By Mike Armson and Manoj Raheja
Flexing our analytics muscles means approaching and solving problems with a scientific and actionable perspective. Our process is called ‘Write the Speech’ and it involves four steps:
- Think carefully about the desired outcome
- Identify the decision we’re trying to inform
- Generate hypotheses
- Do the research to confirm or deny the hypotheses
By following these steps and starting with the end in mind to help focus research, we avoid analysis paralysis and arrive at insights that are not only interesting but also actionable.
As big sports fans, especially when it comes to our local basketball team the Toronto Raptors (GO RAPTORS!), we decided to give ourselves an analytics practice session and showcase our skills with a hypothetical research project commissioned by Raptors coach Nick Nurse himself. Here was the scenario he posed to us (hypothetically!):
‘As we head into this year’s playoffs, we’re looking for an edge. Last year, we lost two critical playoff games by close margins in Round 2 against the Cleveland Cavaliers – Game 1 by 1 point in overtime, and Game 3 by 2 points. We have a great team, so we shouldn’t have to let it come down to the dying seconds like we did in those two games. What can we do during this year’s playoffs to gain an extra point or two each game so we don’t let it come down to the end? What is your recommendation to help us achieve our ultimate goal of winning the 2019 NBA title?’
We began by reframing Coach Nurse’s request in terms of our ‘Write the Speech’ process.
Think carefully about the desired outcome
The desired outcome was to help the Raptors gain one or two extra points during the course of each game to help them win more playoff games, and eventually, to win the NBA championship.
Identify the decision we’re trying to inform
What could the Raptors do to gain an extra point or two every game? One area we identified as a potentially fruitful area of investigation was turnovers.
(Definition time! In basketball, a turnover is when a team loses possession of the ball. There are two kinds of turnovers. Live ball turnovers happen when the ball is intercepted or stolen by the other team. Dead ball turnovers include offensive fouls, double dribbles, or when the ball goes out of bounds.)
Sports statisticians generally treat turnovers the same way. But are all turnovers really created equal? When you watch an NBA game, you might think that live ball turnovers, which result in a scramble to get back on defence, are more costly than dead ball turnovers, which allow more time to get set up and make sure the opponent’s players are covered. Perhaps live ball turnovers more often result in baskets scored by the opponent. If this were the case, then maybe intentionally fouling after live ball turnovers could reduce the number of points allowed per game by turning live balls into dead balls. That is, by fouling, the result would be the opponent getting the ball out of bounds with time to get set up, just like a dead ball turnover. With this in mind, we turned to the data to find out whether Raptors players should foul after live ball turnovers, and if so, how often.
Our hypotheses were two-fold:
- Live ball turnovers lead to more opponent baskets than dead ball turnovers.
- Intentionally fouling on live ball turnovers would reduce the number of baskets allowed per game.
Do the research to confirm or deny the hypotheses
Here are the steps we followed in carrying out our analysis:
- First, we needed to find the right data to answer our question. We came across a freely available play-by-play dataset for the full 2017-18 season at https://eightthirtyfour.com/data. This dataset comprised every play from every game allowing us to test our hypothesis about how often live ball as opposed to dead ball turnovers led to baskets by the other team.
- Since we were interested in how these turnovers affected the Toronto Raptors specifically, we decided to look only at turnovers committed by Raptors players.
- Turnovers were already tagged within the dataset, and we further categorized these turnovers as either ‘live ball’ or ‘dead ball’ turnovers. This was a new approach since all turnovers are typically treated the same way by statisticians.
- The last step was to identify baskets scored by the opponent directly after Raptor turnovers as 2-points and 3-points. These points were summed and used to calculate the percentage of live ball and dead ball turnovers leading to each type of basket by the other team.
So here’s what we found.
During the 2017-18 season, the Raptors committed more live ball than dead ball turnovers – specifically, about 8 live ball turnovers and 5 dead ball turnovers per game. As you can see in the left side of the chart, 32% of the Raptors’ live ball turnovers led to baskets by the other team, while only 23% of dead ball turnovers led to opponent baskets. This amounted to more opponent baskets per game directly after live ball turnovers than directly after dead ball turnovers.
Overall, these data confirmed our first hypothesis. Live ball turnovers were more costly than dead ball turnovers, more often leading to opponent baskets. These additional points from live ball turnovers could make the difference between the Raptors winning and losing in the playoffs.
How could this knowledge that live ball turnovers are more costly than dead ball turnovers help the Raptors gain an extra point or two per game? Or, in data scientist speak, how could we make this finding not only interesting but also actionable?
We initially hypothesized that by intentionally fouling on live ball turnovers, the Raptors could reduce their points allowed per game by turning more costly live ball turnovers into less costly dead ball turnovers. To simulate whether this strategy would actually benefit the Raptors during game play, we did exactly this within our dataset. We statistically treated some of the live ball turnovers as dead ball turnovers, and calculated how this would impact the points allowed by the opponent each game.
Given that each player is only allowed 6 personal fouls per game before being ejected from the game, and that each team is only allowed 5 fouls per quarter before automatically giving the other team free throws with each foul, it would be unrealistic to expect the Raptors to foul on every live ball turnover they commit. So if they can’t foul on every live ball turnover, on what percentage should they foul to reduce their points allowed per game by at least 1 point?
To answer this question, we calculated the percentage of live ball turnovers that would need to be converted to dead ball turnovers to yield a difference of 1 less point per game. As you can see in the right side of the chart, this calculation indicated that the Raptors would need to foul on 67% of live ball turnovers to reduce their points allowed per game by 1.
This seems much more realistic than fouling on every live ball turnover. It could actually be a viable strategy during game play. What it most likely translates to in a real in-game scenario is that the Raptors should consider fouling on live ball turnovers more often especially when the situation allows it. That is, the tendency should be to foul whenever the team is under the 5-foul-per-quarter limit and the Raptors players on the floor aren’t in foul trouble.
Reducing their points allowed by one per game in these playoffs might sound like a small number, but as we saw earlier, one point can be the difference between winning and losing in crucial games. If the Raptors had saved one point against the Cleveland Cavaliers in Game 1 last year, they would have won that game. That win could have changed the whole series last year, and it could very well help them win critical games on the road to their first NBA championship this year.
And that is a prime example of how to tackle a research problem to get an actionable result. Think about the desired outcome of gaining an extra point or two per game. Identify the decision to intentionally foul on live ball turnovers. Generate hypotheses that (1) live ball turnovers lead to more opponent baskets than dead ball turnovers, and (2) fouling on live ball turnovers would help reduce the points allowed per game by the Raptors. Confirm or deny the hypotheses to land on an actionable recommendation of fouling on two out of every three live ball turnovers to allow one less point each game.
Data science for the win!
Ready to learn more? Download our Sklar Wilton Research Decision Wheel for a template to help you categorize decisions to ensure your research plans are focused on the right big areas. Or, learn how we helped our health and beauty client identify key target groups, determine product positioning, and predict the size of the potential audience.
Perhaps you’d like these posts too…
- Eye Movements, Memory, and Market Research
- How Investing in World Class Research and Insights Planning Can Supercharge Your Results
- The Marketer’s Guide to Successful Artificial Intelligence: AI Ethics, Applications, and Examples
Mike Armson is a Research Management Associate at Sklar Wilton & Associates where he exercises his skills in health research, quantitative research, and statistical analysis. Mike also has a knack for developing innovative new research methods, which he uses to draw insights about market trends as well as human psychology. Mike earned a Ph.D. in Cognitive Neuroscience, with a focus on memory, from the University of Toronto.
A Partner at Sklar Wilton & Associates, Manoj Raheja is passionate about helping his clients create an aligned consumer view of the future so they have the necessary foresight and conviction to be the disruptors rather than the disrupted. He skillfully scopes and clarifies problems, listens and engages to inspire creative solutions, synthesizes and story-tells to move teams forward, all while leaving businesses, brands, and leaders better than before.
Sklar Wilton & Associates has worked for more than 30 years with some of Canada’s most iconic brands to help them solve tough business challenges to unlock growth and build stronger brands. SW&A was recognized as a Great Workplace for Women in 2018, and the Best Workplace in Canada for Small Companies in 2017 by the Great Place To Work® Institute. Recognized as the number one Employee Recommended Workplace among small private employers by the Globe and Mail and Morneau Shepell in 2017, SW&A achieved ERW certification again in 2018.