ESPN/J.E.: Rookies provide no impact at all

Home for all your discussion of basketball statistical analysis.
mystic
Posts: 470
Joined: Mon Apr 18, 2011 10:09 am
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by mystic » Tue Aug 26, 2014 1:36 pm

I understand what you want to say and I agree that there will be people interpreting those words in a similar or same fashion as you do/did, but the issue here is that this is only possible, if you skip the first paragraph of the article in which RPM is explained to some degree (at least good enough to interpret those values correctly as values compared to the value of an average player; in fact it is explicitely stated "It suggests he added about 5 more points per game to his team's scoring margin than the league average player."). Ignoring that will ultimately lead to a wrong interpretation of the phrases in the following paragraphs.

And the interpretation is by no means correct, if you say that it would mean "hurting more than a better player would". The correct interpretation is: they are hurting more than an average player would. That should be logical based on the fact that RPM is a value where the average player has 0. So, and it should also be clear, that a negative value will ultimately lead to more losses than wins, which is certainly the reason the authors choose the word "hurting" here, so in that sense his net contribution in comparison to an average player is negative. That does not mean that each of the rookies is "hurting more" than every possible replacement, but as I see it, that's how you interpret those words (given your example of the Cavs 2003 vs. Cavs 2004).

Also, I don't see any reason to believe that the authors intention is to put blame on players/coaches, but rather trying to explain the overall value of rookies in comparison to league average players. You seem to try to read something between the lines, which isn't there. And at that point it should be noted that the main reason for playing rookies is not necessarily the prospects of them helping the team win more games in their first season, but rather that experience will help those players to develop into someone being able to help a team win more games in comparison to an average player (or at least contribute above typical replacement level). There is also an impact by the fact that "hope" is sold to the fans by playing such young players, because somehow fans seems to value potential in cases of bad teams more than actual performance level, thus prefering the young and supposedly talented player while dismissing the better veteran player. But that isn't the point of that article anyway ...

nbo2
Posts: 42
Joined: Thu Apr 25, 2013 4:28 am
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by nbo2 » Tue Aug 26, 2014 4:18 pm

All of this is mostly due to the rookie prior in RPM. That being said, many rookies are indeed worse than a minimum salary player (roughly 2 pts per 200 possessions below avg).

fpliii
Posts: 82
Joined: Fri May 10, 2013 1:38 pm

Re: ESPN/J.E.: Rookies provide no impact at all

Post by fpliii » Tue Aug 26, 2014 7:29 pm

nbo2 wrote:All of this is mostly due to the rookie prior in RPM. That being said, many rookies are indeed worse than a minimum salary player (roughly 2 pts per 200 possessions below avg).
Hm, I don't think there is a distinct rookie prior in RPM in particular (could be wrong, though).

talkingpractice
Posts: 194
Joined: Tue Oct 30, 2012 6:58 pm
Location: The Alpha Quadrant
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by talkingpractice » Wed Aug 27, 2014 9:57 pm

imo, a lot of people have attributed about 234923423 things to je here that he either didnt really say, didnt really mean, or both. all he was really saying is that rookies, almost always, tend to suck. everything else is just sorta conversation of particulars and semantics that don't matter that much.

Mike G
Posts: 4413
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: ESPN/J.E.: Rookies provide no impact at all

Post by Mike G » Wed Aug 27, 2014 10:43 pm

It's to be expected. A young player tends to declare for the draft when he thinks he can be a competitive NBA player. This may not be yet an 'average' player -- around the 120th to 150th best -- though he hopes to be there within a year or two.

Being a bit short of average, though, is a far cry from "hurting his team more than he's helping them". If he's not starter quality, he may be a valued reserve. Again, it depends on the depth and strength of the team. You don't have guys in the rotation so that they can "hurt" you.

mystic
Posts: 470
Joined: Mon Apr 18, 2011 10:09 am
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by mystic » Thu Aug 28, 2014 9:50 am

nbo2 wrote:All of this is mostly due to the rookie prior in RPM.
How can that be the reason, if that "That being said, many rookies are indeed worse than a minimum salary player (roughly 2 pts per 200 possessions below avg)." is true? The rookie prior has an effect on the value for the individual player (some may be "overrated" other might be "underrated" due to the choosen prior), but in average the prior has only a negative effect, if the prior value is lower than the average value of a rookie. So, and the prior isn't selected arbritarily, but based on the fact (as you said) that rookies are in fact in average not that good. Assuming the average value (talking about minute- or possession-weighted average here) of a rookie is indeed "roughly -2" and the choosen prior in RPM for rookies is -2 (not quite sure whether that is the case; this would need to be answered by J.E.), thus the rookie prior in RPM has no effect on the overall conclusion!

If I compare the average value of the #1 picks in my metric with their respective RPM values from 2001 to 2014, they ended up being about 0.2 standard deviations better in my metric than in RPM. For all rookies it ends up about the same in both metrics. Nonetheless, the overall point of the article is also supported by my metric (and I guess by basically every statistical measurement too).
Mike G wrote: Being a bit short of average, though, is a far cry from "hurting his team more than he's helping them".
Again, you can't ignore the reference to RPM here, in which a player's value is compared to the value of an AVERAGE player! Thus, a player below average will hurt a team more than he is helping them in comparison to what an average player would do! Really, you are arguing a strawman here ...

nbo2
Posts: 42
Joined: Thu Apr 25, 2013 4:28 am
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by nbo2 » Sat Aug 30, 2014 5:24 am

How can that be the reason, if that "That being said, many rookies are indeed worse than a minimum salary player (roughly 2 pts per 200 possessions below avg)." is true? The rookie prior has an effect on the value for the individual player (some may be "overrated" other might be "underrated" due to the choosen prior), but in average the prior has only a negative effect, if the prior value is lower than the average value of a rookie. So, and the prior isn't selected arbritarily, but based on the fact (as you said) that rookies are in fact in average not that good. Assuming the average value (talking about minute- or possession-weighted average here) of a rookie is indeed "roughly -2" and the choosen prior in RPM for rookies is -2 (not quite sure whether that is the case; this would need to be answered by J.E.), thus the rookie prior in RPM has no effect on the overall conclusion!
Because due to the nature of ridge regression, the actual best players who perform well in limited sample size will be underrated. Using a better prior that factors in more information might help, but since RPM's formula isn't public, it's hard to tell how much is due to the rookie prior and how much is due to rookies just being bad in general. If J.E. can elaborate on the rookie prior that would probably help.

mystic
Posts: 470
Joined: Mon Apr 18, 2011 10:09 am
Contact:

Re: ESPN/J.E.: Rookies provide no impact at all

Post by mystic » Sat Aug 30, 2014 12:18 pm

nbo2 wrote: Because due to the nature of ridge regression, the actual best players who perform well in limited sample size will be underrated.
True, but that neither was in dispute nor would that be an explanation why "All of this is mostly due to the rookie prior in RPM.". Only if the rookies as a group would be within the group of "actual best players", their average value would be negatively effected by the choosen prior (assuming the used prior in RPM for rookies isn't significant lower than the actual average playing level of the rookies). Otherwise, and adapting your wording here, due to the nature of ridge regression the actual worst players who perform bad in limited sample size will be overrated. Now, if your statement "All of this is mostly due to the rookie prior in RPM." is supposed to be correct, you must believe that in average the rookies do not belong into the group of "worst players", but rather among the group of "best players". Using your next statement then (That being said, many rookies are indeed worse than a minimum salary player (roughly 2 pts per 200 possessions below avg).) can only mean that you are contradicting yourself. Either the rookies usually are in the group of "better" players (better = better than average, thus getting rather "underrated, due to the nature of ridge regression"), which would mean your first statement is correct, but your second is wrong; or the rookies are belonging usually to the group of "worse" players (worse = worse than average, thus getting rather "overrated due to the nature of ridge regression"), which would make your second statement correct and your first statement wrong. I would think, the latter is the much more likely scenario ...

Again, an individual rookie might be over- or underrated by the method overall or specifically due to the rookie prior, but for the whole group of rookies that rookie prior will likely have only little to no effect on their average value. That would also be in agreement with the conclusion based on no-prior informed RAPM values for that group of rookies.
nbo2 wrote: Using a better prior that factors in more information might help, but since RPM's formula isn't public, it's hard to tell how much is due to the rookie prior and how much is due to rookies just being bad in general. If J.E. can elaborate on the rookie prior that would probably help.
So, in conclusion: You don't have a clue what's going on in RPM, thus that statement, that "All of this is mostly due to the rookie prior in RPM" was pulled out of your rear end ... at least that's how I see it.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: ESPN/J.E.: Rookies provide no impact at all

Post by schtevie » Sun Aug 31, 2014 5:25 pm

Sidestepping the discussion on the meaning and implication of the word "average" (some "debates" are impossibly endless) I would like to pose a question (or two) to the practitioners of the dark arts of manipulating +/- data, regarding rookie value and the related issue of aging cuves.

In generating such estimates, is account taken of the systematic effect of the correlation of more talented players (in expectation) starting their careers with worse teams?

I think I can tell a plausible story (or two) why a rookie drafted by a "bad" team, all else equal, might realize a xRAPM estimate that is lower than were he starting his career for a "good" team. First, the better team might be better able to use the rookie/sophomore in situations where he would be less likely to "fail". And second. the nature of the +/- "production function" for teams might well be different across the quality spectrum, where a given "amount" of rookie/sophomore skill yields lower +/- returns on a bad team vs. a good team.

Now, whether these stories are true and/or the effect is empirically interesting is another issue, but the first question is have such potential effects been considered?

Post Reply