Basketball is a complicated sport. It's got fewer traditional positions than any other team sport, but no less specialisation and far more fluid movement than many others. Two "point guards" can play drastically differently and still play the same basic position. Muthu Alagappan's research makes some sense of that.
Basically, Alagappan looks at what you're doing and where you're doing it. Instead of calling the tallest guy on the court the centre, figure out if he's primarily a scorer -- from the post or as a shooter? -- or a rebounder or a defender. And then do similar for everyone else. The data is then laid onto topological charts, which relate the data sets into shapes.
Alagappan's original paper from last year, presented at MIT's Sloan Sports Analytics Conference, had 13 possible positions. The new list has been refined to 10, with many of the old positions changed from qualitative to descriptive terms. So "3-point slasher" and "Two-way All-Star" replace designations like "One-of-a-kind" and "All-NBA-1st-Team". Here are the 10 new positions:
- Two-Way All-Stars
- Shooting Ball-Handlers
- Scoring Rebounders
- Paint Protectors
- Three-Point Slashers
- Role-Playing Big Men
- Rim-Attacking Defensive Ball-Handlers
- Three-Point Specialists
- Low-Usage Ball-Handlers
- Role-Playing Ball-Handlers
Alagappan also added a spatial component to his charting this year. So he's not just looking at what you're doing, like scoring, but where you're doing it at on the court. You can use this to infer how a team plays or possibly diagnose too many players doing the same things. Or, really, just look at the pretty pictures and nod along as your brain makes sense of them.
The Houston Rockets, a slashing, slightly duplicative, long-range shooting team, nearly overlap each other, while the San Antonio Spurs, an always well-spaced and efficient offence, are dispersed across the chart.
The data can also be applied to entire teams, charting stats like pace and types of shot:
Attacking data in sports like other data problems -- the company powering these charts typically works with data sets like mapping 25,000 genetic markers across hundreds of cancer patients -- can make it seem impenetrable. Hopefully, more of it gets articulated with relatively simple, awesome and digestible visuals like these. [Muthu Alagappan via Wired]