Can data science make your team win? Some teams prove so
The overall strength of a professional sports team is measured with one straightforward metric: wins.
However, winning one game, let alone a championship, is extremely difficult in professional sports leagues. As the popularity of sports has grown worldwide over the last century, so has the level of competition in professional sports leagues and what it takes to win.
Athletes today are bigger, faster, stronger, and more skilled than their counterparts from previous generations. Professional sports organizations need to look for any advantage to help put their teams in the best position to win.
More and more pro sports organizations have turned to data science in search of competitive advantages in recent years.
One sports team that has consistently used data science as part of its recipe for success is the English football club Liverpool FC. Although all professional football teams feature an analytics department, Liverpool is unique in its approach.
Liverpool’s diligent analytical work informs its high-powered offense. Their analytics team built pitch control models to help determine how each action on the pitch would affect the probability of its players scoring a goal while in possession of the ball.
Pitch control refers to the probability that a player can control the ball at a given point on the pitch. To collect the data needed to make these models work, cameras are set all around the stadium that tracks the position of all players on the pitch, as well as the football. Over the course of a single match, analytics teams may end up with as many as 1.5 million data points to work with.
While pitch models come in handy during pre-and post-match analysis, they’re also utilized in real-time to help inform the team’s strategy. As such, it’s on the players to execute the strategy. Each precise pass by Liverpool’s players has an effect on the opposing defense, creating open spaces and expanding the areas Liverpool players can run into on their way to scoring.
This can often be seen in the use of long diagonal passes across the pitch that can outflank several defenders in one pass. The pitch control model identified these areas as the best areas of the pitch to use on offense, which in turn enhanced the players’ ability to execute high-percentage offensive possessions.
As a result, Only Manchester City scored more goals than Liverpool in the 2019-20 Premier League season. In fact, Liverpool finished at the top of the standings, with 32 wins.
On the other side of the pond, American sports leagues are no strangers to the power of data, either. Former professional player, Oakland Athletics general manager Billy Beane shook up the American sports world in 2002 with his use of sabermetrics to assemble a winning baseball team with a relatively limited budget.
Instead of looking for players with good basic stats such as stolen bases, runs batted in, and high batting averages, Beane sought players with high on-base and slugging percentages. He believed these metrics were more indicative of a player’s offensive success.
His hunch paid off, and he was able to send the Athletics to the playoffs in 2002 and 2003. By selecting good but undervalued players on the market, he was able to keep his team competitive against teams with far bigger payrolls.
Data science and basketball
Analytics has also made its way into the NBA. A myriad of advanced stats such as Player Efficiency Rating, Win Shares per 48 Minutes, and Box Plus-Minus have emerged in recent years to help ascertain each players’ overall contribution to producing a win.
While all these metrics have aided teams in roster construction, not all analytics in basketball is reliant on data-based models. Former Houston Rockets general manager Daryl Morey’s whole approach was centered around one simple fact: three points is worth more than two.
He worked with the Rockets’ coaching staff to devise an offensive strategy that emphasized the three most efficient ways to score in basketball: free throws, layups, and three-point shots.
Morey relied on one key statistic, which he gathered from teams across the league to put together this game plan: points per shot. This is the average amount of points a shot would score upon a successful attempt, factoring in its point value as well as the probability of the ball going in the hoop.
In 2012, he landed the perfect player to make this strategy work: James Harden. Harden fit this strategy like a glove, due to his three-point shooting ability, driving ability, and knack for drawing fouls. Previously a bench player for the Oklahoma City Thunder, Harden made an immediate impact upon his arrival in Houston. He put up 37 points on his debut, with 16 coming from layups, 6 from free throws, and 12 from three-point shots.
With James Harden at the helm, the Rockets were able to field three of the most historically efficient offensive seasons in league history. Despite this success, they were never able to complete a successful championship run. In a cruel bit of irony, their hyper-efficient, analytically-driven game plan ended up working against them during their best chance to win it all. In the 2018 Western Conference Finals, the Rockets missed 27 straight three-point shots against the eventual champions, the Golden State Warriors.
As powerful a tool as analytics can be in influencing a team’s success, there are also an endless amount of variables that can render any analytical model useless. Players are only human. They can get injured, have off nights, or simply just fold under pressure.
All in all, data science has changed sports. It’s given fans and coaches more tools to compare players and new ways of formulating strategies. But at the end of the day, it’s the athlete playing the game, and not the data.
A new study shows when scientists are most innovative and creative in their careers. We talk to the scientists behind the discovery.