Posted: Tue Mar 04, 2008 6:38 am Post subject: HOF Predictions
It started on RealGM, then I tryed to build something to predict future HOFers and the achievements needed to get in.
I wanted to totally not include boxscore stats and concentrate on personal accolades and influence on team results.
I've yet to fine tune it, it's just some sort of first draft.
HOF point = sum nn points for each of the following:
MVP: 80
Finals MVP: 20
1st team: 20
2nd team: 10
3rd team: 5
DPOY: 10
Ring as obvious #1 player: 40
Ring as sidekick (or in a very balaced team): 20
Final lost as #1 player: 20
Conference Final lost as #1 player: 10
100 might be the threshold to pass to be sure, 70 to be in the bubble.
here some results (if their career ended today!) on players not too obvious like Malone or Olajuwon:
Locks:
Scottie Pippen 210
Nowitzki: 200
Kidd: 150
Gary Payton 130
Clyde Drexler* 120
John Stockton 115
Patrick Ewing 110
Bubble:
Reggie Miller 85
Wade: 85
Joe Dumars* 80
Parker: 80
McGrady: 75
Dominique Wilkins* 70
Likely out:
LeBron: 60
Mutombo: 60
Ginobili: 60
Grant Hill 60
Mourning: 55
Tim Hardaway 55
Kevin Johnson 55
Chris Webber 55
Bernard King 55
Chauncey Billups 55
No chance:
Alex English* 40
Adrian Dantley 20
What do you think?
I realize I can't go too much back in time because some awards are too new, even in the 80s we didn't have the 3rd All NBA team, but you think it could predict future nominations?
The parameters are yet to fine tune, as I previously mentioned, I made them up in a few minuts.
The most tricky part is the one related on team success for no obvious stars, a guy like Dennis Johnson could go as low as 70 or as high as 140, depending on how you define his role in those Seattle or Boston teams (I think 70 is more accurate, between the two).
Joined: 05 Jan 2005 Posts: 613 Location: Columbus, OH
Posted: Tue Mar 04, 2008 7:55 am Post subject:
I have done something similar to this using a logistic regression model. Details are available here. _________________ Regards,
Justin Kubatko
Basketball-Reference.com
I've seen your work, I'm just not sure how it will work with future nominations.
How do you think it predicts players who played in 80s and 90s?
Do you keep a stat on how it predicted recent players?
Wouldn't it be nice to add a column here?
The ones more arguable, predicting Magic or Bird is easy...
Joined: 14 Jan 2005 Posts: 1888 Location: Delphi, Indiana
Posted: Tue Mar 04, 2008 9:22 am Post subject:
The swift induction of Alex English may indicate a flaw in the process. You give him 40 'HOF points', based on his (3) 2nd-teams and 1 trip to the WCF 'as #1 player'.
I suppose his 8 Allstar appearances went into his selection to HOF. Twice he led the league in points scored, 3 times in FG. By b-r.com, his HOF% is .811 .
Dantley, with only a 20 score, also had impressive credentials: 6 ASG, 2 2nd-teams, 2 scoring titles, unprecedented combo of scoring and efficiency. In the '88 run to the Finals, he was the 2nd Piston in minutes and in scoring. In those Finals, easily their top scorer, shooting .672 TS%. Strong 2nd in minutes.
So, rather than arbitrarily deciding that a guy was #1 or #2 or 'other', just quantify. By minutes + production, perhaps, as a % of the total. Then again, you might just end up with what Justin's done (which is great). _________________ `
There's no I in analysys.
You know, I think Alex English actually was a pretty weak candidate, probably the weakest among the guys who played in the last 25-30 years.
He reminds me Vince Carter, who's almost a lock according to B-R but out following my rating. How would you consider him?
I think the an important test will be if guys like him or Pierce, Allen, Dantley etc will eventually sneak in or not, to see how much voters factor stats with moderate team or personal accolades.
BTW I should probably put ASGs in the equation, a guy who has many of them has probably been considered very good for quite a long time.
I apologize in advance if I don't do this the justice it deserves, but Justin's approach is to to test variables for signifiance, apply them if significant, and allow them to "weigh" themselves through regression. The only way to test it definitively is to quantify the hit% based upon actual entrants. New data might require the model to be "refit".
The challenge is that Hall of Fame requirements have shifted by era: compare the change in the game from 47 to 77, and then compare 77 to 2007. Basketball has matured a lot, and there are probably a lot of early guys who would not get in today. The "dummy variable" used for pre-60 players addresses some of this in the HOF probability.
Looking at the post-1980 guys rated in the HOF probability who are not in already:
-16 of the first 17 are all HOFers in my mind, with Carter being the lone exception.
-there are 8 more with HOF prob > 50%: McGrady, Pierce, Allen, Richmond, Hill, Webber, Tim Hardaway, Arenas. I don't see many of these guys making it, but many are still playing, so that may change.
-there are another 33 or so that have a HOF prob > 3%. 20 of them are retired or all but retired. Of those 20, I'd rate their chances: 1-Miller (near lock), 2-Rodman, 3-Mullin, 4-Mourning, 5-Cheeks, 6-Aguirre. Maybe three of these guys get in.
-everyone else below 3% is either way too young, or has no chance, with the exception being Mutombo (big maybe).
Considering the changes in era, I'd say the system does a great job.
Joined: 05 Jan 2005 Posts: 613 Location: Columbus, OH
Posted: Wed Mar 05, 2008 7:33 am Post subject:
Ryoga Hibiki wrote:
I've seen your work, I'm just not sure how it will work with future nominations.
Only looking at the new candidates, the system sees Hakeem Olajuwon and Patrick Ewing as being locks this year; Michael Jordan, David Robinson, and John Stockton being inducted in 2009; Karl Malone and Scottie Pippen gaining entry in 2010; and no first-time candidates making it in 2011.
Quote:
How do you think it predicts players who played in 80s and 90s?
Time will eventually tell, but I think it will do quite well. _________________ Regards,
Justin Kubatko
Basketball-Reference.com
Joined: 05 Jan 2005 Posts: 613 Location: Columbus, OH
Posted: Wed Mar 05, 2008 7:36 am Post subject:
Ryoga Hibiki wrote:
He reminds me Vince Carter, who's almost a lock according to B-R but out following my rating.
A note about active players: Points per game, rebounds per game, and assists per game are used in the probability calculation. For active players this can lead to a misleading result, as many of these players are still at or near their peak level, and therefore have not seen the (almost) inevitable decline in their rate statistics. _________________ Regards,
Justin Kubatko
Basketball-Reference.com
I thought I'd just tack my own HOF monitor method on to this thread. When putting this together, I tried to be mindful of the fact that players HOF selectors probably aren't looking at advanced stats, but rather general statistics, awards/accolades, and memorable playoff runs. This was done about 18 months ago after seeing the HOF probabilty calculator. I also didn't try to recreate the wheel with Justin's regression-based HOF probability method, as he has accounted for era about as well as anyone can. The nomination process has evolved and become more stringent over time, so it focuses on more recent players (where similar criteria is more likely to be consistently applied). I looked at players entering the league from 76/77 on, the bulk of whom will be the NBA nominees going forward, and I tried to think like a selector.
Metrics used:
1-General statistics. I realize I just said selectors probably don't use advanced stats as a criterion, but rather than look at ppg, apg, rpg, etc, I stuck with PER to capture a bit of everything in one easy number. To account for longevity vs. greatness, I looked at excess PER over replacement ( rep = 12, assuming some chaining for subs) and multiplied by total min. Total min were then divided by (82 x 48 minutes) to adjust for seasons. The resulting number is one component of the total HOF score. I prefer this method as it doesn't seem to penalize guys who play extra years at 12-15 PER with limited minutes, but it also doesn't help them much either. General stats represents about 50% of total HOF pts for the top 50 players. The top 5 beneficiaries of this component (as % of their total HOF score) were Nance, Cummings, Dantley, Aguirre, and English.
2-Awards/accolades. PER doesn't capture all defensive aspects and I believe selectors consider this to some degree. General awards such as all-NBA teams and MVPs especially are also remembered. To account for D and other awards, I threw together a simple pt system based upon a subjective hierarchy:
Pts
1 - 2nd Team D selection
2 - 1st Team D
3 - All-Star
4 - 3rd team League
5 - 2nd team League
6- 1st team League
10 - MVP
Awards account for about 30% of the total HOF score. The top 5 beneficiaries of this component were Moncrief, Payton, Richmond, Bird, and Olajuwon.
3-Playoff Runs. Selectors, the media, and fans remember good teams. Not just championship teams, but also classic playoff matchups. In the EC alone: Sixers vs. Celtics of the early 80s, Pistons vs. Celtics and Pistons vs. Bulls in the late 80s, Pacer/Knick/Bull/Heat battles later on, and so forth. By the time the conference finals roll around, there aren't a lot of games to see, and a lot more exposure is given to CF particpants. It's not enough to be on the roster. You need to be on the court too. Pt system:
Round 1 losers = no pts
Round 2 losers = no pts
Round 3 (CF) losers = 3 pts * % of total team plyf min played
Round 4 (champ) losers = 4 pts * % of total team plyf min played
Round 5 (champ) winners = 5 pts * % of total team plyf min played
This represents about 20% of total HOF pts for the top 50 players. The top 5 beneficiaries of this component were Grant, Rodman, DJ, Worthy, and Cheeks.
Results The system seems to do a reasonably good job of separating HOFers from those on the outside looking in. 120 is the approximate cutoff. Everyone above (other than Dantley) is in and everyone below (except Dumars) is out. Of the actives, everyone above 120 I would call a lock with the possible exception of McGrady at 124. Of those who have recently retired or are about to, Miller and Payton are closest to the cutoff at 133 and 195. Mutombo, Mourning, and Webber are the closest below at 103, 105, and 119. The guys who repeatedly get nominated but not enshrined (Dantley and DJ) are right on the edge. This is a list of the highest HOF scores for retired players who entered the league from 1976/77:
Code:
1 Michael Jordan 402
2 Karl Malone 350
3 Magic Johnson 302
4 Larry Bird 286
5 Hakeem Olajuwon 282
6 David Robinson 248
7 Charles Barkley 244
8 John Stockton 233
9 Scottie Pippen 209
10 Gary Payton 195
11 Patrick Ewing 183
12 Clyde Drexler 172
13 Robert Parish 162
14 Dominique Wilkins 154
15 Isiah Thomas 149
16 Kevin McHale 137
17 Reggie Miller 133
18 James Worthy 122
19 Adrian Dantley 120
20 Alex English 120
21 Dennis Johnson 118
22 Alonzo Mourning 105
23 Kevin Johnson 104
24 Sidney Moncrief 102
25 Joe Dumars 99
26 Chris Mullin 97
27 Tim Hardaway 97
28 Shawn Kemp 96
29 Horace Grant 95
30 Dennis Rodman 92
31 Mitch Richmond 90
32 Marques Johnson 90
33 Maurice Cheeks 89
34 Walter Davis 89
35 Jack Sikma 89
36 Bernard King 87
37 Larry Nance 84
38 Jeff Hornacek 81
39 Tom Chambers 80
40 Mark Price 76
41 Mark Aguirre 76
42 Terry Porter 75
43 Terry Cummings 75
And another for active players:
Shaquille O'Neal 327
Tim Duncan 263
Kevin Garnett 210
Kobe Bryant 208
Dirk Nowitzki 153
Jason Kidd 151
Steve Nash 126
Tracy McGrady 124
Chris Webber 119
Grant Hill 103
Dikembe Mutombo 103
Ray Allen 101
Ben Wallace 97
Vince Carter 92
Paul Pierce 91
Shawn Marion 84
Chauncey Billups 83
Rasheed Wallace 83
Sam Cassell 81
LeBron James 76
Elton Brand 74
Dwyane Wade 72
Jermaine O'Neal 71
Stephon Marbury 69
Eddie Jones 66
Joined: 14 Jan 2005 Posts: 1888 Location: Delphi, Indiana
Posted: Tue Mar 11, 2008 4:47 am Post subject:
Chicago76 wrote:
...
Round 1 losers = no pts
Round 2 losers = no pts
Round 3 (CF) losers = 3 pts * % of total team plyf min played
Round 4 (champ) losers = 4 pts * % of total team plyf min played
Round 5 (champ) winners = 5 pts * % of total team plyf min played
This represents about 20% of total HOF pts for the top 50 players. The top 5 beneficiaries of this component were Grant, Rodman, DJ, Worthy, and Cheeks.
I expect it would be >20% for Robert Horry. Do you know his HOF score?
(I'm not pushing Horry, just curious how this 'playoff specialist' stacks up against a Richmond, or a Nance.)
How about Cliff Robinson?
The point scale above looks like it was set up to allow 1 and 2 points for 1st and 2nd-round eliminees; and then they get zero. An individual may play brilliantly in those early rounds.
They represent about 3/4 of all playoff games.
Sympathies aside, your gut is probably right: Hall of Fame is about Fame, in fact. The casual fan doesn't read past the headline. Great job; thanks for sharing. _________________ `
There's no I in analysys.
Horry was the most extreme example of a player accumulating HOF pts from playoffs. I neglected to mention that the playoffs were weighted twice as heavily as th other two metrics to get to a 20% average weight for this measure:
PER-adjustment: 9.4
Playoffs: 46.6
Awards: 0.0
Total 56
That puts Horry w/ Blackman, Perkins, Strickland, Smits and Alvin Robertson, which is about halfway were you need to be to get decent consideration. It's wouldn't put him in those guys' league necessarily, but this isn't supposed to be a ranking, just a HOF score.
I only considered using the first two rounds of playoff data when I set up the pt system, but like you've summed up, I decided against using them because this is about Fame more than anything else. There are a lot of opening round series, and joe sixpack isn't going to watch all 8 of them, nor is he necessarily going to find time to watch the 4 semifinal matchups. Conference finals and the Championship series get higher viewership and more media coverage. The team that makes a decent run to a conference finals will also have a good amount of retrospective coverage on how they got there. There are always some great playoff stories in the earlier rounds (Millers 8 pts in <9 seconds, Denver upsetting Seattle, Wilkins vs. Bird), but the vast majority of the big moments that stay with us 5 years after a player retires come from the Conference Finals on.
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum