APBRmetrics

Posted: **Wed Oct 15, 2014 6:42 pm**

Each season there are some players that, maybe unexpectedly, see their RPM/xRAPM rise or fall significantly. I figured it'd be interesting to see who those players were in the last 14 years, and if I/we can maybe identify certain characteristics about them.

I'm not saying these players improved (or got worse) by the stated amount - there's a chance that a player's RPM rating is off from his "true value" in one or both years. Here, I'm just interested in big changes in rating. It's obviously possible, or even likely, that these players did actually improve but we can't be sure

For now, I'm just collecting and looking at the data. Maybe I'll try to predict such occurances in the future, if I have the time

Here are the biggest risers. Only players with >1000 minutes in both years are listed

Code: Select all

╔══════╦════════════════════╦═══════════╦═══════════╦══════╦═════╦════════════════╦════════════════════╦═════════════╗
║ Year ║        Name        ║ New xRAPM ║ old xRAPM ║ Diff ║ Age ║ Year in League ║   Switched Teams   ║ Pick number ║
╠══════╬════════════════════╬═══════════╬═══════════╬══════╬═════╬════════════════╬════════════════════╬═════════════╣
║ 2007 ║  David Lee         ║ 3.1       ║ -2.4      ║ 5.5  ║  23 ║              2 ║                    ║          30 ║
║ 2010 ║  Kevin Durant      ║ 2.6       ║ -2.6      ║ 5.2  ║  21 ║              3 ║                    ║           2 ║
║ 2005 ║  Drew Gooden       ║ -1.6      ║ -6.5      ║ 4.9  ║  23 ║              3 ║ In-between seasons ║           4 ║
║ 2006 ║  Boris Diaw        ║ 3.1       ║ -1.6      ║ 4.7  ║  23 ║              3 ║ In-between seasons ║          21 ║
║ 2008 ║  Caron Butler      ║ 3.2       ║ -1.5      ║ 4.7  ║  27 ║              6 ║                    ║          10 ║
║ 2013 ║  Tristan Thompson  ║ -1.1      ║ -5.8      ║ 4.7  ║  21 ║              2 ║                    ║           4 ║
║ 2005 ║  Marcus Banks      ║ 0.7       ║ -3.9      ║ 4.6  ║  23 ║              2 ║                    ║          13 ║
║ 2006 ║  Dwight Howard     ║ 1.7       ║ -2.9      ║ 4.6  ║  20 ║              2 ║                    ║           1 ║
║ 2010 ║  Dwight Howard     ║ 8         ║ 3.4       ║ 4.6  ║  24 ║              6 ║                    ║           1 ║
║ 2004 ║  Hedo Turkoglu     ║ 4         ║ -0.5      ║ 4.5  ║  24 ║              4 ║ In-between seasons ║          16 ║
║ 2014 ║  Draymond Green    ║ 4.6       ║ 0.1       ║ 4.5  ║  23 ║              2 ║                    ║          35 ║
║ 2014 ║  Goran Dragic      ║ 4.3       ║ -0.2      ║ 4.5  ║  27 ║              6 ║                    ║          45 ║
║ 2002 ║  Bob Sura          ║ 0.7       ║ -3.7      ║ 4.4  ║  28 ║              7 ║                    ║          17 ║
║ 2006 ║  Dwyane Wade       ║ 7.4       ║ 3         ║ 4.4  ║  24 ║              3 ║                    ║           5 ║
║ 2004 ║  Carlos Boozer     ║ -0.2      ║ -4.6      ║ 4.4  ║  22 ║              2 ║                    ║          34 ║
║ 2002 ║  Marcus Fizer      ║ -3.5      ║ -7.8      ║ 4.3  ║  23 ║              2 ║                    ║           4 ║
║ 2011 ║  Mike Conley       ║ 3.9       ║ -0.4      ║ 4.3  ║  23 ║              4 ║                    ║           4 ║
║ 2002 ║  Kenyon Martin     ║ 1.1       ║ -3.1      ║ 4.2  ║  24 ║              2 ║                    ║           1 ║
║ 2012 ║  Paul George       ║ 4.8       ║ 0.6       ║ 4.2  ║  21 ║              2 ║                    ║          10 ║
║ 2014 ║  Marcus Morris     ║ -0.1      ║ -4.2      ║ 4.1  ║  24 ║              3 ║ during 1st season  ║          14 ║
╚══════╩════════════════════╩═══════════╩═══════════╩══════╩═════╩════════════════╩════════════════════╩═════════════╝

"Age" and "Year in League" relate to the year on the left. So, David Lee was 23 in 2006-2007 and that was his 2nd year in the league. Thus, he signficantly improved already from year 1 to year 2.

Average age in this group 23.4. Most players either improve from year 1 to year 2, or from year 2 to year 3. Grabbing some so far (mildly) underwhelming players after their 1st or 2nd year and hoping for a jump in performance seems like a valid strategy.
Pick numbers seem, to me, all over the place. Three players from 2014 on this list, two Suns, including 2014 MIP Goran Dragic

.. and those whose rating took the biggest fall

Code: Select all

╔══════╦════════════════════╦═══════════╦═══════════╦══════╦═════╦════════════════╦════════════════════╦══════════╗
║ Year ║        Name        ║ New xRAPM ║ old xRAPM ║ Diff ║ Age ║ Year in League ║   Switched Teams   ║ Injured  ║
╠══════╬════════════════════╬═══════════╬═══════════╬══════╬═════╬════════════════╬════════════════════╬══════════╣
║ 2014 ║  Caron Butler      ║ -4.7      ║ 0.5       ║ -5.2 ║  33 ║             12 ║ during 2nd season  ║          ║
║ 2002 ║  Tim Hardaway      ║ -0.7      ║ 4.3       ║ -5   ║  35 ║             12 ║ during 2nd season  ║          ║
║ 2013 ║  Rodney Stuckey    ║ -1.2      ║ 3.7       ║ -4.9 ║  26 ║              6 ║                    ║          ║
║ 2004 ║  Drew Gooden       ║ -6.5      ║ -1.9      ║ -4.6 ║  22 ║              2 ║ during 1st season  ║          ║
║ 2007 ║  Eric Snow         ║ -2.5      ║ 2.1       ║ -4.6 ║  33 ║             12 ║                    ║          ║
║ 2010 ║  Kevin Garnett     ║ 6.4       ║ 11        ║ -4.6 ║  33 ║             15 ║                    ║          ║
║ 2002 ║  Donyell Marshall  ║ -1.8      ║ 2.7       ║ -4.5 ║  28 ║              8 ║                    ║ probably ║
║ 2003 ║  Andre Miller      ║ 1.3       ║ 5.8       ║ -4.5 ║  26 ║              4 ║ In-between seasons ║          ║
║ 2013 ║  Brandon Jennings  ║ -4.5      ║ -0        ║ -4.5 ║  23 ║              4 ║                    ║          ║
║ 2007 ║  Ben Wallace       ║ 1.9       ║ 6.3       ║ -4.4 ║  32 ║             11 ║ In-between seasons ║          ║
║ 2007 ║  Andrei Kirilenko  ║ 4.1       ║ 8.5       ║ -4.4 ║  25 ║              6 ║                    ║          ║
║ 2014 ║  Ersan Ilyasova    ║ -0.9      ║ 3.5       ║ -4.4 ║  26 ║              8 ║                    ║ X        ║
║ 2010 ║  Devin Harris      ║ 1.2       ║ 5.6       ║ -4.4 ║  26 ║              6 ║                    ║ X        ║
║ 2009 ║  Allen Iverson     ║ -0.6      ║ 3.7       ║ -4.3 ║  33 ║             13 ║ during 2nd season  ║ X        ║
║ 2014 ║  Kendrick Perkins  ║ -4.1      ║ 0.2       ║ -4.3 ║  29 ║             11 ║                    ║ X        ║
║ 2014 ║  Luke Ridnour      ║ -3.8      ║ 0.4       ║ -4.2 ║  32 ║             11 ║ during 2nd season  ║ X        ║
║ 2006 ║  Mike Miller       ║ 0.3       ║ 4.4       ║ -4.1 ║  25 ║              6 ║                    ║          ║
║ 2009 ║  Tyson Chandler    ║ 0.8       ║ 4.9       ║ -4.1 ║  26 ║              8 ║                    ║ X        ║
║ 2013 ║  Dwight Howard     ║ 4.1       ║ 8.2       ║ -4.1 ║  27 ║              9 ║ In-between seasons ║ X        ║
║ 2014 ║  Luis Scola        ║ -3.4      ║ 0.6       ║ -4   ║  33 ║              7 ║ In-between seasons ║          ║
╚══════╩════════════════════╩═══════════╩═══════════╩══════╩═════╩════════════════╩════════════════════╩══════════╝

Average age is ~29. Many players with 10+ seasons. A good portion of these were injured during the 2nd season - nagging injuries could've definitely hurt their performance. Four players changed teams during the 2nd season - I guess their team recognized they weren't a great fit and made significant effort to trade them. Another four players switched teams between seasons and were probably forced to play a role with their new team that they weren't as comfortable with

Posted: **Wed Oct 15, 2014 10:42 pm**

Is this a prior informed RAPM? If so, wouldn't it be already biased? y1 would be influenced with a y-1 prior

Have you looked at doing this with just a simple NPI Rapm without any RAPM or box score priors?

Posted: **Thu Oct 16, 2014 2:57 am**

Are these large yr to yr changes associated with a higher rate of coaching changes than the overall average? I don't think so but I haven't done the precise math (est. count of coaching change to at least 25% but probably below 40%). Are they associated with a higher rate of change in the intelligence of team strategy apart from simple change of coach or in quality of teammates in ways that impact individual spm or change in individual basketball intelligence, focus and effort? Are these cases of extreme sliding because of the method (the core or the add-one)? Or randomness in the observed world? Or a little of several or all? I don't think changing teams or injury / healing are large coincidences with appearance in these sets. But if time availability and curiosity were great enough one could correlate a large number of attributes and circumstances with having the level of rapm change to make these lists. If we can't decide for sure how much these factors are impact in general, can judgment make reasonable guesses in at least some cases? Is it worth the time guessing about these things? If so then isn't imprecise rpm still a useful tool for groping after insight or educated hunches that are better than thinking derived from more rigidly limited input? Better or not, it is a choice and a choice that can be judged from experience. Different perhaps from one person, place and experience to the next.

Posted: **Thu Oct 16, 2014 10:01 am**

J.E. wrote:... It's obviously possible, or even likely, that these players did actually improve but we can't be sure
...

If their boxscore stats indicate they've improved, then the likelihood increases.
If you have ASPM handy, why not compare their "expected xRAPM improvement" to their measured xRAPM improvement?
Then we'd see lists of players whose on-court effectiveness improved more than their stats indicate they should have?

Here are some other boxscore summary stats, lined up with your xRAPM for the "improved players".

Code: Select all

xRAPM     improvers             - xRAPM -      - eW/484 -      - - PER - -       - ws/48 -  
diff   name            yr2      yr1   yr2      yr1    yr2       yr1    yr2       yr1    yr2
5.5   David Lee       2007     -2.4   3.1      .76   1.33      15.4   20.2      .116   .191
5.2   Kevin Durant    2010     -2.6   2.6     1.70   2.45      20.8   26.2      .132   .238
4.9   Drew Gooden     2005     -6.5  -1.6      .89   1.61      15.6   19.7      .043   .163
4.7   Boris Diaw      2006     -1.6   3.1      .14   1.29      10.0   17.3      .005   .149
4.7   Caron Butler    2008     -1.5   3.2     1.24   1.45      18.3   20.7      .103   .150

4.7  Tristan Thompson 2013     -5.8  -1.1      .86   1.15      13.3   16.1      .038   .098
4.6   Dwight Howard   2006     -2.9   1.7     1.20   1.58      17.2   19.3      .131   .137
4.6   Dwight Howard   2010      3.4   8.0     2.52   2.42      25.4   24.0      .234   .223
4.5   Draymond Green  2014       .1   4.6      .34    .77       7.1   12.7      .028   .119
4.5   Goran Dragic    2014      -.2   4.3     1.14   1.64      17.5   21.4      .106   .186

4.5   Hedo Turkoglu   2004      -.5   4.0      .46    .95      12.0   14.1      .105   .155
4.4   Carlos Boozer   2004     -4.6   -.2     1.11   1.68      17.9   20.8      .139   .174
4.4   Dwyane Wade     2006      3.0   7.4     2.18   2.62      23.1   27.6      .177   .239
4.3   Mike Conley     2011      -.4   3.9      .67   1.19      13.9   15.9      .074   .111
4.2   Kenyon Martin   2002     -3.1   1.1      .87   1.15      13.4   14.5      .044   .093

4.2   Paul George     2012       .6   4.8      .61   1.32      13.0   16.5      .091   .148
4.1   Marcus Morris   2014     -4.2   -.1      .64    .85      11.3   14.8      .058   .111

Most all of our numbers agree these guys improved from year1 to year2. Exception: Dwight Howard from 2009 to 2010 -- seems not to have beefed up his numbers, but great leap in xRAPM.

NBA averages: xRAPM = 0, eW/484 = 1.00, PER = 15, WS/48 = .100
I've only looked up players active last season (cuz it was easier). Here are those with big yr-to-yr drops in xRAPM:

Code: Select all

xRAPM     droppers              - xRAPM -       - eW/484 -      - - PER - -       - ws/48 -  
diff     name           yr2     yr1   yr2       yr1    yr2       yr1    yr2       yr1    yr2
-4.0   Luis Scola      2014      .6  -3.4      1.28   1.14      16.7   13.4      .096   .084
-4.1   Dwight Howard   2013     8.2   4.1      2.38   1.65      24.2   19.4      .179   .134
-4.1   Mike Miller     2006     4.4    .3      1.38   1.33      16.7   17.2      .148   .157
-4.1   Tyson Chandler  2009     4.9    .8      1.32    .80      17.5   13.4      .172   .114
-4.2   Luke Ridnour    2014      .4  -3.8       .77    .29      13.0    9.0      .084   .005

-4.3  Kendrick Perkins 2014      .2  -4.1       .53    .26       8.2    6.3      .062   .031
-4.4  Andrei Kirilenko 2007     8.5   4.1      1.67   1.07      20.7   14.3      .145   .104
-4.4   Devin Harris    2010     5.6   1.2      1.75   1.09      21.6   16.2      .145   .057
-4.4   Ersan Ilyasova  2014     3.5   -.9      1.40    .90      18.3   13.8      .159   .053
-4.5   Andre Miller    2003     5.8   1.3      1.87   1.02      21.8   15.2      .163   .076

-4.5  Brandon Jennings 2013      .0  -4.5      1.44   1.12      18.4   16.1      .114   .096
-4.6   Drew Gooden     2004    -1.9  -6.5      1.18    .89      15.0   15.6      .043   .043
-4.6   Kevin Garnett   2010    11.0   6.4      2.17   1.84      21.2   19.4      .193   .171
-4.9   Rodney Stuckey  2013     3.7  -1.2      1.12    .69      17.6   13.0      .131   .055
-5.2   Caron Butler    2014      .5  -4.7       .78    .75      12.4   12.2      .091   .062

The anomalies here are: Mike Miller from '05 to '06 -- eWins shows very little dropoff, while PER and WS show small upticks.
Also Caron Butler last year, didn't seem to drop off much statistically, in eW or PER. Was his -4.7 for the whole season, or just in OKC?
Many or most of these year-to-years look like biggest career drops / rises in all stats.

I ran 'diff' (yr2 - yr1) for the other stats, got their correlations with column 1:
eWins .870
PER .845
WS .840

Posted: **Tue Oct 21, 2014 12:01 pm**

Mike G wrote:If you have ASPM handy, why not compare their "expected xRAPM improvement" to their measured xRAPM improvement?

"expected xRAPM improvement", for me, is nothing but "old xRAPM rating"+age factor

Note that xRAPM already includes BoxScore information - that's probably why we see a strong correlation for 'diff' with xRAPM and the other metrics

Also Caron Butler last year, didn't seem to drop off much statistically, in eW or PER. Was his -4.7 for the whole season, or just in OKC?

That's for the whole season. He posted a -5.5 NET in OKC and a -4.5 NET rating in MIL. Considering the quality of teammates, a negative NET in OKC isn't really that bad, but a negative NET for '13-'14 MIL certainly is

Here's another batch of "breakout players". Those that went from playing <1000 minutes to playing >1000 and saw their xRAPM increase by a significant amount

Code: Select all

╔══════╦════════════════════╦═══════════╦═══════════╦══════╦═════╦════════════════╗
║ Year ║        Name        ║ New xRAPM ║ old xRAPM ║ Diff ║ Age ║ Year in League ║
╠══════╬════════════════════╬═══════════╬═══════════╬══════╬═════╬════════════════╣
║ 2003 ║  Earl Boykins      ║ 1.3       ║ -3.3      ║ 4.6  ║  26 ║              5 ║
║ 2003 ║  Mike James        ║ 0.7       ║ -3.5      ║ 4.2  ║  27 ║              2 ║
║ 2004 ║  Brian Cardinal    ║ 5.5       ║ -2.3      ║ 7.8  ║  26 ║              4 ║
║ 2005 ║  Steven Hunter     ║ -0.6      ║ -4.8      ║ 4.2  ║  23 ║              4 ║
║ 2006 ║  DeSagana Diop     ║ 2.3       ║ -3.8      ║ 6.1  ║  24 ║              5 ║
║ 2006 ║  Kirk Snyder       ║ 0.7       ║ -4.5      ║ 5.2  ║  22 ║              2 ║
║ 2007 ║  Ime Udoka         ║ 3.3       ║ -2.3      ║ 5.6  ║  29 ║              3 ║
║ 2007 ║  Dahntay Jones     ║ -2.3      ║ -6.2      ║ 3.9  ║  26 ║              4 ║
║ 2007 ║  Fabricio Oberto   ║ 1.8       ║ -2.4      ║ 4.2  ║  31 ║              2 ║
║ 2007 ║  Matt Barnes       ║ 0.9       ║ -2.9      ║ 3.8  ║  26 ║              4 ║
║ 2008 ║  Ronnie Brewer     ║ 2.3       ║ -3        ║ 5.3  ║  22 ║              2 ║
║ 2009 ║  Matt Bonner       ║ 3         ║ -2.3      ║ 5.3  ║  28 ║              5 ║
║ 2009 ║  Ramon Sessions    ║ 2.1       ║ -1.7      ║ 3.8  ║  22 ║              2 ║
║ 2010 ║  Donte Greene      ║ -2.1      ║ -6.4      ║ 4.3  ║  21 ║              2 ║
║ 2011 ║  Jodie Meeks       ║ 1.8       ║ -2.9      ║ 4.7  ║  23 ║              2 ║
║ 2012 ║  Nikola Pekovic    ║ -1.3      ║ -4.9      ║ 3.6  ║  26 ║              2 ║
║ 2012 ║  Avery Bradley     ║ 0.4       ║ -3.7      ║ 4.1  ║  21 ║              2 ║
║ 2012 ║  Danny Green       ║ 4.7       ║ -0.5      ║ 5.2  ║  24 ║              3 ║
║ 2013 ║  Lance Stephenson  ║ 1.4       ║ -3.8      ║ 5.2  ║  22 ║              3 ║
║ 2013 ║  Enes Kanter       ║ -2.6      ║ -6.2      ║ 3.6  ║  20 ║              2 ║
║ 2013 ║  DeMarre Carroll   ║ 3.3       ║ -0.8      ║ 4.1  ║  26 ║              4 ║
║ 2014 ║  Kendall Marshall  ║ -1.7      ║ -5.6      ║ 3.9  ║  22 ║              2 ║
║ 2014 ║  Patrick Mills     ║ 4.3       ║ -1.7      ║ 6    ║  25 ║              5 ║
║ 2014 ║  Timofey Mozgov    ║ -0.2      ║ -5.2      ║ 5    ║  27 ║              5 ║
╚══════╩════════════════════╩═══════════╩═══════════╩══════╩═════╩════════════════╝

Many of these went from "awful" to decent in xRAPM, in part because I'm only selecting those with <1000 MP and MP has a strong positive coeficient in SPM (which gets fed into xRAPM). Four (ex) Spurs on the list: Oberto, Bonner, Green and Mills. Obviously, all in the Popovich era

Watch out for Kyle Anderson in 2016!

Posted: **Mon Jun 01, 2015 11:01 pm**

There could be some followup on the big movers of 2013 and 2014 in 2015. I might get to it later. Anyone interested in doing it / seeing it?

APBRmetrics

Players with big y-t-y changes in RPM

Players with big y-t-y changes in RPM

Re: Players with big y-t-y changes in RPM

Re: Players with big y-t-y changes in RPM

Re: Players with big y-t-y changes in RPM

Re: Players with big y-t-y changes in RPM

Re: Players with big y-t-y changes in RPM