APBRmetrics

I am looking to create and maintain a stats database that pretty much accounts for most of the advanced statistics that people use now and use that information to tweak/create something new. I just finished an advanced statistics course for my MBA program and I have a few questions.

Besides excel, what are the programs people use to make their databases? I don't have the means to purchase a SQL license but do have part-time access to Synergy currently.

Which stats do people think NBA teams value most?

Is there a place where I can get the stats for past NBA seasons?

Any help would be much appreciated.

Some SQL servers are free, like this one from Microsoft:

http://www.microsoft.com/web/platform/database.aspx

You also might want to try MongoDB, which is open-sourced. I believe it's what Evan Z uses for NBA Wowy.

Thanks...I will look into both of those options. I look forward to the challenge.

tlubinski3 wrote:I am looking to create and maintain a stats database that pretty much accounts for most of the advanced statistics that people use now and use that information to tweak/create something new.

Which stats do people think NBA teams value most?

Is there a place where I can get the stats for past NBA seasons?

Any help would be much appreciated.

I dunno which stats NBA teams value most. Are you targeting them as the main or only audience? That could be a different job than providing more convenient, organized access to a fairly comprehensive set of existing stats for players and teams to the public audience interested in analytics. Might start with the later than contemplate the former task later?

I would probably have more ideas later after seeing what you come up with as a template but at minimum from my perspective it would be nice to see metrics like RAPM, IPV, EZPM, ASPM, winshares/48 and others for players all in one spot and side by side. Averaging these and preparing ratios of them with each other is something I think would be interesting and somewhat useful if it was easier.

Check out what draftexpress has for NBA stats. http://www.draftexpress.com/stats.php?league=NBA There are possible additional ratios of interest for things related to shot distribution and other things like block / personal fouls, etc.

Crow wrote:

tlubinski3 wrote:I am looking to create and maintain a stats database that pretty much accounts for most of the advanced statistics that people use now and use that information to tweak/create something new.

Which stats do people think NBA teams value most?

Is there a place where I can get the stats for past NBA seasons?

Any help would be much appreciated.

I dunno which stats NBA teams value most. Are you targeting them as the main or only audience? That could be a different job than providing more convenient, organized access to a fairly comprehensive set of existing stats for players and teams to the public audience interested in analytics. Might start with the later than contemplate the former task later?

I would probably have more ideas later after seeing what you come up with as a template but at minimum from my perspective it would be nice to see metrics like RAPM, IPV, EZPM, ASPM, winshares/48 and others for players all in one spot and side by side. Averaging these and preparing ratios of them with each other is something I think would be interesting and somewhat useful if it was easier.

Check out what draftexpress has for NBA stats. http://www.draftexpress.com/stats.php?league=NBA There are possible additional ratios of interest for things related to shot distribution and other things like block / personal fouls, etc.

I have access to the full version of Synergy Sports Tech which has tons of different stats and rankings. I am not trying to target the NBA as the only audience, but I would like to focus my energy on some of the stats I know people look at/value most when looking at this stuff. Ideally, I would like to set this up on MySQL and find a XML feed of present and historical stats. The more I think about this, the more difficult and less enthusiastic I feel about it haha.

Does anybody know if someone has created a system to hypothesis test or find the optimal number of minutes a player should play before they start to "hurt" the team or something like that. For example, if player X averages 13 minutes a game he puts up certain stats. However, if X played 15 minutes a game, he should put up this amount of stats while the player he subs in for, Y, those two minutes decreases Y's stats by a certain amount.

I don't recall someone specifically looking at the optimal number of minutes a player should play before they start to "hurt" the team but Mike G has written here in the past about small minute players tend generally to improve with more time for awhile.

It may be that most players decline after the 25-30-35th minute or whatever but I don't immediately recall for sure seeing anything that was comprehensive. Basketball On Paper by Dean Oliver has skill curves on this topic but I don't recall if there were overall summaries. Maybe DSMOK1 has something on this topic as well. http://godismyjudgeok.com/DStats/ Perhaps someone else remembers better.

One simple additional approach could look at the split data at hoopsstats.com and compile who improved / declined across the league as players move up in minutes played in a game but this is biased by natural selection.

There was something a few years ago that also showed that lineups tended to improve performance with more minutes of use, from greater familiarity and / or natural selection of the better lineups.


The topic of creating a new advanced stats database in general has come up here several times including recently. viewtopic.php?f=2&t=8224 You can check the thread list and read more if you want. I think a lot of people are not likely to leap forward and discuss the idea much because there have been several false starts or no starts really. If something really got going, some people might jump in and help. Might.
Some are more interested in public collaboration than others. Some major efforts may be underway privately and for profit or potential profit.