Page 14 of 18
Re: Predictions 2014-2015
Posted: Mon Jan 19, 2015 8:16 pm
by Mike G
Posting because in just 4 days, everyone's errors have shrunk (relative to the b-r.com Forecast) by quite a bit.
Absolute error today (Jan 19), as of Jan. 15, and relative to last season's Pythagorean.
Code: Select all
error proj 1-15 -14py error proj 1-15 -14py
AJb1 6.08 6.42 2.53 ncs 7.25 7.64 1.36
crow 6.15 6.46 2.46 itca 7.40 7.78 1.21
myst 6.25 6.71 2.36 eW 7.41 7.63 1.21
atc 6.46 6.87 2.15 AJb2 7.49 7.67 1.13
HDon 6.47 7.04 2.15 DrP 7.49 7.82 1.12
bbs 6.57 7.06 2.04 snd1 7.77 8.19 .84
v-0 6.65 6.98 1.97 fpli 7.82 8.18 .79
bobb 7.11 7.43 1.51 14py 8.61 8.90 .0
Note that the 2013-14 Pyth "prediction" improved by 0.29 in the interval. A few of us improved by less than that.
AJb1 also leads (by .03) in RMSE.
Re: Predictions 2014-2015
Posted: Mon Jan 19, 2015 8:45 pm
by Crow
If not too much trouble a chart showing everyone's team by team win predictions vs the different predicted win estimates (by method and blended) would be great as we hit the half way point.
Re: Predictions 2014-2015
Posted: Mon Jan 19, 2015 9:46 pm
by AcrossTheCourt
Here's a better projection, I think:
http://fivethirtyeight.com/datalab/intr ... yoff-odds/
One problem is that it'll help anyone who uses RPM, but at least it's not a "dumb" rating system that doesn't understand OKC has Durant back.
Re: Predictions 2014-2015
Posted: Sun Jan 25, 2015 2:24 pm
by Mike G
permaximum wrote:Mike G wrote:Is 26 mpg (x82) for Kobe more optimistic than most are guessing?
His (arbitrary) 1.33 eW/484 would be lowest since his 2nd season (when he also went 26 mpg).
I bet everything I have on Kobe that he will play more than 26 mpg. I don't have to look at any statistics or metrics to see that.
Kobe certainly started out with more minutes.
Code: Select all
mo. G MPG Min dnp tG tMin tMPG
O-Nov 17 36.0 612 0 17 612 36.0
Dec 12 34.2 410 3 32 1022 32.0
Jan 6 30.9 185 6 44 1208 27.5
Just 3 more DNP and he'll have dropped below 26 mpg for the season.
Yet, even at 27.5 mpg to date from Kobe, the Lakes are well worse than I predicted. Returning players, this season on the left, last year on the right:
Code: Select all
eW+ per36 rates Eff% Sco Reb Ast e484 '14 Eff% Sco Reb Ast e484
.25 Ronnie Price .465 7.8 2.6 5.3 .30 Orl .376 5.8 4.0 5.5 .16
.09 Wayne Ellington .537 12.8 4.9 1.9 .53 Dal .551 13.3 3.8 1.7 .47
.07 Wesley Johnson .531 11.3 4.6 1.9 .46 .518 11.0 5.1 1.9 .43
-.01 Nick Young .528 20.3 3.4 1.3 .97 .555 22.1 2.9 1.6 .97
-.11 Ed Davis .584 11.9 10.4 1.4 .99 Mem .534 14.8 10.7 .9 1.04
-.16 Xavier Henry .400 6.8 1.6 1.1 -.33 .498 15.5 3.9 1.8 .56
-.19 Robert Sacre .449 8.2 6.6 1.3 .32 .504 10.5 7.3 1.5 .50
-.22 Ryan Kelly .481 7.7 6.8 1.2 .12 .538 12.6 5.3 2.3 .61
-.30 Jeremy Lin .532 14.3 3.7 5.7 .82 Hou .559 17.1 3.4 5.7 .95
-.32 Jordan Hill .513 15.2 10.1 1.7 1.14 .569 16.8 11.4 1.2 1.27
-.59 Carlos Boozer .530 17.1 9.9 1.7 1.23 Chi .484 20.4 11.7 2.1 1.50
-.76 Kobe Bryant .469 22.0 5.9 5.4 1.51 .559 26.6 5.3 5.6 1.82
Their 3 best players also lead them in
declines. That's Kobe's 2013 line, and his e484 for this year is
better than my prediction.
Jordan Hill is 27; minutes up from 20 to 28, all other numbers down.
J Lin at 26 also with dropoffs across the board.
R Kelly, 23, almost disappears in his 2nd season.
25 yo Sacre has gone from passable to not really.
...
Boozer, 33, has not declined according to BPM and WS/48; PER says he's up (14.4 to 16.3).
Still he's no Pau, nor even the Pau of last year.
Re: Predictions 2014-2015
Posted: Mon Jan 26, 2015 5:17 am
by sndesai1
sorry for the lack of updates...been having some hdd issues
added in the 2 other projections requested...arturo's isn't doing too well (though who am i to talk? lol) but the other one is actually right near the top
the "average" entry is creeping up
the vegas entry isn't doing so hot at the moment, but it's not totally out of the game
Re: Predictions 2014-2015
Posted: Mon Jan 26, 2015 5:48 am
by Crow
Thank you.
Re: Predictions 2014-2015
Posted: Tue Jan 27, 2015 11:45 am
by permaximum
Mike G wrote:permaximum wrote:Mike G wrote:Is 26 mpg (x82) for Kobe more optimistic than most are guessing?
His (arbitrary) 1.33 eW/484 would be lowest since his 2nd season (when he also went 26 mpg).
I bet everything I have on Kobe that he will play more than 26 mpg. I don't have to look at any statistics or metrics to see that.
Kobe certainly started out with more minutes.
Code: Select all
mo. G MPG Min dnp tG tMin tMPG
O-Nov 17 36.0 612 0 17 612 36.0
Dec 12 34.2 410 3 32 1022 32.0
Jan 6 30.9 185 6 44 1208 27.5
Just 3 more DNP and he'll have dropped below 26 mpg for the season.
Yet, even at 27.5 mpg to date from Kobe, the Lakes are well worse than I predicted.
There was a misunderstanding on my part. I thought you meant the usual MPG. In this case I would bet everything against it. The sitiuation with these kind of guys they rather sit instead of playing 20-25 minutes.
BTW if Kobe even plays the next season his career decline will be one of the soft ones. He's played only 1284 minutes in his decline so far and I don't think that number wil surpass 3000.
Re: Predictions 2014-2015
Posted: Wed Jan 28, 2015 1:12 pm
by Mike G
As requested by Crow, here are teams ranked by how much "we" have underestimated them (first column); predictions sorted (left to right) by absolute error (bottom row).
Code: Select all
avEr tm 14py AJb1 crow HDon myst atc bbs v-0 ncs itca eW DrP bobb fpli AJb2 snd1
17.8 Atl 22 17 19 18 18 20 20 16 20 16 18 17 21 18 18 16
16.8 Mil 23 13 18 23 16 18 18 17 18 12 16 7 18 21 14 20
10.6 GSW 11 11 9 6 11 11 8 17 7 9 19 10 10 9 16 9
6.0 Was 6 5 4 4 6 9 4 8 3 9 3 10 12 2 9 5
5.5 Bos 6 3 7 5 5 7 6 5 4 3 1 15 9 13 2 9
5.5 Por 2 4 6 6 9 1 8 10 6 3 6 4 3 3 9 9
5.5 Tor 3 6 6 3 6 4 6 4 5 8 7 8 1 5 7 8
5.1 Mem 9 4 4 5 6 7 2 11 4 8 11 2 9 3 8 3
5.1 NOP 10 7 5 8 5 3 4 4 8 8 1 6 8 2 6 11
4.6 Uta 10 3 4 5 3 6 3 0 9 6 0 3 5 12 3 11
3.6 Hou 0 1 4 4 6 4 6 6 3 1 8 0 2 2 4 3
3.5 Sac 1 9 3 5 3 3 4 1 9 3 2 6 3 7 1 5
1.2 Orl 1 4 1 2 3 1 0 0 3 9 0 9 5 6 1 0
0.6 LAL 0 1 1 1 2 1 5 6 6 7 9 5 1 0 6 2
0.5 Chi 4 5 1 4 0 5 3 2 1 5 8 0 1 7 11 3
avEr tm 14py AJb1 crow HDon myst atc bbs v-0 ncs itca eW DrP bobb fpli AJb2 snd1
0.3 Phl 0 2 2 0 2 1 3 0 1 3 4 3 3 2 0 2
-.2 Dal 4 1 1 3 2 2 3 2 7 3 8 1 8 5 2 2
-.3 Det 2 3 1 3 2 2 4 3 4 7 4 7 1 5 9 3
-1.5 Phx 4 1 2 1 2 5 4 0 1 8 2 6 9 2 2 1
-3.1 LAC 4 1 1 2 0 2 2 5 3 8 6 5 7 6 1 1
-3.3 Den 1 1 5 1 1 7 3 5 4 7 7 11 8 3 8 1
-4.1 Brk 6 1 8 3 6 0 4 7 4 6 1 6 1 7 5 12
-5.0 Ind 21 10 4 1 5 2 8 2 8 1 2 9 7 5 6 8
-5.1 Cha 5 2 7 7 5 4 4 12 2 10 8 3 10 9 1 2
-7.3 SAS 11 2 7 10 5 8 6 5 9 9 6 12 4 13 7 11
-8.0 Mia 17 14 11 9 10 7 9 3 13 4 0 15 3 6 4 15
-11.1 Okl 14 8 7 14 10 11 12 9 13 11 13 12 8 13 15 14
-12.8 Cle 14 10 11 14 12 14 14 14 13 8 15 9 19 14 16 10
-14.5 NYK 19 12 14 4 14 15 11 15 10 17 18 19 18 16 20 16
-15.9 Min 31 19 14 22 20 14 18 15 20 13 20 5 11 11 18 21
0.01 avg 8.7 5.9 6.1 6.3 6.4 6.5 6.6 6.9 7.2 7.2 7.4 7.4 7.4 7.5 7.5 7.7
avEr tm 14py AJb1 crow HDon myst atc bbs v-0 ncs itca eW DrP bobb fpli AJb2 snd1
Errors are relative to:
http://www.basketball-reference.com/fri ... f_prob.cgi
Re: Predictions 2014-2015
Posted: Wed Jan 28, 2015 3:32 pm
by Crow
Thanks Mike.
Re: Predictions 2014-2015
Posted: Tue Feb 03, 2015 1:04 pm
by Mike G
I don't know what it's called to average the
square roots of (absolute) error, and then square that number -- it's the opposite effect of RMSE-ing them -- but the upshot is interesting:
Code: Select all
(avg sqrt)^2 14py (avg sqrt)^2 14py
AJb1 4.8 1.89 ncs 6.0 .73
HDon 4.9 1.76 bobb 6.0 .71
crow 5.0 1.66 snd1 6.3 .38
myst 5.3 1.39 fpli 6.3 .36
atc 5.4 1.27 DrP 6.4 .31
bbs 5.8 .93 itca 6.5 .22
v-0 5.9 .83 14py 6.7 .00
eW 5.9 .78 AJb2 6.7 -.02
This is at about 59% of the season.
AJ has the lead, and he's also worse than "last year's pythagorean".
Call it SRME. Instead of exaggerating big misses, it minimizes them. Effectively giving more credit for direct hits.
Re: Predictions 2014-2015
Posted: Tue Feb 03, 2015 6:44 pm
by Mike G
The Bucks are projecting from 10 (Dr.Pos) to 26 games better than our guesses. Avg error = 19. How did we get them so wrong? They don't have a lot of new guys, and nobody who was expected to be an impact player.
On the left of the name are my guesses at their minutes this year, last year's rates and eWins/484, and subsequent expected eW.
On the right are this year's rates and mpg, and eWins projected to 82 games (after 47 so far).
Code: Select all
mpg? Eff% Sco Reb Ast Stl e484 eW? per36 Min Eff% Sco Reb Ast Stl e484 e82
26 .515 20 4 5 1.0 1.14 5.0 Knight 32 .549 22 5 6 1.9 1.53 8.3
16 .506 10 7 2 1.1 .41 1.1 Antetokounmpo 29 .541 16 8 3 1.1 1.11 5.4
20 .512 17 3 3 .7 .59 2.0 Mayo 24 .524 18 4 4 1.0 1.01 4.2
26 .537 15 5 2 1.2 .66 2.9 Middleton 24 .565 16 6 2 2.0 1.00 4.1
10 .537 11 3 2 .9 .30 .5 Dudley 23 .606 14 5 2 1.4 .78 3.0
14 .493 11 10 3 1.2 .78 1.9 Pachulia 21 .490 13 10 4 1.4 1.06 3.7
16 .493 15 3 4 1.3 .67 1.8 Bayless 21 .558 15 4 5 1.3 .82 2.9
20 .91 3.1 Parker 16 .520 16 7 2 1.5 .96 2.6
18 .470 11 11 1 1.1 .82 2.5 Sanders 12 .500 13 11 1 1.6 1.14 2.4
22 .535 16 10 2 .8 1.16 4.3 Henson 11 .599 16 9 2 .6 1.26 2.4
20 .480 15 9 2 1.1 .90 3.1 Ilyasova 10 .534 20 8 1 1.0 1.26 2.1
18 .492 9 3 10 1.1 .75 2.3 Marshall 9 .559 11 2 7 1.7 .60 .9
0 .0 O'Bryant 5 .407 10 5 1 .4 .20 .2
14 .480 11 4 5 1.0 .60 1.4 Wolters 3 .379 6 4 2 1.2 -.04 .0
. KMart 2 .421 6 7 2 1.6 .36 .1
242 32.1 242 42.3
Virtually every player is shooting better and scoring more. Almost all are rebounding and assisting as much or more. And everyone has been stealing like crazy.
Blocks and turnovers are omitted for brevity. TO have been worse, but the margin vs opponents is still favorable.
I see 9 guys who could start for most teams in the league; not counting Parker.
Re: Predictions 2014-2015
Posted: Tue Feb 03, 2015 6:59 pm
by J.E.
I think several things are at play for the Bucks
1. The Parker injury actually helped their (soon-to-be) win total. He may be great one day but rookies are generally not good in, well, their rookie year. He'd have taken minutes away from Dudley and Middleton
2. I remember Dudley saying in an interview that he played hurt pretty much all year for LAC. Before that season he had some good-to-decent years in PHX (at least in +/-).
3. Improvement due to age. The 3 players with highest MP on their roster are 23, 20 and 23 years old, resp. The alphabet, in particular, should have been expected to make a jump due to him being 19 last season
Re: Predictions 2014-2015
Posted: Tue Feb 03, 2015 7:46 pm
by Mike G
1 - It may turn out to not hurt much that Parker was lost. As noted, they are 2 deep at every position.
2 - Dudley, Ilyasova, Mayo and others had off years last year. So it's still a freakish event that they all bounce back this year.
3 - At 24.7 years, they're the 5th- youngest team in the league. But even younger teams -- Philly, Utah, Minny, Orlando -- aren't seeing most of their players improve a lot, are they?
Knight and Middleton see their numbers jump at age 23.
The team went from 3rd worst to 3rd best in steals this year.
Re: Predictions 2014-2015
Posted: Tue Feb 03, 2015 9:17 pm
by J.E.
MIN only has 2 players younger than 25 on their roster that aren't currently rookies: Muhammad and Bennett. Muhammad seems to have improved significantly in pretty much every metric, Bennett in every metric that's on bbr, but ironically not in my own (RPM)
These two are 6th and 7th in MP for MIN
Re: Predictions 2014-2015
Posted: Wed Feb 04, 2015 1:53 pm
by Mike G
Speaking of b-r.com, for the last couple of years at least, they've had something like 'Projected Performance' (for the upcoming season) on player pages in the preseason. Seems one might have to sift thru every player page to gather these. Has anyone done that, who wants to say how it's done most practically, or whether it's worth much?
An 'aging' factor would be useful. But a 'bounceback' factor -- for younger players who've just had a down year -- would also be good. This Bucks team seems to have gathered a bunch of such players this season.
Are those b-r.com player projections still available somehow?