Page 3 of 6
Re: 2012-2013 predictions
Posted: Thu Nov 15, 2012 6:56 pm
by v-zero
Well these are my current teams strengths... I don't have Wade injured for Miami, I use BBref for injury reports, every not listed as probable is assumed out.
Code: Select all
Miami Heat 7.3
San Antonio Spurs 7.15
Oklahoma City Thunder 6.26
New York Knicks 4.81
Chicago Bulls 4.37
Memphis Grizzlies 3.83
Denver Nuggets 3.64
Los Angeles Clippers 2.61
Los Angeles Lakers 2.33
Utah Jazz 2.15
Boston Celtics 1.21
Minnesota Timberwolves 1.04
Atlanta Hawks 0.15
Brooklyn Nets 0.08
Houston Rockets -0.08
Milwaukee Bucks -0.5
Dallas Mavericks -0.66
Philadelphia 76ers -0.81
Indiana Pacers -1.04
Portland Trail Blazers -1.37
Phoenix Suns -1.64
Orlando Magic -1.75
Toronto Raptors -2.63
Golden State Warriors -3.71
New Orleans Hornets -3.82
Cleveland Cavaliers -4.01
Washington Wizards -4.8
Sacramento Kings -5.09
Detroit Pistons -6.1
Charlotte Bobcats -8.92
Re: 2012-2013 predictions
Posted: Fri Nov 16, 2012 3:10 pm
by Neil Paine
DSMok1 wrote:Should probably post RMS error, Neil. That's how predictions are usually tested, aren't they?
Good idea (I was just going with what you guys seemed to be doing last year)... Here are the RMSEs for each projection thus far:
Code: Select all
--------------------
Projection RMSE
--------------------
TeamRankings 7.12
np "dumb" 7.26
DSMok1 7.42
DSMok1_2 7.49
WoW2 8.27
Bobbofitos 8.58
WoW1 8.62
v-zero 8.88
SCHOENE 9.03
Schaller 9.34
ESPN Summer 9.38
Bovada 9.47
LV Hilton 9.65
Hollinger 10.72
--------------------
Re: 2012-2013 predictions
Posted: Sat Nov 17, 2012 5:28 am
by EvanZ
I'll try to keep updating my Vegas ratings about once a week.
http://www.d3coder.com/thecity/vegas-power-ratings/
Re: 2012-2013 predictions
Posted: Thu Nov 29, 2012 6:52 pm
by Neil Paine
Checking in with these thru last night's games:
Code: Select all
Projection RMSE
--------------------
np "dumb" 7.12
DSMok1_2 7.12
DSMok1 7.13
TeamRankings 7.42
Bovada 7.91
Bobbofitos 8.04
LV Hilton 8.30
Schaller 8.46
ESPN Summer 8.50
v-zero 8.68
SCHOENE 8.78
WoW2 8.97
WoW1 9.02
Hollinger 9.50
Here were the Bayesian win values I'm measuring against:
Code: Select all
Team Mean W
----------------------------
Memphis Grizzlies 57.9
San Antonio Spurs 57.8
Oklahoma City Thunder 54.4
Miami Heat 54.2
New York Knicks 51.8
Brooklyn Nets 51.8
Atlanta Hawks 50.4
Los Angeles Clippers 46.2
Philadelphia 76ers 46.2
Golden State Warriors 44.6
Utah Jazz 44.3
Milwaukee Bucks 42.8
Denver Nuggets 42.7
Boston Celtics 42.7
Chicago Bulls 41.0
Charlotte Bobcats 41.0
Los Angeles Lakers 39.2
Indiana Pacers 39.2
Houston Rockets 39.2
Dallas Mavericks 37.6
Phoenix Suns 37.6
Minnesota Timberwolves 37.3
Portland Trail Blazers 35.7
Orlando Magic 33.7
Detroit Pistons 30.9
New Orleans Hornets 30.1
Sacramento Kings 30.1
Cleveland Cavaliers 25.2
Toronto Raptors 24.1
Washington Wizards 20.2
Re: 2012-2013 predictions
Posted: Thu Nov 29, 2012 7:09 pm
by v-zero
Neil Paine wrote:Checking in with these thru last night's games:
Code: Select all
Projection RMSE
--------------------
np "dumb" 7.12
DSMok1_2 7.12
DSMok1 7.13
TeamRankings 7.42
Bovada 7.91
Bobbofitos 8.04
LV Hilton 8.30
Schaller 8.46
ESPN Summer 8.50
v-zero 8.68
SCHOENE 8.78
WoW2 8.97
WoW1 9.02
Hollinger 9.50
Here were the Bayesian win values I'm measuring against:
Code: Select all
Team Mean W
----------------------------
Memphis Grizzlies 57.9
San Antonio Spurs 57.8
Oklahoma City Thunder 54.4
Miami Heat 54.2
New York Knicks 51.8
Brooklyn Nets 51.8
Atlanta Hawks 50.4
Los Angeles Clippers 46.2
Philadelphia 76ers 46.2
Golden State Warriors 44.6
Utah Jazz 44.3
Milwaukee Bucks 42.8
Denver Nuggets 42.7
Boston Celtics 42.7
Chicago Bulls 41.0
Charlotte Bobcats 41.0
Los Angeles Lakers 39.2
Indiana Pacers 39.2
Houston Rockets 39.2
Dallas Mavericks 37.6
Phoenix Suns 37.6
Minnesota Timberwolves 37.3
Portland Trail Blazers 35.7
Orlando Magic 33.7
Detroit Pistons 30.9
New Orleans Hornets 30.1
Sacramento Kings 30.1
Cleveland Cavaliers 25.2
Toronto Raptors 24.1
Washington Wizards 20.2
Any chance you could do another using pythagorean win percentage from last season as the prior?
Re: 2012-2013 predictions
Posted: Thu Dec 06, 2012 6:59 pm
by Neil Paine
Sure, I can do that. This is through last night's games:
Code: Select all
RMSE, .500 Prior
---------------------
1. np "dumb" 6.574
2. DSMok1 6.858
3. DSMok1_2 6.863
4. TeamRankings 7.283
5. Bovada 7.482
6. Bobbofitos 7.735
7. LV Hilton 7.921
8. v-zero 7.979
9. ESPN Summer 8.016
10. Schaller 8.300
11. WoW2 8.359
12. WoW1 8.417
13. SCHOENE 8.731
14. Hollinger 9.101
---------------------
RMSE, '12 Pyth Prior
---------------------
1. np "dumb" 7.019
2. Bovada 7.259
3. DSMok1_2 7.449
4. DSMok1 7.490
5. v-zero 7.543
6. LV Hilton 7.785
7. Bobbofitos 8.049
8. TeamRankings 8.061
9. ESPN Summer 8.143
10. WoW2 8.553
11. Schaller 8.699
12. WoW1 8.766
13. SCHOENE 9.054
14. Hollinger 9.164
---------------------
Re: 2012-2013 predictions
Posted: Thu Dec 06, 2012 8:58 pm
by v-zero
Neil Paine wrote:Sure, I can do that. This is through last night's games:
Code: Select all
RMSE, .500 Prior
---------------------
1. np "dumb" 6.574
2. DSMok1 6.858
3. DSMok1_2 6.863
4. TeamRankings 7.283
5. Bovada 7.482
6. Bobbofitos 7.735
7. LV Hilton 7.921
8. v-zero 7.979
9. ESPN Summer 8.016
10. Schaller 8.300
11. WoW2 8.359
12. WoW1 8.417
13. SCHOENE 8.731
14. Hollinger 9.101
---------------------
RMSE, '12 Pyth Prior
---------------------
1. np "dumb" 7.019
2. Bovada 7.259
3. DSMok1_2 7.449
4. DSMok1 7.490
5. v-zero 7.543
6. LV Hilton 7.785
7. Bobbofitos 8.049
8. TeamRankings 8.061
9. ESPN Summer 8.143
10. WoW2 8.553
11. Schaller 8.699
12. WoW1 8.766
13. SCHOENE 9.054
14. Hollinger 9.164
---------------------
And now it looks like I just wanted to look better
Re: 2012-2013 predictions
Posted: Thu Dec 06, 2012 10:24 pm
by Crow
NP "dumb" was not so simple or dumb. Hollinger having an off year, as is Schaller. Some other metrics near the bottom are in familiar territory. DSMok1's entries may not be hampered by the problem that hurt last season.
Re: 2012-2013 predictions
Posted: Fri Dec 07, 2012 1:46 am
by DSMok1
Crow wrote:NP "dumb" was not so simple or dumb. Hollinger having an off year, as is Schaller. Some other metrics near the bottom are in familiar territory. DSMok1's entries may not be hampered by the problem that hurt last season.
The main issue for me last season was I didn't have any decent playing time projection.
Re: 2012-2013 predictions
Posted: Fri Dec 07, 2012 3:25 pm
by Neil Paine
Crow wrote:NP "dumb" was not so simple or dumb. Hollinger having an off year, as is Schaller. Some other metrics near the bottom are in familiar territory. DSMok1's entries may not be hampered by the problem that hurt last season.
The "Dumb" metric's success early in the season is probably an artifact of its heavy regression to the mean component (since the standard I'm measuring it against is also regressed to the mean, adding 11.8 games of .500 W-L to each team's record). I take no credit for it being #1 right now because it's only there as a sanity check against the other systems. I fully expect it to decline as the season goes on, because some of the bolder (i.e. less regressed) predictions will be closer to the regressed WPcts as teams get more games under their belts and the RTM becomes progressively weaker.
If it's the end of the season and "Dumb" still wins, it would represent a credibility problem for advanced stats. It would mean I could take the Win Shares framework, add in a very conservative usage/efficiency tradeoff, use a basic 3-year weighted + RTM projection with an extremely naive age adjustment, plug it into ESPN's fantasy playing-time projections, and out-predict far more complex systems.
Re: 2012-2013 predictions
Posted: Fri Dec 07, 2012 3:31 pm
by v-zero
Neil Paine wrote:The "Dumb" metric's success early in the season is probably an artifact of its heavy regression to the mean component (since the standard I'm measuring it against is also regressed to the mean, adding 11.8 games of .500 W-L to each team's record). I take no credit for it being #1 right now because it's only there as a sanity check against the other systems. I fully expect it to decline as the season goes on, because some of the bolder (i.e. less regressed) predictions will be closer to the regressed WPcts as teams get more games under their belts and the RTM becomes progressively weaker.
If it's the end of the season and "Dumb" still wins, it would represent a credibility problem for advanced stats. It would mean I could take the Win Shares framework, add in a very conservative usage/efficiency tradeoff, use a basic 3-year weighted + RTM projection with an extremely naive age adjustment, plug it into ESPN's fantasy playing-time projections, and out-predict far more complex systems.
That is more than went into my predictions, I must admit. My minute projections were based on assuming that coaches knew who to play, and the number of minutes in the season was the same for all players at the same depth position on their team... Though I did account for long-term injuries somewhat. I did predict rookie performance for those in the first round out of college, but that was the depth of my rookie work.
Re: 2012-2013 predictions
Posted: Fri Dec 07, 2012 3:48 pm
by DSMok1
Neil Paine wrote:Crow wrote:NP "dumb" was not so simple or dumb. Hollinger having an off year, as is Schaller. Some other metrics near the bottom are in familiar territory. DSMok1's entries may not be hampered by the problem that hurt last season.
The "Dumb" metric's success early in the season is probably an artifact of its heavy regression to the mean component (since the standard I'm measuring it against is also regressed to the mean, adding 11.8 games of .500 W-L to each team's record). I take no credit for it being #1 right now because it's only there as a sanity check against the other systems. I fully expect it to decline as the season goes on, because some of the bolder (i.e. less regressed) predictions will be closer to the regressed WPcts as teams get more games under their belts and the RTM becomes progressively weaker.
If it's the end of the season and "Dumb" still wins, it would represent a credibility problem for advanced stats. It would mean I could take the Win Shares framework, add in a very conservative usage/efficiency tradeoff, use a basic 3-year weighted + RTM projection with an extremely naive age adjustment, plug it into ESPN's fantasy playing-time projections, and out-predict far more complex systems.
If baseball is any indicator, it is very hard to beat a system like the one you have described. Very hard.
Re: 2012-2013 predictions
Posted: Wed Dec 19, 2012 4:09 pm
by Neil Paine
Checking in with the RMSEs through last night's games:
Code: Select all
-----------------------------------------------------------
System .500 Prior Rk Pyth Prior Rk
-----------------------------------------------------------
np "dumb" 7.030 1 7.727 1
DSMok1_2 7.513 2 8.264 4
Bobbofitos 7.658 4 8.252 3
Bovada 8.051 6 8.250 2
DSMok1 7.615 3 8.385 6
v-zero 8.244 7 8.301 5
TeamRankings 7.883 5 8.753 8
LV Hilton 8.479 9 8.714 7
WoW2 8.394 8 8.883 9
WoW1 8.480 10 9.069 10
ESPN Summer 8.712 11 9.106 11
Schaller 8.969 12 9.540 12
SCHOENE 9.286 13 9.808 13
Hollinger 9.833 14 10.154 14
-----------------------------------------------------------
Re: 2012-2013 predictions
Posted: Wed Dec 19, 2012 8:52 pm
by Crow
Bobbofitos joins the battle for 2nd place.
Re: 2012-2013 predictions
Posted: Wed Jan 02, 2013 5:07 pm
by Neil Paine
Checking in again, through 1/1/2013:
Code: Select all
Rk Naive pace RMSE Rk .500 prior RMSE Rk Pyth prior RMSE
----------------------------- ----------------------------- -----------------------------
1 np "dumb" 9.00 1 np "dumb" 7.04 1 np "dumb" 6.37
2 Bovada 9.12 2 Bovada 7.81 2 Bovada 7.20
3 Bobbofitos 9.45 3 Bobbofitos 7.82 3 DSMok1_2 7.28
4 DSMok1_2 9.71 4 DSMok1_2 7.85 4 Bobbofitos 7.31
5 WoW2 9.72 5 DSMok1 7.88 5 DSMok1 7.33
6 LV Hilton 9.75 6 WoW1 8.16 6 TeamRankings 7.56
7 DSMok1 9.76 7 TeamRankings 8.17 7 WoW2 7.68
8 WoW1 9.79 8 WoW2 8.18 8 WoW1 7.68
9 ESPN Summer 9.94 9 LV Hilton 8.37 9 v-zero 7.71
10 v-zero 9.94 10 v-zero 8.52 10 LV Hilton 7.75
11 TeamRankings 10.24 11 ESPN Summer 8.55 11 ESPN Summer 8.05
12 Schaller 10.25 12 Schaller 8.78 12 Schaller 8.38
13 Hollinger 10.65 13 SCHOENE 9.17 13 SCHOENE 8.67
14 SCHOENE 10.73 14 Hollinger 9.44 14 Hollinger 9.02
----------------------------- ----------------------------- -----------------------------
"Naive pace" is simply the # of wins you'd predict if every team won at their current WPct rate for the rest of the season.