APBRmetrics

My concern with Weber isn't limited data (he only missed something like 2 games to injury), but rather that he isn't in consideration for teams due to the fact that he will be unable to play next season.

I'll write up the official summary soon, probably tonight, but just so everyone here is clear how to best interpret my results:

My numbers are based on 4 models, a regression and a bagged neural network (to help with stability) trained on two slices of data, both all historical prospects and just those prospects that enter and play in the NBA. I use high school rank, combine measurements and tests (not shooting), paced per 40 box score stuff, age/height/weight/wingspan (combine if available, listed if not), mins, and 3pt%pts. I average an entire player's career, weighted by minutes played. The last year/two years only get weighted more if the player played more. For the vast amount of missing data for the players who do not participate in the combine, I impute regression based estimates of body dimensions (hand length, body fat, etc) based on listed height and weight (body dimensions are mostly very easy to estimate, for obvious reasons), and impute missing verticals and agility via decision trees based on a player's age and body dimensions.

Thus, as layne and crow have already pointed out, youre gonna see some love for the highly heralded high schoolers.

Just a reminder that tomorrow is May 20th - the day each contributor's write up is due.

If people need to make further edits after tomorrow but before the 25th, that will likely be ok. I just need to start getting the full article/write up organized and making everything look uniform and cohesive.

While there is no template, here is what I wrote to one poster:
a quick bio followed by what your model is conceptually trying to display (and why this is important), how it actually works (any problems it might have), players who do really well in it/players who fair really poorly in it (and why).
I suggest making the write up short, and if possible, provide a link to your website where readers can learn more about you/the model.

ampersand5 wrote:Just a reminder that tomorrow is May 20th - the day each contributor's write up is due.

If people need to make further edits after tomorrow but before the 25th, that will likely be ok. I just need to start getting the full article/write up organized and making everything look uniform and cohesive.

While there is no template, here is what I wrote to one poster:
a quick bio followed by what your model is conceptually trying to display (and why this is important), how it actually works (any problems it might have), players who do really well in it/players who fair really poorly in it (and why).
I suggest making the write up short, and if possible, provide a link to your website where readers can learn more about you/the model.

I'm sorry I won't be able to make that deadline. I still plan on being involved, but it'd be closer to week 2 of June if it happens.

I really appreciate your helping pull together these different models - hopefully the info you pull together will be a real go to for fans come draft time (twitter?) who are looking for something much more than mocks that are merely trying to mirror the draft as oppose to ranking based on projected relative pro production/performance.

Are all the dependent variables going to be specified in the write up? For the 2 listings shared, the DV has not yet been specified in public as part of this project. Perhaps earlier, elsewhere? But would be helpful to see, conveniently.

I have several models, but in terms of sheer expectation, this is the one I trust the most:

OFF DEF SCORE:
1) Justise Winslow 1.62 1.38 3.00
2) Karl Towns 1.52 1.49 3.00
3) Tyus Jones 1.80 1.19 2.99
4) Stanley Johnson 1.62 1.36 2.98
5) D'Angelo Russell 1.76 1.19 2.95
6) Jahlil Okafor 1.45 1.50 2.94
7) Kevon Looney 1.52 1.40 2.92
8) Myles Turner 1.50 1.43 2.92
9) Frank Kaminsky 1.53 1.35 2.88
10) Bobby Portis 1.49 1.38 2.87
11) Delon Wright 1.62 1.23 2.85
12) Willie Cauley-Stein 1.35 1.50 2.85
13) Rondae Hollis-Jefferson 1.48 1.36 2.84
14) Christian Wood 1.44 1.40 2.84
15) Trey Lyles 1.46 1.38 2.84
16) Dakari Johnson 1.33 1.49 2.82
17) Cameron Payne 1.72 1.09 2.82
18) Chris McCullough 1.43 1.37 2.81
19) Cliff Alexander 1.31 1.50 2.80
20) Wesley Saunders 1.60 1.20 2.80
21) Aaron White 1.49 1.30 2.80
22) Kelly Oubre 1.53 1.25 2.78
23) RJ Hunter 1.64 1.14 2.77
24) Jerian Grant 1.60 1.16 2.76
25) Robert Upshaw 1.25 1.51 2.75
26) Larry Nance Jr 1.44 1.31 2.75
27) Richaun Holmes 1.34 1.41 2.75
28) Andrew Harrison 1.62 1.12 2.74
29) Sam Dekker 1.49 1.24 2.74
30) Montrezl Harrell 1.33 1.37 2.70
31) Devin Booker 1.61 1.09 2.70
32) Joshua Smith 1.28 1.42 2.69
33) Vince Hunter 1.35 1.33 2.68
34) Darrun Hilliard 1.58 1.10 2.68
35) Josh Richardson 1.56 1.11 2.67
36) Derrick Marks 1.55 1.09 2.64
37) TJ McConnell 1.54 1.10 2.64
--------------------------------------------------(my undrafted mark for my model)
38) Alan Williams 1.28 1.36 2.64
39) Branden Dawson 1.31 1.33 2.64
40) JP Tokoto 1.45 1.19 2.63
41) Jarell Martin 1.33 1.30 2.63
42) Terry Rozier 1.58 1.05 2.62
43) Justin Anderson 1.49 1.11 2.60
44) Chris Walker 1.22 1.37 2.59
45) Rashad Vaughn 1.50 1.09 2.59
46) Keifer Sykes 1.55 1.03 2.58
47) Pat Connaughton 1.44 1.13 2.57
48) Quinn Cook 1.56 1.01 2.57
49) Jordan Mickey 1.20 1.35 2.55
50) Treveon Graham 1.40 1.14 2.53
51) Norman Powell 1.44 1.09 2.53
52) Michael Qualls 1.47 1.06 2.53
53) Michael Frazier 1.52 1.01 2.53
54) Brandon Ashley 1.28 1.24 2.52
55) Aaron Harrison 1.51 1.01 2.52
56) Tyler Harvey 1.52 0.99 2.51
57) Rakeem Christmas 1.13 1.37 2.50
58) Shannon Scott 1.44 1.05 2.50
59) Olivier Hanlan 1.50 0.99 2.48
60) Juwan Staten 1.42 1.07 2.48
61) Dez Wells 1.35 1.11 2.46
62) Joseph Young 1.47 0.99 2.45
63) Anthony Brown 1.39 1.06 2.45
64) Jonathan Holmes 1.25 1.19 2.44
65) DJ Newbill 1.39 1.03 2.42
66) Terran Petteway 1.31 1.09 2.40
67) Corey Hawkins 1.39 0.98 2.37


Others:
Emmanuel Mudiay 1.67 1.26 2.93
George Lucas 1.59 1.30 2.89
Seth Tuttle 1.49 1.35 2.84
Corey Walden 1.66 1.17 2.83
Alpha Kaba 1.43 1.40 2.83
Aleksandar Vezenkov 1.59 1.23 2.82
Nedim Buza 1.52 1.25 2.77
Mario Hezonja 1.59 1.17 2.76
Kristaps Porzingis 1.43 1.30 2.73
Royce O'Neale 1.56 1.15 2.71
Nikola Milutinov 1.29 1.42 2.71
Luka Mitrovic 1.49 1.22 2.71
Wael Arakji 1.54 1.16 2.70
Rayvonte Rice 1.50 1.18 2.68
Denzel Livingston 1.52 1.16 2.68
Cedi Osman 1.48 1.20 2.67
Satnam Singh Bharama** 1.30 1.38 2.67
Lucas Dias 1.44 1.22 2.66
Guillermo Hernangomez 1.32 1.34 2.65
Timothe Luwawu 1.52 1.13 2.65
Marc Garcia 1.51 1.14 2.65
Andzejs Pasecniks 1.39 1.26 2.65
Ailun Guo 1.53 1.12 2.65
Dimitris Agravanis 1.38 1.26 2.64
Simone Fontecchio** 1.48 1.15 2.63
Guillem Vives 1.52 1.11 2.63
Charles Jackson 1.20 1.42 2.62
JJ Avila 1.33 1.28 2.62
Kevin Harley 1.45 1.17 2.61
Mouhammadou Jaiteh 1.24 1.38 2.61
Rade Zagorac 1.50 1.11 2.61
Kevin Pangos 1.59 1.00 2.60
Moussa Diagne 1.25 1.35 2.60
Oriol Pauli 1.44 1.16 2.59
Adin Vrabac 1.36 1.22 2.58
Nikola Radicevic 1.49 1.09 2.58
Andrey Desyatnikov 1.222354201 1.349883936 2.572238138
Arturas Gudaitis 1.19 1.37 2.56
Briante Weber 1.46 1.09 2.54
Daniel Diez 1.37 1.16 2.53
Cady Lalanne 1.12 1.41 2.53
Paul Zipser 1.37 1.15 2.52
Beka Burjanadze 1.37 1.11 2.47
Walter Pitchford 1.24 1.22 2.46
Andre Hollins 1.43 1.00 2.43
Chad Frazier 1.39 1.03 2.42
Jherrod Stiggers 1.37 1.03 2.39
Todd Mayo 1.30 1.07 2.37
Aaron Thomas 1.28 1.08 2.36
Trevor Lacey 1.34 0.96 2.30

Lots of stat model love for Looney. How far does he rise? DX has him at 19 at the moment.

I assume some models adjust player projections based on quality of own team. I assume some may do it by specific stat (level of opportunity or need) and specific teammates they actually play with, minutes weighted. Do any adjust the age adjustment based on the quality of teammates? Is a two step process (teammates, age) an ok way to stay / go or is it potentially better to integrate them? If you are young, productive but the need is greater than normal are you getting too much credit? Or if need is less than normal, too little?


If this is not a burning analytical question, what questions do the modelers wrestle with that they are willing to talk about now or after lockdown / publication?



Any intention to compare contest projections to any other stat models in the media (Pelton, wp based, 538 if they give it a go, etc.)? Either at time of projection or at future scoring?

At scoring far enough down the road, it might be interesting to see what the optimal blend would have been for that draft class, then possibly multiple classes if models stay the same or at least guided by similar philosophy / formula. If one were sufficiently curious and skilled, it would be interesting to try to produce a machine learning derived stat model that mimics the optimal blend findings. (Since folks don't want to give up a specific formula, use computers to try to find the approximate optimal formula.)

Maybe it is too much to dream / ask but there could be fun had if DX or someone else offered a convenient interface to select and weight models to give draft enthusiasts their own custom projection blend (similar to what Layne has done with weights in player comparisons). If this is not done, could an excel spreadsheet be offered with the article to easily facilitate archiving and further independent study, analysis, manipulation?

Can you guys explain the inputs that are making Russell rate 1st over KAT? I thought KAT's combination of elite blocks, strong rebounding, efficiency and overall production for a freshman (31 PER+) would be more analytic catnip than Russell's good not amazing steals, impressive rebounding/assists for a guard, and high volume/efficiency. Is KAT hurt by his steal rate?

Sorry guys, I don't think I'll have something by the deadline. Last weekend was graduation weekend for me and I won't be getting home until the 23rd at earliest. I'll post my results here when I have them though, hopefully within the next two weeks or so. Looking forward to seeing the DX article though!

I guess I'll go first. Let me know if this is appropriate as a summary/bio. I tried my best to keep it short, general, and accessible.



My name is Nick Restifo. In my basketball life, I write for Nylon Calculus and am a special assistant for the D2 powerhouse that is the University of New Haven Chargers. If you like, you can follow me on Twitter at @itsastat.
My overall predictions are based on an ensemble of four base models predicting a two year career peak blend of RAPM and Win Shares. The four models include a regression and a bagged neural network (to help with stability) trained on two different subsets of data; all prospects with statistics listed on DraftExpress since 2001-2002 and just those prospects that were actually drafted since 2001-2002. I use standing reach, RSCI high school rank, standing vertical leap, lane agility test time, true shooting percentage, points per 40 minutes pace adjusted, total rebounds per 40 minutes pace adjusted, assists per 40 minutes pace adjusted, steals per 40 minutes pace adjusted, blocks per 40 minutes pace adjusted, turnovers per 40 minutes pace adjusted, personal fouls per 40 minutes pace adjusted, minutes per game, age on February 1st of a player’s draft year, strength of schedule, and percentage of points from three (to account for some spacing benefits). I average an entire player's pre-NBA career, each year weighted by minutes played. Unlike other models, I do not assign any extra weight to the most recent years. The most recent years only get weighed more if the player played more in those years, (and this is usually the case anyway). For the vast amount of missing data for the players who did not participate in the combine, I impute regression based estimates of body dimensions (hand length, body fat, etc) based on listed height and weight. Body dimensions are mostly very easy to estimate, for obvious reasons. For the vertical and agility tests, I impute missing values via decision trees trained on a player's age and body dimensions.
Each model in my ensemble plays an individual role conceptually. By applying two of the base models to some 20,000 plus prospects since 2002, I set the framework for my model to have the ability be applied to any basketball player anywhere, not just those who make top 100 prospects lists. For those prospects that never play in the NBA and don’t have RAPM and Win Shares values, I fill those missing values with -4 and 0 respectively, which are each very close to the absolute minimum career peaks of all NBA players since 2001-2002. Only a handful of NBA players have ever peaked at below -4 RAPM or 0 Win Shares. For those prospects without an RSCI high school rank, I fill those missing values with 600, which is a very rough estimate of what the average high school rank would be for the remaining unranked prospects each year. The problem with developing these models on every potential player is that, in conjunction with imputing all these missing values, the models become not only a reflection of NBA success, but also a reflection of whether or not a player will be drafted, which isn’t always the same thing.
To counter these effects, I trained two additional base models just on players who were drafted. Though some of these players also never played in the NBA, these models are able to get a better handle on whether or not a player will actually succeed in the NBA, more independent of the sometimes clouding effects of what will get a player drafted in the NBA.
In comparison to other models, since I include high school ranking as a variable, my model will favor those highly heralded high school players significantly more than other models. High school rank is an especially important predictor in the regression model trained on all available prospects This results in additional predicted value for the highly ranked high schoolers that might not be as favored in other models, players like Cliff Alexander and Myles Turner, and less predicted value for unranked high school players that do a little better in other models, like Frank Kaminsky. (In Frank Kaminsky’s case in particular, he does not do very well by the models trained on all prospects, but does considerably better by the models trained on just those prospects who were drafted.)
With regards to methodology, my ensemble has its strengths and weaknesses like any other prediction system. I used neural networks as part of my ensemble because they were the most accurate out of sample prediction method on my data, and accuracy is obviously valuable. Neural networks are flexible and often better than other methods like regression at teasing out complex, non-linear relationships amongst the training data, and with regards to draft prospects, neural networks are also good at capturing just how much better the premier players are than the middle class. While neural networks are powerfully accurate, they also tend to overfit training datasets and attach themselves to noise in data in their pursuit of accuracy. To alleviate these concerns, I applied a process known as bagging to my neural networks, which helps to increase the stability of the predictions by taking the consensus of several neural networks over subsets of the training data, rather than a single neural network over the complete training data, as the latter is more likely to interpret noise as signal.

Dr Positivity wrote:Can you guys explain the inputs that are making Russell rate 1st over KAT? I thought KAT's combination of elite blocks, strong rebounding, efficiency and overall production for a freshman (31 PER+) would be more analytic catnip than Russell's good not amazing steals, impressive rebounding/assists for a guard, and high volume/efficiency. Is KAT hurt by his steal rate?

Haven't updated my weight/ratings - but I would guess KAT's high foul rate will be a definite limiter in his career projections in terms of playing time in my model, especially in the early years.

In some models steal rate might hurt him.

Some models might not be adjusting much for pace, or maybe not much for teammates.

My name is Layne Vashro. I am presenting my simple “Estimated Wins Peak” (EWP) here, but I've put together a number of different projection models and tools to help evaluate incoming talent. These include several NCAA/International models, a player-season comparison finder, a tool that that shows how each statistic has historically translated to the NBA for players under different coaches, and a tool that allows you follow each prospect's progression/regression throughout the season. You can find these over at NylonCalculus under the “Our Stats” tab [http://nyloncalculus.com/stats/].

The goal of the EWP model is to project how good each prospect will be at the peak of his NBA career. In order to do that, I must quantify “peak NBA performance” in some acceptable way. I do this by calculating the number of wins a player is responsible for in each season of his career using a blend of Win Shares (box-score metric) and RAPM (+/- metric). I then use a two-year rolling average and select the highest value as that player's “wins peak”. Here is a link to the list of previously drafted players included in the sample [https://goo.gl/yYmeEQ]. If this list largely agrees with the order in which you would select these players in a redraft, you can at least be comfortable with my model's validity.

To build the model itself, I take the above list of players along with their collegiate box-score and play-by-play statistics, anthropometric information, and a selection of team statistics. I then use a regression process to identify what each bit of pre-NBA information says about a player's future peak production in the NBA. This knowledge is then applied to current prospects whose future is still unknown.

re: D'Angelo over Towns...

They are effectively tied. Steals are probably the biggest reason why Towns doesn't run away with it. I think that is something worth attending to, since high steal rates historically help separate guys like Robinson, Hakeem, Shaq, Ewing, Davis... from the rest of the big kids who beat up college opponents. (Duncan, who Towns puts up very similar numbers to, is a good exception.)

masseffectlenk wrote:I have several models, but in terms of sheer expectation, this is the one I trust the most:

OFF DEF SCORE:
1) Justise Winslow 1.62 1.38 3.00
2) Karl Towns 1.52 1.49 3.00
3) Tyus Jones 1.80 1.19 2.99
4) Stanley Johnson 1.62 1.36 2.98
5) D'Angelo Russell 1.76 1.19 2.95
6) Jahlil Okafor 1.45 1.50 2.94
7) Kevon Looney 1.52 1.40 2.92
8) Myles Turner 1.50 1.43 2.92
9) Frank Kaminsky 1.53 1.35 2.88
10) Bobby Portis 1.49 1.38 2.87
11) Delon Wright 1.62 1.23 2.85
12) Willie Cauley-Stein 1.35 1.50 2.85
13) Rondae Hollis-Jefferson 1.48 1.36 2.84
14) Christian Wood 1.44 1.40 2.84
15) Trey Lyles 1.46 1.38 2.84
16) Dakari Johnson 1.33 1.49 2.82
17) Cameron Payne 1.72 1.09 2.82
18) Chris McCullough 1.43 1.37 2.81
19) Cliff Alexander 1.31 1.50 2.80
20) Wesley Saunders 1.60 1.20 2.80
21) Aaron White 1.49 1.30 2.80
22) Kelly Oubre 1.53 1.25 2.78
23) RJ Hunter 1.64 1.14 2.77
24) Jerian Grant 1.60 1.16 2.76
25) Robert Upshaw 1.25 1.51 2.75
26) Larry Nance Jr 1.44 1.31 2.75
27) Richaun Holmes 1.34 1.41 2.75
28) Andrew Harrison 1.62 1.12 2.74
29) Sam Dekker 1.49 1.24 2.74
30) Montrezl Harrell 1.33 1.37 2.70
31) Devin Booker 1.61 1.09 2.70
32) Joshua Smith 1.28 1.42 2.69
33) Vince Hunter 1.35 1.33 2.68
34) Darrun Hilliard 1.58 1.10 2.68
35) Josh Richardson 1.56 1.11 2.67
36) Derrick Marks 1.55 1.09 2.64
37) TJ McConnell 1.54 1.10 2.64
--------------------------------------------------(my undrafted mark for my model)
38) Alan Williams 1.28 1.36 2.64
39) Branden Dawson 1.31 1.33 2.64
40) JP Tokoto 1.45 1.19 2.63
41) Jarell Martin 1.33 1.30 2.63
42) Terry Rozier 1.58 1.05 2.62
43) Justin Anderson 1.49 1.11 2.60
44) Chris Walker 1.22 1.37 2.59
45) Rashad Vaughn 1.50 1.09 2.59
46) Keifer Sykes 1.55 1.03 2.58
47) Pat Connaughton 1.44 1.13 2.57
48) Quinn Cook 1.56 1.01 2.57
49) Jordan Mickey 1.20 1.35 2.55
50) Treveon Graham 1.40 1.14 2.53
51) Norman Powell 1.44 1.09 2.53
52) Michael Qualls 1.47 1.06 2.53
53) Michael Frazier 1.52 1.01 2.53
54) Brandon Ashley 1.28 1.24 2.52
55) Aaron Harrison 1.51 1.01 2.52
56) Tyler Harvey 1.52 0.99 2.51
57) Rakeem Christmas 1.13 1.37 2.50
58) Shannon Scott 1.44 1.05 2.50
59) Olivier Hanlan 1.50 0.99 2.48
60) Juwan Staten 1.42 1.07 2.48
61) Dez Wells 1.35 1.11 2.46
62) Joseph Young 1.47 0.99 2.45
63) Anthony Brown 1.39 1.06 2.45
64) Jonathan Holmes 1.25 1.19 2.44
65) DJ Newbill 1.39 1.03 2.42
66) Terran Petteway 1.31 1.09 2.40
67) Corey Hawkins 1.39 0.98 2.37


Others:
Emmanuel Mudiay 1.67 1.26 2.93
George Lucas 1.59 1.30 2.89
Seth Tuttle 1.49 1.35 2.84
Corey Walden 1.66 1.17 2.83
Alpha Kaba 1.43 1.40 2.83
Aleksandar Vezenkov 1.59 1.23 2.82
Nedim Buza 1.52 1.25 2.77
Mario Hezonja 1.59 1.17 2.76
Kristaps Porzingis 1.43 1.30 2.73
Royce O'Neale 1.56 1.15 2.71
Nikola Milutinov 1.29 1.42 2.71
Luka Mitrovic 1.49 1.22 2.71
Wael Arakji 1.54 1.16 2.70
Rayvonte Rice 1.50 1.18 2.68
Denzel Livingston 1.52 1.16 2.68
Cedi Osman 1.48 1.20 2.67
Satnam Singh Bharama** 1.30 1.38 2.67
Lucas Dias 1.44 1.22 2.66
Guillermo Hernangomez 1.32 1.34 2.65
Timothe Luwawu 1.52 1.13 2.65
Marc Garcia 1.51 1.14 2.65
Andzejs Pasecniks 1.39 1.26 2.65
Ailun Guo 1.53 1.12 2.65
Dimitris Agravanis 1.38 1.26 2.64
Simone Fontecchio** 1.48 1.15 2.63
Guillem Vives 1.52 1.11 2.63
Charles Jackson 1.20 1.42 2.62
JJ Avila 1.33 1.28 2.62
Kevin Harley 1.45 1.17 2.61
Mouhammadou Jaiteh 1.24 1.38 2.61
Rade Zagorac 1.50 1.11 2.61
Kevin Pangos 1.59 1.00 2.60
Moussa Diagne 1.25 1.35 2.60
Oriol Pauli 1.44 1.16 2.59
Adin Vrabac 1.36 1.22 2.58
Nikola Radicevic 1.49 1.09 2.58
Andrey Desyatnikov 1.222354201 1.349883936 2.572238138
Arturas Gudaitis 1.19 1.37 2.56
Briante Weber 1.46 1.09 2.54
Daniel Diez 1.37 1.16 2.53
Cady Lalanne 1.12 1.41 2.53
Paul Zipser 1.37 1.15 2.52
Beka Burjanadze 1.37 1.11 2.47
Walter Pitchford 1.24 1.22 2.46
Andre Hollins 1.43 1.00 2.43
Chad Frazier 1.39 1.03 2.42
Jherrod Stiggers 1.37 1.03 2.39
Todd Mayo 1.30 1.07 2.37
Aaron Thomas 1.28 1.08 2.36
Trevor Lacey 1.34 0.96 2.30

Thanks.

You are missing the rankings for: Chasson Randle, Julian Washburn, Lebryan Nash, Marcus Thornton, Ryan Boatright, Tashawn Thomas and Travis Trice. If possible, can you please provide them.