Bad calculations?

[JohnQ]

I have a question about calculations in Wins Produced

http://wagesofwins.com/how-to-calculate-wins-produced/

For example, Derrick Rose committed 8.2% of Chicago’s personal fouls in 2010-11. Chicago’s opponents made 1,450 free throws, so Rose is charged with 119.5FTM(opp.).

8,2%?

Derrick Rose had 136 fouls
Chicago had 1639 fouls
so... 136/1639 is 8,297% ~ 8,3%

next...

Rose PROD= 128*0.064 + 583*0.032 + 476*0.017 + 886*-0.034 + 79*-0.015 + 81*0.034 + 249*0.034 + 278*-0.034 + 85*0.033 + 119.5*-0.017 + 51*0.020 = 7.47

equation '128*0.064 + 583*0.032 + 476*0.017 + 886*-0.034 + 79*-0.015 + 81*0.034 + 249*0.034 + 278*-0.034 + 85*0.033 + 119.5*-0.017 + 51*0.020' give us 7,1925 ~ 7,19

where is the error?
who did it wrong?

[xkonk]

My guess would be from the part directly between the general equation and Rose's specific numbers:

NOTE: One should note that the above values have been rounded off to three decimals. The results reported below for each player were ascertained without any rounding.

[jbrocato23]

Your numbers are right, and it's also true that the calculations on that page are done with non-rounded numbers. Be careful though, there are quite a few errors on that page...

[JohnQ]

Ok, but if we have:

Derrick Rose had 136 fouls
Chicago had 1639 fouls
so... 136/1639 is 8,297% ~ 8,3%

we cant use 8.3 instead of 8.2 in next steps, because this is first step and later we will have different result and that difference will be big

I looked at the further calculations and...

Bulls Team Defensive Adjustment = [(427*-0.064 + 2378*-0.032 + 567*0.033 + 57*-0.034 + 415.0*0.033 – 468*0.200)/19,830]*48 = -0.195

-0,195?!

This is no longer a matter of the adopted approximations. Good result is -0,40316 !

So if every result on that site is bad, it is so strage, that Wins Produced are so popular...

[Mike G]

Popular?

[JohnQ]

I thought so..

[v-zero]

It is very popular amongst what I'd describe as the pseudo-science crowd. Guys who have degrees/high school level scientific understanding but whose understanding of regression analysis is extremely meagre. The biggest single issue with Wins Produced (other than the constant use of straw men/appeals to scientific authority in answering criticism) is that it takes a team-level regression and assumes it will correctly apportion credit/blame at the player level. Good examples of this on defence are the fact that a DRB is worth half a defensive stop (madness!) and that a foul should be penalized with the full value of foul shots made as a result (very dumb).

[talkingpractice]

I have a question about calculations in Wins Produced

http://wagesofwins.com/how-to-calculate-wins-produced/

where is the error?

The day some book editor said "sure, we'll publish that book".

[DSMok1]

There are other threads where Wins Produced is discussed... I don't think we need another one. Could we please stay on topic here? Thanks!

[v-zero]

I think it's good to point out that Wins Produced is a bad metric beyond the inaccuracy of that particular page. WoW is a dangerous honey trap for people interested in getting into basketball stats who might be fooled by its following/the arrogance of Berri.

This post is not intended to patronize, unfortunately the letters PHD are far too convincing in a 'he has a PHD so he must be right' kind of way.