BuzzerBeater Forums

BB Global (English) > The Quest for the Formulas

The Quest for the Formulas (thread closed)

Set priority
Show messages by
From: Asasasa

This Post:
00
46657.69 in reply to 46657.62
Date: 11/29/2008 12:23:44 AM
Overall Posts Rated:
22
They do have.

A couple of seasons ago, when salaries updated during the ASG, one of my players had a $30 increase, because he had 1 stamina training during those 7 weeks.

Its almost minimal but the effect is there.

This Post:
00
46657.70 in reply to 46657.68
Date: 11/29/2008 12:46:49 AM
Overall Posts Rated:
458458
It very well could. Unfortunately, I don't keep the data for experience.

Once I scored a basket that still makes me laugh.
From: Monique

This Post:
00
46657.71 in reply to 46657.60
Date: 11/29/2008 3:31:35 AM
Overall Posts Rated:
2323
I only have a SG that is about 10% off, but I guess that's not that exeptional :)

This Post:
00
46657.72 in reply to 46657.62
Date: 11/29/2008 5:44:49 AM
Le Cotiche
III.1
Overall Posts Rated:
772772
your formulas predict the correct role for all my players
but for some of them the salary is quite different (in %)
especially for this guy (5789231) because his predicted salary is actually higher than the real one, so it's not about hidden decimals
his skills didn't change since the salary update

i hope it can be helpful

PS i know i couldn't link him, but i put him in the TL (at a very HIGH price for his skills so nobody's gonna buy him) only to post him there

This Post:
00
46657.75 in reply to 46657.74
Date: 11/29/2008 11:24:19 AM
Overall Posts Rated:
224224
I assume these come from simple linear regressions, right? What percentage of the variance in salary do the models explain? (or, in other words, what's the R-squared of the models above).

Also, have you considered the possibility that the models are non-linear (that is, salary is calculated by multiplying the variables instead of adding them)? In such case, the right way to do the regression would be to use the natural logarithms of skills, rather than their actual values.

Last edited by GM-kozlodoev at 11/29/2008 11:25:41 AM

"I don't know half of you half as well as I should like; and I like less than half of you half as well as you deserve."
This Post:
00
46657.77 in reply to 46657.76
Date: 11/29/2008 11:51:42 AM
Overall Posts Rated:
224224
Sorry, it seems you took time to explain earlier, but I was too lazy to look back. Will fix that, and thanks for taking the time to point to the correct posts.

"I don't know half of you half as well as I should like; and I like less than half of you half as well as you deserve."
This Post:
00
46657.78 in reply to 46657.73
Date: 11/29/2008 2:01:45 PM
Le Cotiche
III.1
Overall Posts Rated:
772772
his real salary is 6089 and the predicted salary is 6409, this is only 5.2% off. It doesn't matter whether it's higher or lower, sublevels can explain both directions, because the model will unavoidably adapt its coefficients to the average of the sublevels it encounters in the training data, and this average is likely somewhere close to the middle of the range.


ooops sorry i'm an idiot, i typed the wrong actual salary in the excel file i'm using... forget about it

Advertisement