Re: Major Update of WINVAL Ratings Through 2003-04
- Dear DanD and JohnH:
First of all, in comparing the game with an 11 point margin with 1
minute left in the third versus a game with a 4 point margin with 11
minutes left in the fourth, here is what my formula comes up with.
11 point margin, 1 minute left in third: 9.0
4 point margin, 11 minutes left in fourth: 10.8
After I normalize the weights so that the average weight in the
fourth quarter is the same as the rest of the game, here is what I
end up with (with the average weight being 1).
11 point margin, 1 minute left in third: 1.00
4 point margin, 11 minutes left in fourth: 1.29
So I am counting latter situation much heavier.
And on the matter of not counting garbage time at all. I could set
a minimum weight so that every observation is counted, but then I
suspect there are going to be a lot of people complain about me
measuring a bunch of garbage time phenoms. It is a bit of a no-win
But this is not a trivial issue. Here is the distribution of the
weights. So 9.2 percent of the minutes played (4.5 minutes per
game) are garbage minutes that according to my weighting system get
zero weight. Seventy percent of the zero weight minutes take place
in the fourth quarter, where they make up 28.3 percent of minutes
Range Percent Cumulative Percent
0 9.2 9.2
0.0-0.1 0.3 9.5
0.1-0.2 0.9 10.4
0.2-0.3 1.2 11.6
0.3-0.4 1.2 12.8
0.4-0.5 1.5 14.3
0.5-0.6 1.7 16.1
0.6-0.7 1.8 17.9
0.7-0.8 2.2 20.1
0.8-0.9 2.6 22.7
0.9-1.0 0.5 23.2
1.0-1.1 3.0 26.2
1.1-1.2 62.5 88.8
1.2-1.3 0.8 89.5
1.3-1.4 1.4 91.0
1.4-1.5 1.3 92.3
1.5-1.6 1.4 93.7
1.6-1.7 1.3 95.0
1.7-1.8 1.2 96.2
1.8-1.9 1.3 97.5
1.9-2.0 0.9 98.4
2.0-2.1 0.6 98.9
2.1-2.2 0.4 99.3
2.2-2.3 0.2 99.5
2.3-2.4 0.5 100.0
--- In APBR_analysis@yahoogroups.com, "Daniel Dickey"
> >From: "dan_t_rosenbaum" <rosenbaum@u...>
> >Reply-To: APBR_analysis@yahoogroups.com
> >To: APBR_analysis@yahoogroups.com
> >Subject: [APBR_analysis] Re: Major Update of WINVAL Ratings
> >Date: Thu, 29 Apr 2004 18:02:03 -0000the
> >Here is my exact code. Clock measures the minutes elapsed in the
> >game at the beginning of the observation. Three minutes left in
> >game (in regulation or in overtime) is counted as 45. Margin isthe
> >absolute value of the difference in scores at the beginning of theany
> >Basically, in the first three quarters, full weight is given to
> >part of a game where the margin is less than 10 and no weight isphased
> >given if it is more than 20. Between 10 and 20, the weight is
> >from full to zero.except
> >This is basically what happens in the fourth quarter as well,
> >that I decrease the margin from 10 (20) to 3 (6) from thebeginning
> >to the end of the fourth quarter. Also, ceteris paribus, the endof
> >the quarter counts more than the beginning of the quarter.the
> >At the end of all of this, I renormalize the weights so that on
> >average minutes in the fourth quarter count the same as those in
> >first three quarters.value than a
> So a 4 point lead with 11 minutes left in the game has "less"
> 11 point lead with 13 minutes left in the game? Of course - Ivery well
> could be reading this wrong.bit - EVEN
> I also tend to think that EVERY minute should count just a little
> if it's in a blowout. These garbage players aren't going to nottry (if
> they wanna play in the future) - and it doesn't seem quite fairthat a guy
> could come in and play extremely well for 5 minutes and not haveit count at
> all. I mean - I know that garbage time had no true value to theteam
> win/loss wise - but doesn't it show a little bit the value of theplayer
> Maybe this is part of the Frahm situation? He has so few minutes -
> probably most of them are garbage time. Maybe his poorer playjust happened
> during this time - and his better play happened the few minutes(percentage
> wise) that counted?today -
> Express yourself with the new version of MSN Messenger! Download
- --- In APBR_analysis@yahoogroups.com, "Kevin Pelton" <kpelton08@h...>
> > Maybe this is part of the Frahm situation? He has so few minutesany
> > and probably most of them are garbage time. Maybe his poorer play
> > just happened during this time - and his better play happened the
> > few minutes (percentage wise) that counted?
> Most people have made the opposite argument -- that Frahm's good
> unadjusted plus-minus (+10.0 Roland Rating, almost twice that of
> other Sonics player save the four-minute effort of Leon Smith) ison
> largely due to him playing in garbage time and leading the Sonics
> runs along the lines of the 12/10 Boston game I mentionedpreviously
> and the New Jersey game the night before.I am not sure that true . .I think the cause and effect is when Frahm
> I was genuinely surprised to see Frahm rate so extraordinarily well
> in DanVAL given the "clutch" adjustment.
> The Sonics simply have played well with Frahm in the game,
> regardless the circumstances.
plays well (read shoots well) he get to stay in the game. When he
doesn't shoot well . .the plays only a couple of minutes and sits. I
think he probably has 8 or so good games that skew his results (not
sure if skew is the right word, but indicate a greater value of his
play than otherwise would be true).
> Dan -Well, obviously that message was supposed to only go to DanR, but I
> The article is now live at Hoopsworld.
humbly think the column is worth pimping here anyway, especially
since Stuart already hit on the Holy Grail metaphor:
The latest from "Page 23" at Hoopsworld.com by Kevin Pelton:
The Last NBA Crusade
If you've seen the third and final Indiana Jones movie, "Indiana
Jones and the Last Crusade", you probably recall what the hero
encounters on his way to find the Holy Grail -- skulls and skeletons
of those who have gone before him, also searching for the Holy Grail
only to give their lives in the chase.
Things aren't quite as dire when it comes to the search for the Holy
Grail of rating systems by basketball's statistical analysts, but
there are plenty of skeletons drying in the sun if you look through
the history of such efforts.
Over there is TENDEX, Doug Heeren's revolutionary first strike for
using player statistics in a comprehensive rating form. Beside it
lies Manley Credits, the simplification of TENDEX with no
complicated pace factor to complicate, now resurrected as
NBA.com's " Efficiency Rating System".
Scattered throughout are other linear-weights formulas, with various
values for blocks, steals, and whatever else. Squint hard enough and
you might even find my own VORP system.
None of these efforts, however, has succeeded in its quest to
provide one rating system that's all we need to evaluate players.
Dean Oliver, the author of Basketball on Paper, thinks there's a
good reason for that -- the Holy Grail, at least the NBA version, is
apocryphal, a myth blindly chased despite the fact that it does not
Oliver has won me over as a convert. No single rating system can
ever provide enough subtlety, enough nuance, that it will be all we
need. A balanced perspective, one that considers multiple rating
systems as well as personal observation, is the only way to go.
But that doesn't mean we can't make better rating systems. . . .
Read the rest of the column:
>From: "dan_t_rosenbaum" <rosenbaum@...>OK- cool - that makes sense.
>Subject: [APBR_analysis] Re: Major Update of WINVAL Ratings Through 2003-04
>Date: Mon, 03 May 2004 02:42:40 -0000
>Dear DanD and JohnH:
>First of all, in comparing the game with an 11 point margin with 1
>minute left in the third versus a game with a 4 point margin with 11
>minutes left in the fourth, here is what my formula comes up with.
>11 point margin, 1 minute left in third: 9.0
>4 point margin, 11 minutes left in fourth: 10.8
>After I normalize the weights so that the average weight in theCount me as one that WOULD NOT complain about measuring garbage time. You
>fourth quarter is the same as the rest of the game, here is what I
>end up with (with the average weight being 1).
>11 point margin, 1 minute left in third: 1.00
>4 point margin, 11 minutes left in fourth: 1.29
>So I am counting latter situation much heavier.
>And on the matter of not counting garbage time at all. I could set
>a minimum weight so that every observation is counted, but then I
>suspect there are going to be a lot of people complain about me
>measuring a bunch of garbage time phenoms. It is a bit of a no-win
obviously would have it weighted quite a bit less. PLUS you are already
taking into account opposing lineups and other teamates on the court, etc.
I do think we aren't getting a complete picture of a player that already has
limited minutes to analyze when we ignore much of his performance (because
he mainly is a "garbage time" player). So - I for one feel there should be
a minimum small weight for garbage time.
By the way - it looks like you might be weighting max clutch time at about
twice the value of "normal" time- when it is all said and done - is this
right? That kinda fits how I've envisioned looking at clutch performance -
except I always thought the max having a weight of 2, normal 1, and
"garbage" 0.5. Maybe 3, 1, and 0.33 would be even better. Who knows.
Express yourself with the new version of MSN Messenger! Download today -
it's FREE! http://messenger.msn.com/go/onm00200471ave/direct/01/
Having been one of the defenders of limiting the value of garbage
time credit, count me as one who really doesn't care rather you have
a zero value for garbage time or some minimum value (in other words
you wouldn't hear me screaming if you added a minimum value). While I
agree there should probably be some value for performing well when
you team is down by 25 pts in the fourth, the truth is that if you
are performing well enough to matter, you will play more non-garbage
minutes. . .I think it would be fine to give some value but I don't
think it hurts to give none.
Andre Miller this season
To me, this would seem to bear out that Denver does have better
spacing and that playing with a very active big man like Camby,
instead of low post player like Brand, along with a wing that often
demands a double-team like 'Melo could easily be the reason for
Miller's improvement. I suspect if you charted where he got his
points from, you would find offensive rebounding due to 'Melo double
teams and more focus on Camby, Nene resulted in more flexibity for
Andre to crash the boards (rather than actual drives, in which case
we would expect the close shots to maintain a 56% instead of
improving to 61% although it could just be deviation noise).
BTW: Despite the stats, I think it was a toss-up between
Miller, 'Melo, and Camby for Denver's MVP as all three were clearly
more important than everyone else. Basically against any good team,
Denver needed to have two of those three play well to win.
Furthermore, Miller is probably the best position / technical
rebounder on the team (i.e. he understands when and where to crash
the boards and how to box out better than any other Nugget . .I
beleive those benefits are underated by both the stats and perception