into more than one of these categories. Analysts on staff could work on both in-game strategy and sports science, or on only player personnel strategy. The distinction isn’t always clear.

Some teams may use analytics in one of these areas but not the others. For example, a team might have an analyst that only works on in-game strategy, and no analyst working on player personnel or sports science. Another team might have a group of analysts that share responsibilities across all three categories..

The point is, there is no singular way to construct an analytics department to address these three categories of use cases. Regardless, classifying use cases among these three categories is a helpful way to frame how teams apply analytics.

In-Game Strategy

The use of analytics to inform in-game strategy comes down to using data to make better coaching decisions. This is often done as a complement to film study of opponents.

Before every game, coaches and scouts comb through tape of their upcoming opponent to identify weaknesses, mismatches, and any other useful information. Teams can analyze data in parallel to this process to generate insights that might have been missed in the film and flag situations that the scouts and coaches ought to review on film.

For example, let’s say a baseball team, the Bears, is preparing for an upcoming series against the Cougars. Opposition scouts for the Bears watch film of the Cougars’ projected starting pitchers and hitters while, in parallel, the analytics staff pours over the data.

The Cougars just brought up a promising shortstop prospect from the minors, let’s call him Ryan, and the Bears want to know how to match up against him. The scouts watch Ryan’s film from the minors and make careful notes, but in the end, Ryan has had hundreds of at-bats in the minors, so even with careful film study, it’s difficult to extract accurate high-level aggregate trends in his play through film alone.

However, the analytics staff has data on the trajectory of all of Ryan’s hits in the minors and the pitchers those hits were against. Using this data, they can build precise spray charts (charts of the directions and distances of a player’s hits) for Ryan against left-handed and right-handed pitchers. Because data can be aggregated in a way that film cannot in the human mind, the analytics staff is able to paint a more complete picture of Ryan’s tendencies than would be possible by watching hundreds of at-bats. With this data in hand, the analytics staff can recommend defensive shifts for the Bears to deploy against Ryan depending on which pitcher is in the game.

In this case, data complements film study by surfacing high-level trends uncovered by aggregating hundreds of data points. On the flip side, however, film can provide critical context for trends uncovered in the data.

Consider the Storm, a football team, whose upcoming game is against the Bulldogs. The Storm’s analytics staff finds that one of the Bulldogs’ wide receivers, Steve, who was widely considered the Bulldogs’ best offensive player last season, is having a relatively unproductive season this year. Taken at face value, this analytics-driven insight might suggest that the Storm should worry less about Steve, who isn’t the threat he used to be. Maybe the Storm considers playing a slightly different defensive scheme that shifts resources away from the secondary, or uses one of their weaker defensive backs to cover Steve.

However, the Storm’s analytics staff understands that complementary film study is critical to sports analytics, so they take this insight to the opposition scouts. On reviewing the tape of Steve, the scouts notice that while Steve has indeed had a fairly unproductive season, the Bulldogs’ opponents have been consistently playing a particular type of coverage against him, a coverage that few teams had used against Steve in prior seasons.

So, yes, Steve is having an unproductive season, but not necessarily because he’s become a worse receiver. It’s possible that he’s just as good as he was last year, but that teams are recognizing his talent and more effectively countering it this season.

In this case, the analytics staff’s dataset might not include data on the nuanced type of defensive coverage. Because the data set doesn’t contain information on all of the factors at play here, like defensive coverage, analysis of the data is left susceptible to context gaps that can be filled by film study.

If the Storm were to predicate their defensive strategy on the data alone, they would have concluded that Steve wasn’t a threat and, most likely, would’ve been burned for it. But because the analytics staff instead used the insight to ask questions of the film, they came to the correct conclusion that if Steve is matched up against a particular type of coverage, he can be neutralized.

The relationship between quantitative analysis and scouting is symbiotic: quantitative analysis can help overcome biases caused by cognitive heuristics and time constraints, while traditional scouting and film study can provide critical context to paint the numbers in the most accurate light.

Ultimately, analytics is typically used to answer a question or one type or another. What shift should we deploy against Ryan based on which of our pitchers is in the game? Which offensive players on the Bulldogs pose the biggest threat?

With this in mind, below are several additional in-game strategy questions that teams in a variety of sports could turn to analytics to help answer. Most of these questions have probably already been asked by a team somewhere. In later posts, we’ll get into some of the metrics and strategies that could be used to answer the questions.

Additional in-game strategy questions:

  • Football: In what situations should we go for it on 4th down?
  • Basketball: Should we let her shoot the three? Or should we play her tight?
  • Soccer: Should our goalkeeper pass out of the back against our next opponent? Or play it long?
  • Baseball: What is the most efficient way to use our bullpen? In what situations should we bring in our closer?
  • Hockey: When is the optimal time to pull our goalie?

To Be Continued...

As discussed above, in-game strategy is only one of the three broad categories of team use cases. Stay tuned for the next installment of Sports Analytics 101, which will focus on the other two categories: player personnel strategy and sports science.