By Kurt Bullard

Steph Curry captured the Internet with his 46-point performance last Saturday, going from 12-for-16 from beyond the arc and launching a game-winner from a few feet short of half court. But somewhat hidden in Stephen Curry’s marvelous showing against the Thunder was his adding to his streak of making at least one three-pointer, extending the record to 129 games in a row in which Steph hit at least one from deep. Based on Steph’s rate of chucking up shots from deep and the absurd accuracy that he exhibits in making these shots, this streak could have no end in sight.

The streak can end in one of two ways. Either Steph does not take a three-pointer and thus cannot make one, or he takes some but misses all of them. In order to find the percentage at which Steph would either not take a three or miss all of the ones he took, one can look at his stats from the last two season, since the 129-game stretch has spanned the past two years. Steph has made 4.22 three-pointers per game over the last two seasons. You can model the three-pointers made by Curry per game using a Poisson distribution, with a rate parameter of 4.22.

Doing so, you can estimate the percent of games Steph would be expected to go without a three. The percentage of games Steph would go without a three is e^{-4.22}, or 1.47% of games he plays. Then, you can use that percentage in a negative binomial distribution to find the probability distribution of the number of games more that the streak will last.

Below is a summary of the cumulative probability that the streak lasts a certain amount of games.

So there is an approximately 1-in-4 chance the the streak lasts at least 100 more games, and about a 1-in-20 chance the streak lasts at least 200 more games. From the distribution, you can also ascertain that on average, this streak would be expected to last 67 more games, assuming an injury early in the game derails the streak by chance.

This model does not take into account Curry’s desire to keep this streak going. If Curry wants this streak to continue, he might chuck up a few threes at the end of games to make at least one, meaning that this model might underestimate the amount of threes per game that Curry ends up taking. So, the above probabilities may indeed be an underestimate.

So for the extreme minority of you waiting for the end of Steph’s streak, feel free to breathe out.

### Like this:

Like Loading...

*Related*

Kurt,

I have had the exact same idea but didn’t know how to go about it. Are you using game by game stats of the number of 3 pointers made? Have you updated your analysis?

Below are the figures I have from http://www.basketball-reference.com/players/c/curryst01.html

Best,

Mark

Season

15/16 14/15

0 0. 1 (11/11/14).

1 5. 12

2 6. 20

3 13. 18

4 11. 13

5 12. 14

6 6. 12

7 8. 6

8 8. 4

9 3. 0

10 2. 1

11 1. 0

12 1. 0

When I wrote this, I was looking at the average threes per game made over the last two season, which at the time was 129 games. I modeled threes made as a Poisson RV with a rate parameter of the average (4.22), and then found the probability of getting a zero from a Poisson(4.22), which is 1.7%. I used this as my parameter in a negative binomial distribution, which models how many trials will occur before a certain event happens.