Posted: Thu Sep 28, 2006 10:28 pm Post subject: New website: BasketballValue.com
Hello everyone,
My name is Aaron and I wanted to let you know about a website I'm developing at http://basketballvalue.com. It's probably not even at the beta stage yet, but with this post I'm letting the public know about it.
With a couple of friends, we've been working on developing adjusted +/- ratings for NBA players. That work is still in progress(as Dan Rosenbaum has mentioned in his posts it's a tough problem), but it has taken quite a lot of effort just to set up the infrastructure and build the database for use in our analysis. I've seen a few posts asking where people can get their hands on this kind of data (e.g. all the game logs from 2005-6), so that information is available on our website. Also, we've built a simple application that lets you drill into data on the +/- of 5 man units to demonstrate the data we have.
The next steps are to also show the +/- of individual players, which will be coming soon, and adjusted +/- of units as well as players, which may take longer. I also plan to maintain it with data from the upcoming season as well. If there's anything else you'd like to see, please either post to this thread or drop me a line at webmaster@basketballvalue.com I look forward to hearing your comments as well as thoughts on what else you'd like to see.
This is excellent. Thanks! Now I can find out +/- information for an arbitrary set of teammates. How often will this be updated during this season?
If I'm not mistaken, you're counting an offensive possession and the following defensive possession as 2 possessions, right? Perhaps you could split this up into offensive possessions and defensive possessions (since they won't be the same). Then, the +/- would be points forced per 100 off. possessions minues points allowed per 100 def. possessions. Intuitively, this makes more sense to me.
Also, I notice that your total possessions for Kobe (which is the only example I checked) is about 600 more than what 82games.com numbers would indicate (12994 to ~12405), with only minor differences in minutes played and points scored/allowed.
Last edited by deepak_e on Fri Sep 29, 2006 1:37 am; edited 2 times in total
This is excellent. Thanks! Now I can find out +/- information for an arbitrary set of teammates. How often will this be updated during this season?
If I'm not mistaken, you're counting an offensive possession and the following defensive possession as 2 possessions, right? Perhaps you could split this up into offensive possessions and defensive possessions (since they won't be the same). Then, the +/- would be points forced per 100 off. possessions minues points allowed per 100 def. possessions. Intuitively, this makes more sense to me.
Also, I notice that your total possessions for Kobe (which is the only example I checked) is about 600 more than what 82games.com numbers would indicate (12994 to ~12405), with only minor differences in minutes played and points scored/allowed.
Thanks for the feedback, I'm glad to hear you like it.
Yes, I'm counting offensive possessions as a single possession, and defensive possessions as a single possession by the other team. That is why on the top units page I show +/- per 200 possessions as an approximation of +/- over one game, rather than 100.
If I understand your suggestion correctly, wouldn't I basically just divide possessions by 2 and show +/-per 100 possessions? If not, what would balance out the scoring of the offensive possessions? Or are you saying one could try to trim off the cases where there's only "half an inning" in a baseball analogy (e.g. 3 possessions in the matchup, 2 on offense and one on defense)? I guess I'm hoping that if there's enough data to produce reasonable conclusions, the % of possessions on offense will tend to 50% as the extra possessions will balance out.
As far as matching up with other sources, I have not yet done an in depth comparison with data from other sites. I do expect to do more of that at the beginning of this season as it will be easier to track down mismatches. I'm sure there are some small bugs in my code for breaking down games into matchups, but I believe it's 99% accurate. I have compared the automated breakdown with manually breaking down games by matchup to enhance and verify the accuracy, I have corrected all the issues I was able to find.
Finally, I would love to say that I'll be updating these stats daily during the season, but I don't have the process automated at this point so I suspect that won't be the case (in fact, the game logs have enough scoring errors that it would be hard to automatically correct for them). I'll do my best to make updates as close to daily as possible, though.
Joined: 03 Jan 2005 Posts: 460 Location: Washington, DC
Posted: Fri Sep 29, 2006 11:15 am Post subject:
I find the way you're counting possessions sorta confusing. We're all accustomed to "per 100 possessions" stats. It would make more sense to me to see the numbers broken down by offensive performance (pts per 100 offensive possessions), defensive performance (pts per 100 defensive possessions), and then have a net +/- per 100 possessions. _________________ If you can't explain it simply, you don't understand it well enough.
Hey Aaron. I am very interested in this effort to allow exploration of simple +/- data and the promise of public access to adjusted +/-.
I took forward to being able to drill down on your site on any specific 5 man unit or a summary of all such for a specific player in the way you have demonstrated for top units. Would you estimate you are weeks or months away from sharing adjusted +/- data?
Would you be willing to consider adding later an additional dataset that makes just one change to adjusted +/-, giving 50% of the credit for any points scored to the individual directly responsible? It might be interesting to see the contrast between the adjusted +/- scores received by high scorers and lesser scorers from these two approaches (even over just a subset of data).
Last edited by Mark on Sat Sep 30, 2006 10:27 pm; edited 1 time in total
Yes, I'm counting offensive possessions as a single possession, and defensive possessions as a single possession by the other team. That is why on the top units page I show +/- per 200 possessions as an approximation of +/- over one game, rather than 100.
If I understand your suggestion correctly, wouldn't I basically just divide possessions by 2 and show +/-per 100 possessions? If not, what would balance out the scoring of the offensive possessions? Or are you saying one could try to trim off the cases where there's only "half an inning" in a baseball analogy (e.g. 3 possessions in the matchup, 2 on offense and one on defense)? I guess I'm hoping that if there's enough data to produce reasonable conclusions, the % of possessions on offense will tend to 50% as the extra possessions will balance out.
Dividing possessions by 2 to get +/- per 100 possessions won't be exactly the same. That would assume that there are the same number of offensive and defensive possessions, which isn't quite true. To take an example, consider the 5-man unit for the Lakers:
{Parker, Bryant, Odom, Brown, Cook}
You have them listed with a NET +/- of 10.45. According to 82games.com, Kobe Bryant saw 0.4% more offensive possessions than defensive possessions. Doesn't seem like a lot, but if we suppose that the above 5-man unit saw 0.4% more offensive possesions than defensive possessions, their NET +/- turns out to be 10.03.
In most cases, the difference will be much more, since 5-man units aren't on the court for 3000 minutes in a season. In general, I'd expect the difference in offensive and defensive possessions for a 5-man unit to be far greater than 0.4%, so the plus-minus numbers for your 5-man units will be further off.
I think the presentation would be a bit more clear (at least for people who frequent this forum) if the columns were:
Team,
Players (preferably ordered by position as best as possible),
raw +/-,
points for,
offensive possessions,
points against,
defensive possessions,
minutes,
PF per 100 off. poss,
PA per 100 def. poss,
NET +/-
Quote:
As far as matching up with other sources, I have not yet done an in depth comparison with data from other sites. I do expect to do more of that at the beginning of this season as it will be easier to track down mismatches. I'm sure there are some small bugs in my code for breaking down games into matchups, but I believe it's 99% accurate. I have compared the automated breakdown with manually breaking down games by matchup to enhance and verify the accuracy, I have corrected all the issues I was able to find.
The big difference I'm noticing is the possession count. I totaled up all the possessions for Kobe from your 5 man units, and it turned out to be nearly 600 more than what the numbers at 82games.com indicates. I'm not saying its a bug on your end, but it would be great for the numbers on both sides to be pretty much the same.
Quote:
Finally, I would love to say that I'll be updating these stats daily during the season, but I don't have the process automated at this point so I suspect that won't be the case (in fact, the game logs have enough scoring errors that it would be hard to automatically correct for them). I'll do my best to make updates as close to daily as possible, though.
That would be great. I think this will turn out to be a great resource. Thanks again.
BTW, are you planning on determining player posititions for your 5 man units (ie. PG, SG, SF, PF, C)? 82games does that, though as WizardsKev has pointed out it isn't always successful. As you update your program, that might be something to consider adding.
Last edited by deepak_e on Fri Sep 29, 2006 1:58 pm; edited 1 time in total
Here's a comparison of the numbers from your site and some of the other frequently used sites here:
Code:
Lakers
min PF PA Poss
BV.com 3971 8154 7949 15656
82games.com 3964 8161 7950 14905
B-R.com 3971 8154 7949 --
Code:
Rockets
min PF PA Poss
BV.com 3966 7387 7517 15215
82games.com 3958 7388 7521 14347
B-R.com 3966 7387 7517 --
I think B-R.com's tally of minutes, points forced, and points allowed are precisely from the box scores, so your answers on those are clearly correct. However, there is a discrepency between the total possessions from your site, and the possessions from 82games.com (which I calculated using the player On/Off data).
This indicates to me that it's not a discrepency in how matchups are being calculated and tracked, but rather in how possessions are being counted. Either your site is overcounting them, or 82games is undercounting them.
Joined: 03 Jan 2005 Posts: 460 Location: Washington, DC
Posted: Fri Sep 29, 2006 2:19 pm Post subject:
deepak's comments about possessions has me wondering: How do you define what a possession is? _________________ If you can't explain it simply, you don't understand it well enough.
There have been some discussions here about the definition of a possession before. I think getting a black-and-white agreed upon definition is important, especially if one ever hopes to have the NBA's official scorers record possessions in boxscores.
A possession starts when one team gains possession of the ball and ends when the other team gains possession. Basically one possession ends and a new one starts upon one of these events:
- made shot where player is not fouled
- made final free throw (i.e. not the first of two or a technical FT where the team that shoots it keeps possession)
- defensive rebound
- turnover
Events that don't end a possession:
- non-shooting foul (when not in bonus)
- technical foul on the defense
- kicked ball
I think in some thread there's a discussion of the situation where Team A makes a shot and is fouled, misses the free throw, gets the offensive rebound, and scores again. I think most would consider that to be one possession even though two field goals were made.
Some questions I have deal with jump balls. First, does anyone know if a team is credited with a turnover if they are on offense, get tied up, and lose the jump? If that doesn't count as a turnover, that situation should be added to the list of possession-changing events. And when the offensive team does win the jump, my assumption would be that that doesn't start a new possession.
Also, there can be possessions that take zero seconds off the clock - Team A makes a basket, Team B fails to inbound within 5 seconds, Team A gets the ball out of bounds.
Team rebounds are another source of confusion when trying to get from what's in the boxscore to actual possessions. Some team rebounds should be considered possession-changing events.
I'm sure I'm missing some stuff but that's the basic definition I would use.
Joined: 30 Dec 2004 Posts: 410 Location: Near Philadelphia, PA
Posted: Fri Sep 29, 2006 4:02 pm Post subject:
John Quincy wrote:
There have been some discussions here about the definition of a possession before. I think getting a black-and-white agreed upon definition is important, especially if one ever hopes to have the NBA's official scorers record possessions in boxscores.
A possession starts when one team gains possession of the ball and ends when the other team gains possession. Basically one possession ends and a new one starts upon one of these events:
- made shot where player is not fouled
- made final free throw (i.e. not the first of two or a technical FT where the team that shoots it keeps possession)
- defensive rebound
- turnover
Events that don't end a possession:
- non-shooting foul (when not in bonus)
- technical foul on the defense
- kicked ball
Or a T on the offense that gives the D foul shots before going back to the O possession.
John Quincy wrote:
I think in some thread there's a discussion of the situation where Team A makes a shot and is fouled, misses the free throw, gets the offensive rebound, and scores again. I think most would consider that to be one possession even though two field goals were made.
Yes, one possession.
John Quincy wrote:
Some questions I have deal with jump balls. First, does anyone know if a team is credited with a turnover if they are on offense, get tied up, and lose the jump? If that doesn't count as a turnover, that situation should be added to the list of possession-changing events. And when the offensive team does win the jump, my assumption would be that that doesn't start a new possession.
That's a turnover and end of possession.
John Quincy wrote:
Also, there can be possessions that take zero seconds off the clock - Team A makes a basket, Team B fails to inbound within 5 seconds, Team A gets the ball out of bounds.
I'm sure I'm missing some stuff but that's the basic definition I would use.
The tough part is deciding what to do at end of quarters. If a team dribbles down and accidentally runs out the clock before the end of the quarter, is that a possession? I would say yes. If a team gets the ball with 0.1 on the clock and doesn't shoot before end of Q, is that a possession? Seems like a no, but technically it should be a yes. Or both could be a no.
Generally, do we define a possession by its end or its start? That is the only question I see. _________________ Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
The tough part is deciding what to do at end of quarters. If a team dribbles down and accidentally runs out the clock before the end of the quarter, is that a possession? I would say yes. If a team gets the ball with 0.1 on the clock and doesn't shoot before end of Q, is that a possession? Seems like a no, but technically it should be a yes. Or both could be a no.
I would definitely say yes in both cases. Saying no would lead to some weird stuff - for those seconds neither team would be considered to be in possession of the ball. It seems like if you have the ball and could potentially score (or turn it over, or whatever), that's a possession, even if it's only for a second.
Or a T on the offense that gives the D foul shots before going back to the O possession.
This is a weird situation. Unless you say that the D team going to the foul line counts as a possession, they could score points without ever technically having possession.
Joined: 30 Dec 2004 Posts: 410 Location: Near Philadelphia, PA
Posted: Fri Sep 29, 2006 4:22 pm Post subject:
John Quincy wrote:
HoopStudies wrote:
Or a T on the offense that gives the D foul shots before going back to the O possession.
This is a weird situation. Unless you say that the D team going to the foul line counts as a possession, they could score points without ever technically having possession.
It is weird and it is rare. But I decided a long time ago that this is the best way to do it, scoring points without having possession. _________________ Dean Oliver
Author, Basketball on Paper
http://www.basketballonpaper.com
I wanted to investigate further why Kobe's On/Off defensive impact last season didn't look too good.
Many Laker fans have pointed out that the main reason for this is opponents faced. Kobe is usually on the floor when the best offensive lineup for the opposing team is playing, and he rests when the opposing team's best offensive units are not on the floor. Makes sense, but up to now I haven't been able to confirm this statistically.
Thanks to the data from this website, I can now confirm this with a little work.
I used this information to determine the strength of the opposition faced when Kobe was on the floor versus off the floor. I essentially took a weighted average of the offensive efficiency of the opposing 5-man units (weighted by number of minutes they played against the Lakers).
Here's a breakdown for the Lakers. Note that the possession calculation is slightly different than 82games.com, so points per 100 possessions doesn't quite match.
Key:
PA_48m: points allowed per 48 minutes
Opp48: opposition's points scored per 48 minutes throughout season
PA_100P: points allowed per 100 possessions
Opp100P: opposition's points scored per 100 possessions throughout season
Code:
Lakers
PA_48m Opp48m PA_100P Opp100p
on 97.82 96.72 102.68 102.63
off 87.54 89.24 95.73 95.63
Total 96.08 95.46 101.55 101.44
League average points per 48m: 96.16
League average points per 100P: 101.94
You can see that the opponents faced with Kobe on the court is over 7 points per 48 minutes better offensively compared to when he's off the court. That's a big part of why his On/Off defensive numbers look so poor -- he has to face much better players. From a per possession perspective, it is evident that the Lakers gave up about as many points as would be expected given the opposition with or without Kobe on the floor. Could it be that Kobe had to play with worse defensive teammates? Perhaps. But from this, it appears that his On/Off impact on defense over the whole season was essentially negligible.
I did a similar break down for Dwyane Wade on the Miami Heat, for comparison:
Code:
Heat
PA_48m Opp48m PA_100P Opp100p
on 95.50 95.81 100.24 101.74
off 96.10 93.32 101.85 98.71
Total 95.66 95.14 100.66 100.92
League average points per 48m: 96.16
League average points per 100P: 101.94
You can see that there isn't as drastic a difference in opponent faced for the Heat; still, opposition Wade faced was 2.5 points per 48 minutes better offensively that the opposition faced when Wade was sitting. This makes Wade's On/Off defensive impact look more impressive. The Heat gave up less points per 100 possessions with Wade on the floor, and they did so facing better offensive units.
Basically one possession ends and a new one starts upon one of these events:
- made shot where player is not fouled
- made final free throw (i.e. not the first of two or a technical FT where the team that shoots it keeps possession)
- defensive rebound
- turnover
The clock expiring for the quarter should also be on the list.
Quote:
Events that don't end a possession:
- non-shooting foul (when not in bonus)
- technical foul on the defense
- kicked ball
To be thorough, I'd also add:
- missed shot with an offensive rebound
All times are GMT - 5 Hours Goto page 1, 2, 3, 4Next
Page 1 of 4
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
