## Re: replacement level players

Expand Messages
• I ll take this a step further and say that you may be seeing a self- selected sample. There s two possibilities: 1) With marginal players like Scott Williams
Message 1 of 11 , May 6, 2002
I'll take this a step further and say that you may be seeing a self-
selected sample. There's two possibilities:

1) With marginal players like Scott Williams and John Wallace, it
only makes sense to play them when there's a particular matchup
that's in their favor, which is why they look good when they play a
lot of minutes (and have a matchup in their favor), and not so good
when they don't.

2) With a guy like John Wallace, he may get 15+ minutes mostly in
games that are blowouts, where he tends to be playing against the
other teams scrubs, where he thus produces more offensively than one
might reasonably expect.

--- In APBR_analysis@y..., "mikel_ind" <msg_53@h...> wrote:
> --- In APBR_analysis@y..., Ed Weiland <weiland1029@y...> wrote:
> >
> > --- harlanzo <harlanzo@y...> wrote:
> > > I know this has been touched upon, but does anyone
> > > have the
> > > capability to look the stats of a player only when
> > > he plays greater
> > > than 15 or 20 minutes? If we had this data we might
> > > have better idea
> > > of the ability of the guys at the end of the bench
> > > and how readily
> > > replaceable some front-line guys are. For example,
> > > John Wallace
> > > seems like a guy who could surely put up acceptable
> > > numbers with
> > > enough minutes. IF we compared his and others
> > > numbers in games where
> > > they played 15+ minutes I think we might get a
> > > better picture of
> > > their ability.
> >
> > Here's Wallace's log:
> > He played in 10 games where he got over 15 minutes. If
> > my quick math is right, he averaged 21.3 min, with
> > 12.4 pts, 4.7 boards and a .558 FG pct in those games.
> > The Suns' record in those games was 4-6.
>
> Interesting. Since Wallace only averaged 11 minutes for 46 games,
> and shot only .435 FG pct, he must have been Really Bad in those
> other 36 games.
>
> Most likely, he is a streaky shooter, as well as a limited player.
> In other words, when he isn't hitting, he doesn't bring much game
at
> all.
>
> Based on his 11-minute average performance, I'd expect about 12.5
pts
> and 5 reb in 36 minutes, not in 21 minutes.
>
> This might be where coaches earn their money (and why we don't get
> paid for This). If Wallace is on, play him. If not, back to the
> bench.
>
>
> > I looked at Scott Williams numbers in his 20+ minute
> > games last season and found that not only was he a
> > better player in such games, but the Bucks looked like
> > an elite team in those games and a .500 team when
> > Williams played less than 20 minutes.
>
> Again, when your 4th- or 5th-best player has a strong game, you are
> much more likely to win. That doesn't mean playing this guy 30-40
> minutes, regardless, is going to get you the same results.
>
>
> > I didn't think
> > it was much more than a fluke at the time, but in
> > light of Milwaukee's mediocrity this season perhaps
> > Williams was a better fit there than most
> > folks--including Bucks management--realized. At the
> > very least it was a situation they probably didn't
> > need to correct.
>
> Consistency is a much-sought quality in players.
>
>
> > There are some players I'd like to do such a study for
> > this season, one that comes to mind immediately is
> > Keon Clark.
> >
> > Ed Weiland
>
> Keon is intriguing. He has better rebounding and scoring numbers
> than Antonio or Hakeem. Another up-and-down player.
>
> > __________________________________________________
> > Do You Yahoo!?
> > Yahoo! Health - your guide to health and wellness
> > http://health.yahoo.com
• ... Both are excellent points, although implicitly covered by MikeG when he ... [...] ... [...] ... [...] ... Yup, another way to re-state what MikeG and
Message 2 of 11 , May 6, 2002
On Mon, 6 May 2002, alleyoop2 wrote:

> I'll take this a step further and say that you may be seeing a self-
> selected sample. There's two possibilities:
>
> 1) With marginal players like Scott Williams and John Wallace, it
> only makes sense to play them when there's a particular matchup
> that's in their favor, which is why they look good when they play a
> lot of minutes (and have a matchup in their favor), and not so good
> when they don't.
>
> 2) With a guy like John Wallace, he may get 15+ minutes mostly in
> games that are blowouts, where he tends to be playing against the
> other teams scrubs, where he thus produces more offensively than one
> might reasonably expect.

Both are excellent points, although implicitly covered by MikeG when he
wrote:

> --- In APBR_analysis@y..., "mikel_ind" <msg_53@h...> wrote:

[...]

> > Most likely, he is a streaky shooter, as well as a limited player.
> > In other words, when he isn't hitting, he doesn't bring much game
> at
> > all.

[...]

> > This might be where coaches earn their money (and why we don't get
> > paid for This). If Wallace is on, play him. If not, back to the
> > bench.

[...]

> > Again, when your 4th- or 5th-best player has a strong game, you are
> > much more likely to win. That doesn't mean playing this guy 30-40
> > minutes, regardless, is going to get you the same results.

Yup, another way to re-state what MikeG and Alleyoop2 wrote is that we
have to look at cause-and-effect, not just the raw stats and the
correlations in those stats. Sometimes, especially for star players, good
stats ==> victory. But for a 12th man, and sometimes for 4th or 5th men,
the causality will typically be the opposite: victory ==> good stats.

> > Consistency is a much-sought quality in players.

Yeah, those standard deviations again, that we were looking at a couple of
months ago.

--MKT
• ... one ... Yeah. This is also why the relationship you d think that would exist between a team s winning% and good player s minutes doesn t show up in
Message 3 of 11 , May 6, 2002
--- In APBR_analysis@y..., "alleyoop2" <alleyoop2@y...> wrote:
> I'll take this a step further and say that you may be seeing a self-
> selected sample. There's two possibilities:
>
> 1) With marginal players like Scott Williams and John Wallace, it
> only makes sense to play them when there's a particular matchup
> that's in their favor, which is why they look good when they play a
> lot of minutes (and have a matchup in their favor), and not so good
> when they don't.
>
> 2) With a guy like John Wallace, he may get 15+ minutes mostly in
> games that are blowouts, where he tends to be playing against the
> other teams scrubs, where he thus produces more offensively than
one
> might reasonably expect.

Yeah. This is also why the relationship you'd think that would exist
between a team's winning% and good player's minutes doesn't show up
in first-order analyses. I remember looking at Bull games with
Jordan at varying minutes. I think they won more when he played
fewer minutes, basically because there was much more garbage time
when he played <30 minutes (which is also when replacement players
get >20 minutes).

(Personally, I do feel that Williams was important to the Bucks last
year. His playoff numbers last year were quite good. His regular
season numbers somewhat similar. Either way, he did a lot of the
dirty defensive work that no one stepped in to do without him.)

It points at using play-by-plays to analyze players. Only count a
player's contributions when the game is competitive. When is a game
competitive? When each team has at least a 5% chance of winning, I
think. How do we know what a 5% chance of winning is? From play-by-
plays. I've got a bunch of paper pbps, but putting them into a
database is not on my priority list.

DeanO
• On Mon, 6 May 2002, HoopStudies wrote: [...] ... Interesting, 5% strikes me as a bit low, but maybe not. Working backwards, I d guess that a team that is 3
Message 4 of 11 , May 6, 2002
On Mon, 6 May 2002, HoopStudies wrote:

[...]

> It points at using play-by-plays to analyze players. Only count a
> player's contributions when the game is competitive. When is a game
> competitive? When each team has at least a 5% chance of winning, I
> think. How do we know what a 5% chance of winning is? From play-by-
> plays. I've got a bunch of paper pbps, but putting them into a
> database is not on my priority list.

Interesting, 5% strikes me as a bit low, but maybe not.

Working backwards, I'd guess that a team that is 3 points down, withOUT
the ball, with 10 seconds left, has about a 5% chance of winning. Let's
test that hunch (best way to do this would be to work through thousands of
play-by-play sheets, but who's going to do that, so let's do it
theoretically): foul immediately. With a FT% of 70%, there's a 9% chance
the opponent will miss both FTs. Then maybe a 30% chance that the team
can make the 3pter. And presumably about a 50% chance of winning in OT.

Oops, that comes out to 1.35%, far less than 5%.
On the other hand, if they do have the ball, they've got about a 15%
chance of winning. Assuming that 10 seconds is enough time for them to
get a 30% probability 3-pt attempt.

Well so much for my intuition. I was hoping to be able to estimate the
probability of winning when there were say 5 minutes left, and a point
differential of x points, where x was big enough to give one team only a
5% chance of winning. Result: garbage time, under DeanO's definition of
non "competitive" game time. But it is clear that I probably do not have
a good idea of what x is.

A friend of mine assumes a game to be out of reach ("in the refrigerator",
as Chick Hearn would put it, although he put one into the refrigerator
prematurely recently, I think it was the Laker's last regular season game
against Portland) when the point differential is at least 2t+7, where "t"
is the number of minutes left. I haven't done an extensive study, but I
haven't seen it violated yet. E.g. Reggie Miller's epic comeback against
the Knicks was I think 8 points in 25 seconds, not 9 points. (Can we
quantify Reggie's playoff antics? He sure *seems* to have had more
last-second playoff heroics than most other players, even superstar
players.)

Ralph Lawler, the Clippers play-by-play guy, likes to talk about "Lawler's
Law": the team which first reaches 100 points wins the game. (In today's
slowdown NBA, one might even be able to lower the threshold to 90 points.)
I can't recall seeing Lawler's law fail (partly because so few games reach
100 points these days), but I'm sure it does fail occasionally -- perhaps
5% of the time?? Anyway, that might be yet another way of identifying
the 5% games.

--MKT
• ... withOUT ... Let s ... thousands of ... chance ... team ... in OT. ... them to ... I worked through this logic once in the past, looking to calculate odds
Message 5 of 11 , May 6, 2002
>
>
> Interesting, 5% strikes me as a bit low, but maybe not.
>
> Working backwards, I'd guess that a team that is 3 points down,
withOUT
> the ball, with 10 seconds left, has about a 5% chance of winning.
Let's
> test that hunch (best way to do this would be to work through
thousands of
> play-by-play sheets, but who's going to do that, so let's do it
> theoretically): foul immediately. With a FT% of 70%, there's a 9%
chance
> the opponent will miss both FTs. Then maybe a 30% chance that the
team
> can make the 3pter. And presumably about a 50% chance of winning
in OT.
>
> Oops, that comes out to 1.35%, far less than 5%.
> On the other hand, if they do have the ball, they've got about a 15%
> chance of winning. Assuming that 10 seconds is enough time for
them to
> get a 30% probability 3-pt attempt.
>

I worked through this logic once in the past, looking to calculate
odds of winning, but the branching of the tree gets pretty huge.

I'm not sure where the cutoff should be actually. I throw 5% out
there because it also eliminates a fair number of the stats of high
scoring individuals on really bad teams (their teams are out of many
games by halftime). If we end up cutting the last couple minutes out
of a really good player's pool of numbers, I'm not as concerned.
They still have a lot of stats that show they are good.

> A friend of mine assumes a game to be out of reach ("in the
refrigerator",
> as Chick Hearn would put it, although he put one into the
refrigerator
> prematurely recently, I think it was the Laker's last regular
season game
> against Portland) when the point differential is at least 2t+7,
where "t"
> is the number of minutes left. I haven't done an extensive study,
but I
> haven't seen it violated yet. E.g. Reggie Miller's epic comeback
against
> the Knicks was I think 8 points in 25 seconds, not 9 points. (Can
we
> quantify Reggie's playoff antics? He sure *seems* to have had more
> last-second playoff heroics than most other players, even superstar
> players.)
>

Reggie's antics in that game were of the <0.01% variety, I'm
betting. Against NJ the other night, being down 3 with 4 s left and
the ball should be on the order of about 15%, assuming ~50% win% in
OT and 30% on 3 pt shot.

> Ralph Lawler, the Clippers play-by-play guy, likes to talk
> Law": the team which first reaches 100 points wins the game. (In
today's
> slowdown NBA, one might even be able to lower the threshold to 90
points.)
> I can't recall seeing Lawler's law fail (partly because so few
games reach
> 100 points these days), but I'm sure it does fail occasionally --
perhaps
> 5% of the time?? Anyway, that might be yet another way of
identifying
> the 5% games.

We really need a Project Play-By-Play, where we develop a db of play-
by-plays. I am willing to volunteer a bunch of play-by-plays. I can
probably get most of them into electronic format, though some will be
in different formats than others. Who wants to develop the db
structure and import things?

I can post a couple example formats. I like the Printable Play-by-
Plays at NBA.com, but those are limited historically.

DeanO
• I can t agree with you more. This needs to happen. Incidentally, CBS sportsline has the most games with logs for some reason, but nobody has them all from this
Message 6 of 11 , May 9, 2002
I can't agree with you more. This needs to happen. Incidentally, CBS
sportsline has the most games with logs for some reason, but nobody
has them all from this year, and going back further it's just a mess.

>
> I can post a couple example formats. I like the Printable Play-by-
> Plays at NBA.com, but those are limited historically.
>
> DeanO
• ... CBS ... mess. If we can set up APBR as a non-profit org, we can probably get this done by paying a student and writing off those payments as contributions
Message 7 of 11 , May 10, 2002
--- In APBR_analysis@y..., "alleyoop2" <alleyoop2@y...> wrote:
> I can't agree with you more. This needs to happen. Incidentally,
CBS
> sportsline has the most games with logs for some reason, but nobody
> has them all from this year, and going back further it's just a
mess.

If we can set up APBR as a non-profit org, we can probably get this
done by paying a student and writing off those payments as
contributions to a non-profit.

RobertB?
• ... I m willing to look at it and give an estimate of the effort involved. DeanL __________________________________________________ Do You Yahoo!? LAUNCH -
Message 8 of 11 , May 22, 2002
--- HoopStudies <deano@...> wrote:
> --- In APBR_analysis@y..., "Michael K. Tamada"
> >
> We really need a Project Play-By-Play, where we
> develop a db of play-
> by-plays. I am willing to volunteer a bunch of
> play-by-plays. I can
> probably get most of them into electronic format,
> though some will be
> in different formats than others. Who wants to
> develop the db
> structure and import things?
>

I'm willing to look at it and give an estimate of the
effort involved.

DeanL

__________________________________________________
Do You Yahoo!?
LAUNCH - Your Yahoo! Music Experience
http://launch.yahoo.com
Your message has been successfully submitted and would be delivered to recipients shortly.