Posted by schwarze on 2/10/2016 2:17:00 PM (view original):
I am having difficulty proving that the salary changes have a net effect of zero.  I'm not sure how they are handling players with multiple teams in the same year.  If I assume they only count one combined version of that player, then I get the following results...

39,858 Total Hitters
    9,506 increase salary  (+105,507,108 million or 11,099 per player)
  29,829 decrease salary  (-108,333,183 million or -3,632 per player)
       523  same salary
Total salary difference: -$2,826,075

28,585 Total Pitchers
    5,149 increase salary   (+58,312,570 million or 11,325 per player)
  23,347 decrease salary   (-52,316,848 million or 2,241 per player)
         89 same salary
Total salary difference:  +$8,717,737

Overall salary difference: +$5,891,662

If I include all versions of multi-team players, then the overall salary difference is -$26,065,742 (so this clearly isn't what they are doing).

Another thought as to why the overall difference may not be zero.  Perhaps the "zero" refers only to the dynamic changes (players moving up and down based on usage) and the OTHER changes they made (renormalizing the Astros to the AL, downward adjusting some catcher arms) are separate from that?

Posted by contrarian23 on 2/12/2016 8:50:00 AM (view original):

Posted by schwarze on 2/10/2016 2:17:00 PM (view original):
I am having difficulty proving that the salary changes have a net effect of zero.  I'm not sure how they are handling players with multiple teams in the same year.  If I assume they only count one combined version of that player, then I get the following results...

39,858 Total Hitters
    9,506 increase salary  (+105,507,108 million or 11,099 per player)
  29,829 decrease salary  (-108,333,183 million or -3,632 per player)
       523  same salary
Total salary difference: -$2,826,075

28,585 Total Pitchers
    5,149 increase salary   (+58,312,570 million or 11,325 per player)
  23,347 decrease salary   (-52,316,848 million or 2,241 per player)
         89 same salary
Total salary difference:  +$8,717,737

Overall salary difference: +$5,891,662

If I include all versions of multi-team players, then the overall salary difference is -$26,065,742 (so this clearly isn't what they are doing).

Another thought as to why the overall difference may not be zero.  Perhaps the "zero" refers only to the dynamic changes (players moving up and down based on usage) and the OTHER changes they made (renormalizing the Astros to the AL, downward adjusting some catcher arms) are separate from that?

Seems reasonable.

Posted by contrarian23 on 2/12/2016 8:50:00 AM (view original):

Posted by schwarze on 2/10/2016 2:17:00 PM (view original):
I am having difficulty proving that the salary changes have a net effect of zero.  I'm not sure how they are handling players with multiple teams in the same year.  If I assume they only count one combined version of that player, then I get the following results...

39,858 Total Hitters
    9,506 increase salary  (+105,507,108 million or 11,099 per player)
  29,829 decrease salary  (-108,333,183 million or -3,632 per player)
       523  same salary
Total salary difference: -$2,826,075

28,585 Total Pitchers
    5,149 increase salary   (+58,312,570 million or 11,325 per player)
  23,347 decrease salary   (-52,316,848 million or 2,241 per player)
         89 same salary
Total salary difference:  +$8,717,737

Overall salary difference: +$5,891,662

If I include all versions of multi-team players, then the overall salary difference is -$26,065,742 (so this clearly isn't what they are doing).

Another thought as to why the overall difference may not be zero.  Perhaps the "zero" refers only to the dynamic changes (players moving up and down based on usage) and the OTHER changes they made (renormalizing the Astros to the AL, downward adjusting some catcher arms) are separate from that?

Have the hitters and pitchers been separated here, or is this figured as 14,655 players with increased salaries and 53,176 decreased, with 612 remaining the same? 

The reason I ask is because it might change the math a bit that way. There are nearly 40,000 more players listed in this who's salaries decreased than those who increased...Surely that must move the scale just by the volume of decreases, even if individually, they are smaller than the increases.
Example, if 4 salaries were to go up by $5m apiece, that would be $20m...If 8 were to drop by just 2.5m, that would also be 20m. 

Has the volume been accounted for here? (I hope the way I worded all of this makes sense.) 

Thanks for doing what had to have been an incredible amount of research on this!!!!!

Yes - the volume has been accounted for. 

I too think the difference might be due to all those non-usage related changes. 

We'll obviously get a better idea of the validity of that assumption when the next round of adjustments comes out.

OK...I have noticed something that I don't understand.

Before the recent salary change, it was the case that players traded within the same league would have 2 versions of their combined stats, one for each team, and those versions would have the same salary. The only exceptions to this that I was aware of were if one of the teams played a different number of total games that season.

This no longer seems to be true.

Take 1979 Oscar Gamble. His combined versions for both Texas and New York used to cost $3,523,308. Now the Texas version costs $3,542,661 and the New York version costs $3,559,527.

I assume there is no statistical difference between the two players...all of his numbers, both actual and normalized, show up exactly the same. So there is no reason to believe there is ANY performance advantage from having one version of this player over another. This assumption does not hold if the player was traded between leagues, obviously, because the normalization would be different, so I am only interested here in players traded within the same league.

And I assume that WIS has not done anything new to the formula, like normalize based on park, as this would be a major change and would presumably have been accompanied by some documentation.

So the only thing I can come up with is that when they did the salary adjustment, they must have considered these versions separately. And their data must have showed that the NY version of Gamble was chosen slightly more often than the Texas version, and so received a slightly greater price increase.

If true, I think this represents a slight flaw in their methodology...and that in the future they should consider these versions to be identical. It should be the total number of "1979 Gamble, Combined" that drives the salary adjustment, rather than individual adjustments for "1979 Gamble, Combined, NY" and "1979 Gamble, Combined, Texas."

Thoughts?

Your reasoning for why they are now different is definitely correct. That's the only explanation for why they would now be different. I noticed this as well when drafting some recent teams, and took advantage of it. However, I'm not sure how much of a difference it really made at this point since the salary differences are generally minimal. Regardless, I agree that they should change how they handle this from June onward.

you want salaries based on popularity you got it