Basketball Statistics (Help Requested)

by amitp06 on Apr 7, 2011 4:20 PM PDT

I'm in a college course strictly focusing on probability, and I need to do a project on a real world application of my choice. As a Blazers fan, I thought it might be interesting to investigate some aspect of NBA basketball that I wouldn't have time to research otherwise. I have some ideas for topics, but I thought BlazersEdge might have better ones. If there is interest and I actually do a good job, I'd also post the results later. I'm looking for your guys' help on two things.

1. What question or subject related to the NBA would you be interested in analyzing using probability? It doesn't matter whether it's specific to the Blazers, NBA, or something else; it only matters that there is enough data available. Keep in mind that this class is about probability, not general statistics. I can't make an advanced statistic or find correlations, but I can use all the major distributions.

2. Since there are a lot of stats people around here, what resources do you find most helpful for finding data? I know there's Basketball-Reference for data and WhatIfSports for simulation. Any other suggestions are welcome, even if they're similar to those two. I'm open to any extra information.

If you don't have any ideas, please rec the comments you find most interesting. To get started, here are a few general areas I had in mind. Feel free to develop them more or post something completely new. Even though some of these ideas are cliche, I'm trying to pick something unique, so put a twist on it if you can. Thanks for your help!

3 vs. 5 vs. 7 round playoff series. Given the probability that the better team wins each game (through simulation or data), I could do a lot of analysis on the series as a whole. How often does the better team win? How has the elimination of the 5 game round affected the NBA? What would a 3 game series look like?
Free Throws. I'm not too sure what I would research, but there are many ways to analyze a binomial situation like that. An obvious things to look at would be: how does free throw accuracy change in different situations? Clutch, away, playoffs, etc.
Home vs. Away. It's been done a lot, but there are always new angles. Similarly, how does home court advantage change in different situations? When and where is it the most/least helpful?

Tweet Comment 20 comments | 3 recs |

Do you like this story?

Comments

Something on referee bias

If refs call more fouls on away teams in the playoffs, or more frequently late in the game when the crowd gets into it.

Free throw percentage home vs road

by tkblazeys on Apr 7, 2011 4:27 PM PDT reply actions 1 recs

If you draft a center with an incredibly high pick

what is the probability he will turn up injured and unable to fulfill his potential for more than a brief span, if at all?

—Dave

by Dave on Apr 7, 2011 4:49 PM PDT reply actions

I love the black humor, but ouch...

I’ll turn that question into “What is the probability that said center recovers from injury to be successful?”

by amitp06 on Apr 7, 2011 4:56 PM PDT up reply actions

Speaking of draft picks...

What is the relationship between draft number and minutes played in a rookie season? Are higher draft picks always likely to play more minutes than those picked later? In other words, what is the probability that a #1 pick is a bust their rookie year?

Stealth > Wealth

by 500dogs on Apr 7, 2011 6:53 PM PDT up reply actions

Blake Griffin missed his whole first season

hasn’t seemed to hinder him too much this year

"If I had a dime for every basket I made today, you'd still suck!" - from the book 'John Dies @ the End'

by sammymohawk on Apr 8, 2011 10:57 AM PDT up reply actions

No it hasn't.

We’ll see if he can stay healthy playing the way he does (hard, fast and dumb).

Stealth > Wealth

by 500dogs on Apr 9, 2011 10:02 AM PDT up reply actions

The mike's had a good comment

what’s the probability that Wesley hits his first FT as compared to the second of 2 FTs. A bit too simple but a good stater question.

Other ideas: probabiltiy of a make-up call on the next possesion for a home and awy teams, probablity that lakers draw more fouls in the final 2 mintues of a playoff game than it’s opponent, or an advanced question what is the shooting percentage (from the field or combined with FTs) for a team during a game to have a 50% win probablity outcome.

by NWfan on Apr 7, 2011 5:11 PM PDT reply actions

This is actually a good one. Lots of other peoples' ideas I have already seen results on, but I can't recall seeing any on first FT vs second FT.

Especially given it could be 1-and-1 versus 2 shots.

You can measure skill and talent with your eyes, but productivity is shown through statistics.

by austinpwnz on Apr 7, 2011 7:07 PM PDT up reply actions

People have looked into that and found the percentages are almost exactly the same on the 1st and 2nd FT. In addition, the % on the 2nd FT is the same after missing the first FT as it is after making it. “Basketball on Paper” and “Scorecasting” both talk about that.

long live the jd.

by jksnake99 on Apr 8, 2011 12:14 AM PDT up reply actions

Other sources of data are 82games.com and Hoopdata.com, and if play-by-play data is needed, ESPN.com that for at least a couple years going back

I think a good question to ask revolves around the idea of being “in the zone” or being “hot” and look at the expected probability of a player making a certain amount of shots and then compare that to what actually happens. In other words you would work out some theoretical distribution of made shots (i.e. a player with a 50% fg% will be between 4-6 shots out of 10 {x} percent of the time) and then compare that to actual results.

by tingeyga on Apr 7, 2011 5:28 PM PDT reply actions

yeah, you could assume a player’s game fg% fits a Gaussian distribution and see if the data support that or if they are skewed.

long live the jd.

by jksnake99 on Apr 7, 2011 5:45 PM PDT up reply actions

I was playing around recently using the binomial distribution for clutch free throws, making use of 82games.com’s clutch stats. You’ve got some solid ideas there.

long live the jd.

by jksnake99 on Apr 7, 2011 5:44 PM PDT reply actions

I second the mention of 82games.com and hoopdata.com

For play by play data, I think this is one of the better sources:

http://www.basketballgeek.com/data/

You can measure skill and talent with your eyes, but productivity is shown through statistics.

by austinpwnz on Apr 7, 2011 7:09 PM PDT reply actions

Since I see they haven't been mentioned yet, I like NerdNumbers, even though Wins Produced has issues

Also, Kevin Pelton’s Basketball Prospectus has some stats using WARP. I don’t know how much you can still access without paying, though. Both sites would mainly provide you with stats that measure overall player value/production.

You could look at the probability that a star or superstar player will become less productive when paired with another star or superstar.

Phase 1: Collect underpants
Phase 2: ???
Phase 3: Profit!

by HailOden! on Apr 7, 2011 7:29 PM PDT reply actions

Good win tonight

Thank you to all responders!

I checked out all of the sites. I’m glad there are things like clutch and play-by-play data. As for the suggestions, they’re all pretty interesting, so I’ll start thinking about which one is most feasible for me. I still have a little time, so keep the comments coming.

by amitp06 on Apr 7, 2011 10:05 PM PDT reply actions

All this stuff makes me feel dumb

And maybe it should. The ancient gods regarded humility as a virtue. Not so much anymore, which is why nobody worships Zeus these days.

I’ve wondered about aging. What’s the peak age for a player, and for what type of player? Not just position, but style of play? Obviously some talents weather aging more gracefully than others. Is there a consistent theme to this or is it genetic luck of the draw? In baseball power pitchers tend to lose their skills faster than control pitchers. Is there a basketball corollary? I’m thinking how Grant Hill resurrected his career while Brandon struggles with how to re-design his style. Or how a John Stockton or Reggie Miller stayed potent for years, possibly because they weren’t “create your own shot” types — Stockton was a passer, Miller an off-the-ball-pick specialist. Is there such a thing as basketball IQ that can extend a useful career (if so, today’s Celtics have it) and how would you measure it?

by twinsbrewer on Apr 8, 2011 10:06 AM PDT reply actions

The thought I have is that a lot of the ideas, while interesting,

sound correllational in nature. Just sayin’ you might want to be sure you don’t get “off topic” on your assignment.
While it is not very original, probability is perfect for working with card games. Analyzing how lucky some world series of poker winner actually was, for instance. That could be interesting.

Jordan's not a bad guy.. he gave us Gerald Wallace.

by Berkeley on Apr 8, 2011 10:51 AM PDT reply actions

What is the probability that a random basketball statistic has anything to do with probability?

Basketball is a game of spontaneous human skill. If it were simply a stochastic process, it could be modeled with ease and there would be little reason to play the games. When basketball is reduced to equations for video game design, I think it gives people the idea that statistics give probability, but this isn’t so.

IMO, most basketball ‘stats’ are merely a tally of unrelated events. Just because something can be expressed as a percentage does not make it a probability. The reason for this is that most stats only track one outcome of a huge system that can’t be measured. It’s like tracking the weather by day of the week and attempting to predict future weather. “B-b-b-ut it’s been windy and rainy on 67% of Fridays since 2009. There is a 2/3 chance of wind and rain, because it’s Friday.” Uh, no. Wind and rain depend on the weather, not day of the week. A perfect example of this is a streaky shooter like Rudy. If his shooting percentage truly was a probability, he would be much more consistent from game to game. A streaky shooter’s skill is like the weather — its current state can’t be predicted by a historical record tracking only one specific aspect of its outcome.

The very worst basketball stats usually surface for the playoffs. Wheels/Barrett will undoubtedly tell us that the Blazers have won an opening round playoff series ‘X’ percentage of the time when they don’t have home court advantage. They may go on to tell us our historical playoff record against our specific opponent in the first round. These stats are about as useful as tracking the weather by day of the week, from random cities around the globe. This is my single biggest statistical pet peeve, and yet they are always used in pregame shows and newspaper articles as if they have any bearing on the series.

by manfredi on Apr 8, 2011 4:33 PM PDT up reply actions

This is a good point

I was surprised no one called me out on it before. I know I’ll have to make some assumptions no matter which topic I choose, and not all of the assumptions will be valid. I’ll try to pick something where the impact is as small as possible. Even though very little in basketball is pure probability, there are still things to learn from the models.

by amitp06 on Apr 8, 2011 4:58 PM PDT up reply actions

True, I'm trying to be careful with my selection

If none of the basketball ones are feasible with probability, my professor has some fascinating ideas on his site. If you’re interested about the cards, here’s an article about card counting and probability that he put up. Poker could be interesting.

by amitp06 on Apr 8, 2011 5:02 PM PDT up reply actions