2013-14 Rookies

[Mike G]

After 50 NBA games, about 4% of the season, rookies are on a pace to total 34 eWins.

Code: Select all

e82    per36 rates    tm   Min  Eff%   Sco   Reb   Ast   Stl   TO   Blk   e484
10.1  M C-Williams   Phl   35  .516   19.5   5.0   6.9   3.4   2.2   .0   1.71
6.4   Oladipo,Victor Orl   28  .495   18.0   6.4   5.2   2.2   4.7   .9   1.34
4.2   Wolters,Nate   Mil   32  .448   11.7   2.6   7.0    .4   1.4   .4    .79
2.0   Mekel,Gal      Dal   13  .650   10.9   3.4   9.5    .0   3.8   .8    .92
2.0   Adams,Steven   Okl   17  .480    8.2  10.3   1.3    .0    .6  1.3    .72

1.9 Caldwell-Pope,Ke Det   23  .500   18.2    .9   2.4   1.5   1.2   .0    .74
1.8 Christmas,Dionte Phx   11  .650   23.4   9.1    .0   1.5   1.3   .0   1.48
1.3   Mclemore,Ben   Sac   20  .446   14.0   3.1   1.1    .6   1.9   .6    .39
1.2   Zeller,Cody    Cha   15  .375   11.1   7.6    .0   1.5   2.5   .7    .47
0.6   Hill,Solomon   Ind    9  .667    9.9   8.0   2.3    .0   2.9  1.9    .67

0.5   Onuaku,Arinze  NOP   13  .300    3.7   9.8   3.7    .0   1.2   .0    .39
0.5   Hardaway,Tim   NYK   22  .389   11.3   2.2   1.4    .0    .3   .0    .14
0.5   Davies,Brandon Phl    7  .462   12.5   2.2   2.0   4.5    .0   .0    .84
0.5   Len,Alex       Phx   11  .500    8.5   4.6    .0   1.5   2.6  3.1    .41
0.5   Calathes,Nick  Mem    6  .500   11.5   9.5   5.7    .0   2.2   .0   1.00

[steveshea]

Do past seasons' data indicate significant improvement (in general) in eWins for rookies over the course of the season? If so, is it from any particular input in the formula? Any indication of a fatigue factor in the second half?

Thanks

[Mike G]

In general, rookies will play more minutes as the season goes on. At almost every update, the rookie crop's projected eWins will increase. This is mostly due to players who aren't ready for prime time at the beginning of the season, but who eventually "get it". There's also the possibility of "tanking", in which less than the best players are thrown onto the floor.

Those rookies who hit the floor running and get big minutes early: most of them will falter a bit or a lot, at some point in the season. It may be a midseason slump, followed by a fast finish. Or hit a wall and they're useless in the playoffs. Every situation is unique.

[Crow]

If EFF is eEFG%, only 4 of the 15 rookies are above .500 (the league average is probably higher than that) and only Carter-Williams meets that standard and plays 15+ minutes yet. About half are even less efficient than Oladipo so far.

[Mike G]

Eff% = effective shooting percent = Pts/(2*FGA + FTA)
It's generally .002 to .010 less than TS%, depending on FTA/FGA
League avg is .531, at the time of the OP. This is close to historic avg over 25 years or so.
Last year, rookies shot .505 eff%

League eFG% is .493 at this time *
http://www.basketball-reference.com/lea ... _2013.html
Last year, it was .496

* EDIT Nov. 8 -- League eFG% now up to .494

[Mike G]

Rookies are looking pretty dismal so far. Still checking in at 35 eWins on the year, about half of a normal year's total.

Code: Select all

e82   per36 rates     tm   Min   Eff%   Sco   Reb   Ast   Stl   TO   Blk  e484
10.1 MCarter-Williams Phl   35  .493   18.9   5.1   6.6   3.1   2.0   .2  1.68
5.5   Oladipo,Victor  Orl   27  .528   17.7   5.4   4.6   2.2   4.7   .7  1.20
5.1   Wolters,Nate    Mil   32  .476   11.1   4.3   7.1    .3   1.1   .3   .95
2.6   Adams,Steven    Okl   18  .545    8.6  11.2    .9    .9    .4  1.3   .89
2.6   Mclemore,Ben    Sac   21  .500   15.4   4.6   1.2   1.2   1.2   .4   .73

1.8   Mekel,Gal       Dal   15  .500    8.4   4.1   7.8    .4   2.8   .4   .69
1.5 Caldwell-Pope,Ke  Det   16  .563   18.8    .7   1.9   1.3   1.1   .0   .71
1.5   Zeller,Cody     Cha   16  .483   11.4   7.4   1.0   1.1   2.3   .4   .54
1.0  Christmas,Dionte Phx   12  .625   13.9   8.4    .0   1.6   1.4   .0   .85
0.6   Hill,Solomon    Ind    7  .643   11.2   5.9   2.4    .0   1.6   .9   .59

0.3   Bennett,Anthony Cle   13  .114    1.7   7.9    .4   2.8    .0   .9   .15
0.3   Davies,Brandon  Phl    7  .462   10.7   1.9   1.7   4.0    .0   .0   .59
0.2   Onuaku,Arinze   NOP   13  .300    3.3   9.2   3.4    .0   1.2   .0   .29
0.1   Datome,Luigi    Det    7  .143    3.3   7.1   2.2   2.1    .0   .0   .21
0.1   Hardaway,Tim    NYK   20  .368   10.0   2.1   1.5    .0    .3   .0   .03
0.1   Len,Alex        Phx   11  .500    7.4   4.2    .0   1.4   2.4  2.7   .07

This is everyone with a positive eW rate. Only about 10 guys worth playing.

[v-zero]

Would you consider posting eWins transformed into either wins or point margin? I like to look at these, but they're difficult to interpret...

[Mike G]

eWins are already transformed into wins. They're equivalent wins for an average team. Unlike WS or the like, not everyone for a good team is good, and below average teams aren't full of terrible players.

Is a team that scores 95% of their opponent's points considered to be 95% as good? Well, they'll only win about 1/3 of their games. So are they only half as good?
eWins splits the difference between these interpretations.

If you want to know a team's expected wins based on their eWins, it's this:
xW = eW*2 - G/2
A player's eWins are constant from good team to bad team, and there's no known purpose for just declaring a player to have created wins in a vacuum. Is there?

NBA avg is 1.00 eWins per 484 minutes -- the avg player-minutes in 1 game.
Zero eW/484 is a player who adds no wins -- a replacement level player, perhaps.

[v-zero]

eWins are already transformed into wins. They're equivalent wins for an average team. Unlike WS or the like, not everyone for a good team is good, and below average teams aren't full of terrible players.

Is a team that scores 95% of their opponent's points considered to be 95% as good? Well, they'll only win about 1/3 of their games. So are they only half as good?
eWins splits the difference between these interpretations.

If you want to know a team's expected wins based on their eWins, it's this:
xW = eW*2 - G/2
A player's eWins are constant from good team to bad team, and there's no known purpose for just declaring a player to have created wins in a vacuum. Is there?

NBA avg is 1.00 eWins per 484 minutes -- the avg player-minutes in 10 games.
Zero eW/484 is a player who adds no wins -- a replacement level player, perhaps.

I honestly have no idea what you are trying to say here. If eWins need to be transformed in order to predict wins then they do not represent wins produced for an average team. This is not an attack, there's no need to jump to being defensive.

[Mike G]

Rest assured, you aren't the first person to not get this.

On an average team, eWins = xWins (expected wins). A team that totals 41 eWins for a full season would be expected to win 41 games.
A team whose players total 51 eWins would have xWins = 61.
A team with 31 eWins would win around 21 games.

The team that wins 61 could be said to be ~3 times as good as the 21-win team.
But the players for the good team aren't 3X as good, on average. They won't get paid 3x as much, nor score 3X as many points, obviously.

But if the avg player for one team is said to be 3X as good at producing wins -- just because in that season he was on a much better team -- then you've only got a measure of the team assigned to the player.

When statistical production is simply scaled to team and opponent points and rebounds, all players can be measured on a level playing field, as far as the strength of their teams being no influence. Then when a player is traded, or others join or leave his team, his team's wins can still be predicted based on players' previous eWin rates.

[Mathketball]

It will be interesting to see how this rookie class progresses. Obviously not many big contributors right now but 4 of the top 10 picks haven't touched the court yet in Porter, Burke, Noel (and he won't this year), and McCollum. I'll be interested to see how these guys stack up when they hit the floor.

Keep of the good work Mike!

[v-zero]

Rest assured, you aren't the first person to not get this.

On an average team, eWins = xWins (expected wins). A team that totals 41 eWins for a full season would be expected to win 41 games.
A team whose players total 51 eWins would have xWins = 61.
A team with 31 eWins would win around 21 games.

The team that wins 61 could be said to be ~3 times as good as the 21-win team.
But the players for the good team aren't 3X as good, on average. They won't get paid 3x as much, nor score 3X as many points, obviously.

But if the avg player for one team is said to be 3X as good at producing wins -- just because in that season he was on a much better team -- then you've only got a measure of the team assigned to the player.

When statistical production is simply scaled to team and opponent points and rebounds, all players can be measured on a level playing field, as far as the strength of their teams being no influence. Then when a player is traded, or others join or leave his team, his team's wins can still be predicted based on players' previous eWin rates.

Ignoring the attack on metrics which sum to point differential (I can't be bothered to argue), tell me where I'm wrong here.

From what you've said:

Team wins = (sum of team player eWins)*2 - games/2

Now, that is an affine transformation, which is a rather wonderful thing, because we can easily just then say this:

Player wins = ((player minutes)*e484/484)*2 - ((player minutes)/484)

Then we can also do something like this:

Player margin per 48 = (48*(Player wins)/(player minutes) - 48/484)*30

Here the multiplier of 30 comes from my own studies and those of others.

Anyway, unless you can find some issue with the above then I cannot see your argument to not post such values, just as most do.

[v-zero]

Code: Select all

e82 	 per36 rates     tm 	 Min 	 Eff%	   Sco 	 Reb 	 Ast 	 Stl 	 TO 	 Blk 	e484	wp48
10.1	MCarter-Williams Phl 	 35 	.493 	 18.9 	 5.1 	 6.6 	 3.1 	 2.0 	 .2 	1.68	0.23
5.5 	 Oladipo,Victor  Orl 	 27 	.528 	 17.7 	 5.4 	 4.6 	 2.2 	 4.7 	 .7 	1.20	0.14
5.1 	 Wolters,Nate    Mil 	 32 	.476 	 11.1 	 4.3 	 7.1 	  .3 	 1.1 	 .3 	0.95	0.09
2.6 	 Adams,Steven    Okl 	 18 	.545 	  8.6 	11.2 	  .9 	  .9 	  .4 	1.3 	0.89	0.08
2.6 	 Mclemore,Ben    Sac 	 21 	.500 	 15.4 	 4.6 	 1.2 	 1.2 	 1.2 	 .4 	0.73	0.05
											-0.10
1.8 	 Mekel,Gal       Dal 	 15 	.500 	  8.4 	 4.1 	 7.8 	  .4 	 2.8 	 .4 	0.69	0.04
1.5	Caldwell-Pope,Ke  Det 	 16 	.563 	 18.8 	  .7 	 1.9 	 1.3 	 1.1 	 .0 	0.71	0.04
1.5 	 Zeller,Cody     Cha 	 16 	.483 	 11.4 	 7.4 	 1.0 	 1.1 	 2.3 	 .4 	0.54	0.01
1.0 	Christmas,Dionte Phx 	 12 	.625 	 13.9 	 8.4 	  .0 	 1.6 	 1.4 	 .0 	0.85	0.07
0.6 	 Hill,Solomon    Ind 	  7 	.643 	 11.2 	 5.9 	 2.4 	  .0 	 1.6 	 .9 	0.59	0.02
											-0.10
0.3 	 Bennett,Anthony Cle 	 13 	.114 	  1.7 	 7.9 	  .4 	 2.8 	  .0 	 .9 	0.15	-0.07
0.3 	 Davies,Brandon  Phl 	  7 	.462 	 10.7 	 1.9 	 1.7 	 4.0 	  .0 	 .0 	0.59	0.02
0.2 	 Onuaku,Arinze   NOP 	 13 	.300 	  3.3 	 9.2 	 3.4 	  .0 	 1.2 	 .0 	0.29	-0.04
0.1 	 Datome,Luigi    Det 	  7 	.143 	  3.3 	 7.1 	 2.2 	 2.1 	  .0 	 .0 	0.21	-0.06
0.1 	 Hardaway,Tim    NYK 	 20 	.368 	 10.0 	 2.1 	 1.5 	  .0 	  .3 	 .0 	0.03	-0.09
0.1 	 Len,Alex        Phx 	 11 	.500 	  7.4 	 4.2 	  .0 	 1.4 	 2.4 	2.7 	0.07	-0.09

.

As an example here is your most recent table in this thread with the addition of a wins per 48 minutes column. Adding up those wins for a set of players (scaled to minutes) is entirely equivalent to summing their eWins and performing your calculation, but it allows you to pull things apart nicely. An average 48.4 player minutes in the NBA is worth 0.1 wins, so that gives us the base line with which to work and makes your numbers directly comparable to WS48 numbers.

[Mike G]

It looks as though everyone below .50 eW/484 -- a level that is half (.50) as good as average, and about 2/3 of the median -- is below zero in your wp48?
Is that intentional? Inevitable?

I'm not discerning any "attack" by anyone at anything. Maybe I'm just naive?
For now, I'm not looking up "affine transformation". Perhaps others also don't know what that means.

[v-zero]

Sorry, I didn't mean to throw in unexplained maths speak. An affine transformation is the sister to a linear transformation. When we are thinking in terms of simple formulas then a linear transformation is one of the form f(x) = Kx, where K is some multiplicative constant. A linear transform maintains the origin (zero stays in the same place). An affine transform is one of the form f(x) = Mx + B where both M and B are constants. It is trivial to see that all linear transforms are affine transforms (let B = 0), but it is not generally the case that an affine transform maintains the origin.

However, the key point is that a linear transform has the property f(sum of x) = sum of f(x), whereas an affine transform does not, because of the intercept (B) value. However, it is possible in general to create some additional affine transform (g(x)) which will have the property f(sum of x) = sum of g(x), and g(x) will be unique, the only affine transform having the required property.

In this situation your function for the sum of team eWins is f(sum of x), and I have then provided the g(x) function.

Negative wins are not intentional, they are inevitable, they are a property of your ratings, everything I have posted is a property of you ratings (other than the 30 multiplier for margin production, which doesn't enter in to win production), and I hasten to add that these are really your wp48 values, not mine. There is no other equation you could use to properly show the win production of players given your eWins transformation for calculating team wins - you'd have to change your team wins transformation if you wanted to change the player wins values.

To reiterate, all I have done is take what your ratings do at a team level and bring it to a player level.