Saturday on SportCenter, I heard a statistic that showcased the evolution of analytical capabilities and the uncanny ability to drop a fact in almost any scenario. Concluding the highlights showing the New York Knicks losing to the Indiana Pacers, @randyscott dropped the nugget. WTH SportCenter.
Nembhand’s shot, 31 feet, is the second longest go-ahead field goal in the final 20 seconds of a 4th quarter or overtime in the last 25 seasons. Randy Scott, SportCenter
NBA Facts Exercise
Let’s break down how you could uncover this fact from a database. To do this, imagine an Excel spreadsheet of rows being individual shots and endless amounts of columns indicating attributes to the shot. Let’s see how fast we can filter the data.
First, the quote mentions the name of the person shooting, and in this case, just his last name. However, this component is not required because the fact is the occurrence and not whom.
Moving to the comment of “31 feet” requires that each shot has the distance from the basket. To do this, we start by sorting this column in descending order. Filter column number one is complete.
Next, we filter the column titled “go-ahead field goal.” Assuming that a “go-ahead field goal” can occur multiple times in a game, this column should be binary as yes or no. So we filter it only showing the “yes” shots. Because this is a “go-ahead field goal,” we are not required to filter the column showing only made shots. Moving on, now we have filtered our second column.
Looking at the “in the final 20 seconds” requires a column that shows the time the ball was shot. Filtering the data showing a time beyond 47 minutes and 40 seconds into the game will achieve this. Because this column is the amount of time into the game, filtering with this column would show rows in the 4th quarter and overtime. We have filtered our third column by showing all data greater than or equal to 2860 seconds.
Our next filter requires a date. Because Randy indicated “in the last 25 years,” I assume he does not have data older than 25 years. It is hard to imagine only capturing these data points manually, not to mention things like location on the floor, team, years of playing, or shoe size of the defender if the fact required. For this fact, we are using all available data.
SportCenter asterisk*
One asterisk is attached to Randy’s quote because he did not mention “playoffs” and the playoffs start on different dates yearly. Therefore, I will assume we would need to filter out the regular season shots on the column titled “Playoffs” because there have been many regular season game-winning shots that I can remember from beyond half-court. Our last filter indicating Yes on the Playoff column is now complete.
Every year, NBA players take about 200,000 shots. Each season, 30 teams combine to play 1,230 games, and at the end of the regular season, you can bet the sum total of shots taken will be very close to 200,000. Kirk Goldsberry, ESPN analyst
With what we have is Playoff (Yes), distance to the goal in descending order (Feet), go ahead field goal (Yes), and time into the game >2860 (Seconds). The four filtered data points show that Andrew Nemhband’s shot was ranked second.
Damian Lillard's "bad shot"
Randy did not mention who was first, but I remember the crazy “bad shot” Damian Lillard made a few years ago (I Googled it to conclude it was 36 feet). Four filters on a seemingly endless assortment of columns and 5 million rows of data to generate a single fact. One fact to sequence a conversation, answer a trivia question, or impress a friend in a bar. In Randy’s case, one fact that got me thinking about is the velocity of computations of good data.
This blog isn’t about Nemhband’s shot. Nor is it about Randy’s uncanny ability to drop factual dimes each weekend on SportCenter. Instead, this blog emphasizes the exponential speed demonstrated in the last 25 years on good data. And if you think this is fast, GPT3 can do it in less than 4 seconds, and GPT4 will be faster. As we get closer to instantaneous, we are getting closer to only being limited to asking good questions.
Comments