2016 Three-Point Contest: Simulation Results
By Justin
There are few events during All-Star Weekend appropriate for statistical analysis — the dunk contest is a pageantry of youth and crowd woos while the celebrity game is halfway between a Division III college game and hell — but the three-point contest is more grounded in reality and somewhat predictable. Based on research I did previously, I wrote a basic simulation model for the three-point contest using results from 2000 to 2015. The model is simple partly because so few factors actually matter: I used a combined three-point percentage from the past three seasons and a dummy variable for the final round, as players typically shoot a little better. That’s it. Other things like height and volume did not seem to matter, and I’m still searching for another variable to help explain the madness.
This year, Stephen Curry is torching the NBA with the most incredible shooting season anyone has witnessed, so it may surprise people to see how mortal he is with these numbers. But JJ Redick, like Kyle Korver last year, has some sensational percentages, thus the simulation sees him as the most likely winner in the summary below. Based on the odds I’ve seen[1. In case you didn’t see the update, Chris Bosh was replaced by CJ McCollum in the three-point contest.], Redick is being underrated but he’s in competition with Curry, so I understand that. But one sleeper might be Khris Middleton, who’s the long shot even over guys like James Harden and the rookie. In fact, my model is probably overrating Devin Booker because he has so few attempts.
Player | Expected 1st round score | Odds% |
JJ Redick | 16.9 | 21.9 |
Stephen Curry | 16.8 | 21.8 |
Klay Thompson | 16.4 | 16.7 |
Khris Middleton | 15.7 | 11.5 |
Devin Booker | 15.6 | 10.3 |
CJ McCollum | 15.2 | 8.2 |
Kyle Lowry | 14.4 | 4.9 |
James Harden | 14.3 | 4.6 |
After watching the three-point contest and building a simulation, there is one important fact I should impart on people: it’s largely random, driven by chance and leading to improbable upsets regularly. Betting the field is always the smartest bet here. For an illustration of this, I replaced CJ McCollum with Timofey Mozgov, who’s shot 18.9% on three-pointers over the past three seasons — and someone let him take 37 of these. The numbers here aren’t perfect, of course, mostly because the stats have never seen someone like him in this contest, but in 10,000 iterations he actually won three contests. I don’t know if that’s a failing of my model or just a result due to a large number of trails and randomness, but I wish we existed in that universe. For what it’s worth, his expected score in the first round is 9.0, which makes his performance of 22 in the finals (as seen below) somewhat miraculous.
Table: results from the best contest possible
Player | 1st round score | Finals |
Stephen Curry | 13 | |
James Harden | 10 | |
Klay Thompson | 20 | 16 |
Khris Middleton | 10 | |
Kyle Lowry | 17 | |
JJ Redick | 25 | 21 |
Timofey Mozgov | 18 | 22 |
Devin Booker | 8 |
Another fun diversion from this model is that I can extract the extreme values and find virtual contests with bizarre results — the calibration I did is not perfect, but it hews to the real-world surprisingly well. For instance, one simulation run ended with Stephen Curry smashing Curry’s old record for points in a round with 31. Curry had 27 last year, but there was an additional moneyball rack. Using the old rules, that score of 31 would be at least 27, beating the old record of 25 held by Craig Hodges and Jason Kapono.
Table: simulation run with record score
Player | 1st round score | Finals |
Stephen Curry | 23 | 31 |
James Harden | 18 | |
Klay Thompson | 21 | 22 |
Khris Middleton | 12 | |
Kyle Lowry | 18 | |
JJ Redick | 19 | 23 |
CJ McCollum | 17 | |
Devin Booker | 12 |
We’re living in Stephen Curry’s world right now, but the numbers say JJ Redick and Klay Thompson have a good shot at defeating him. Tonight’s three-point contest won’t feature 100,000 simulation runs like my model. We’ll get only one, and somehow that’s our reality.