igor eduardo küpfer

http://members.rogers.com/strudel/deleteable/NBA03-04.htm

I'll update it every once in awhile. Has a bunch of stathead stats that you
don't see anywhere. Plus, it has two graphs -- non-ascii graphs, at that. Yes,
this page is in hi-fi!

--

--------------------------------
| best, | Sticking it to |
| ed | The Man since 1971 |
--------------------------------
Watch the spam trap -- the domain is rogers

ron coscorrosa

Cool.

I disagree that the "best teams" are necessarily in the upper-right
quadrant, the best "balanced" teams certainly are, but it's differential
that's important, not the (offensive or defensive) absolutes. I think.

What's the magical exponent you guys are using in the Expected Win%? I
forget!

Anyway, I was thinking of doing that Excel stats thing again[1], but...
there's too much I would want to add to it, and it would take forever.
Also, I'd want to do some gamelog ****yzing stuff, and, while not looking
into it that much, it would probably be insanely difficult (though fun).

Hmm.

[1] http://coscorrosa.com/programs/misc/nbastats/

--
Ron Coscorrosa
http://coscorrosa.com

igor eduardo küpfer

Geometry is obviously not my strong suit. You're right, and I was wrong. What
it should be is a diagonal line bisecting the plane from the top left corner
to the bottom right -- this represents the "average" relationship between
defensive ratings and offensive ratings. the further away from that line, the
better (or worse) the team.

Now, if only I could figger out a way of getting Excel to draw that line...


I went with 14, which worked best for last season. It doesn't make all that much difference.

Yeah, I'm going to maybe bug you to do that thing again at the end of the
season. It came in real handy, that's for sure. There isn't all that much to
add, unless you wanted to strip the numbers from the boxscores instead of the
player logs.

--

--------------------------------
| best, | Sticking it to |
| ed | The Man since 1971 |
--------------------------------
Watch the spam trap -- the domain is rogers

dave

could you just say distance to the upper right corner?

ron coscorrosa

<snip>

Basically, you're looking for the "best-fit" line right?

Unfortunately, statistics and graphs aren't my strong suit, so we're
screwed (the only way I got an A- in statistics was blind stupid luck).
I'm sure it'd be easier to do in a math package than Excel, if it's even
possible to do (easily) within Excel.
<snip>


The ideal thing would be getting the gamelogs, and calculating the stats
from those (then I could get all the normal stuff, plus stupid things like
lineup stats (like that URL you gave earlier), or average passes per
possession, or meta-assists (passes to the person who gets the assist, or
passes to someone who draws a foul and makes a free throw), exact number
of possessions, etc.). I'd store this in a database of course. Then I
could run queries on it, finding league averages in things like average
time before taking a shot per possession, etc. The fun part of this one,
besides the parsing, would be having some natural language parser for
querying...

Barring that, the next ideal way would be to download the boxscores,
record the info in a database permanently (so I'd only have to download at
most 15 pages per day), and then calculate the stats from there (including
defensive stats which I don't have now). Then I could run queries on it,
like top 5 Manley Credit games in November (you know, the relevant
stuff).

The worst case scenario is tweaking the script as-is and making sure it works
this year.

My guess is I'll do the middle one, but I won't have time for a few weeks
to work on it at best, probably longer.

--
Ron Coscorrosa
http://coscorrosa.com

igor eduardo küpfer

Not quite. Here's my theory: there is a relationship between offense and
defense, one that is hard to break. For example, it is relatively easy for a
team to increase its offensive ability -- by sacrificing defense. There are...

[The ellipses is actually a pause for about 24-hours, while I thought about
what I was going to say.]

I just tested my theory, and found it to be full of holes. I won't bother to
spell it out, then.

Well, _I_ wouldn't need it till season's end. I'm prepared to do some legwork
-- downloading the game stats, or something. Something that has nothing to do
with programming. Maybe involving limericks.

--

--------------------------------
| best, | Sticking it to |
| ed | The Man since 1971 |
--------------------------------
Watch the spam trap -- the domain is rogers

ron coscorrosa

<snip relevant context>


Well, on the other end of the spectrum, I didn't really give any thought to
my post here, so, between our extremes, a happy balance exists. Let's
hope we never reach it.

You overestimate our ability to find holes in theories.
<snip NBA stats program thingy>


I've started toying with the database already, and my head's getting ready
to explode. It's a delicate balance between making it so general as to
encapsulate every possibility, or simplifying it and making it easier to
work with.

Of course, I'm going for the overly complex general "head about to
explode" approach, since it's cooler.

The gamelog parsing would take forever, so I'm not going to do that at the
start - if at all - but I'll design it so that if I do do it, it will fit
in with everything else nicely.

I'll give myself 2 to 1 odds against doing it, but who knows!

--
Ron Coscorrosa
http://coscorrosa.com