Congratulations to Lisle, he now has a four digit number after his name. Any one else achieve their highest rating so far?

Congratulations to Lisle, he now has a four digit number after his name. Any one else achieve their highest rating so far?

Thanks Justin. However, this has made me a little skeptical about the process for how players are rated. Based on what I see, the round ratings are calculated using a recursive function. The round ssa is probably calculated using a linear regression where the player ratings are the inputs, and the scores are the outputs. Once the best-fit line is found, they predict what score a 1000-rated player should get. Since scores and ratings are not normally distributed, a regression should not be run.

This will inflate your rating if you play against higher rated players and decrease your rating if you play against lower-rated players, even if you shoot the same score under the same conditions. If Pro-only or AM-only tournaments are played, Am's ratings will spiral down over time and Pro ratings will spiral up.

Anyhow, I think I'm really a 970 player - but thanks again

So if I play Open from now on, finish DFL every time, I'll be rewarded by my rating increasing?
Sweet!:D

Btw, my rating is at its highest point ever, right now: a whopping 896!
My goal for last year was to get over 900 and I was well on the way until the knee injury derailment.
So, I'm rolling over last year's goal to this year.

Thanks Justin. However, this has made me a little skeptical about the process for how players are rated. Based on what I see, the round ratings are calculated using a recursive function. The round ssa is probably calculated using a linear regression where the player ratings are the inputs, and the scores are the outputs. Once the best-fit line is found, they predict what score a 1000-rated player should get. Since scores and ratings are not normally distributed, a regression should not be run.

This will inflate your rating if you play against higher rated players and decrease your rating if you play against lower-rated players, even if you shoot the same score under the same conditions. If Pro-only or AM-only tournaments are played, Am's ratings will spiral down over time and Pro ratings will spiral up.

Anyhow, I think I'm really a 970 player - but thanks again

In a regression, only the Y (outcome) variable needs to meet the assumption of normal distribution. It is okay that the X (input) variable is not normally distributed. Presumably, they perform a mean-centered correction on Y which fixes the normality issue, thus allowing reasonable conclusions to be drawn from those numbers. However, if they are not performing some sort of correction to meet the assumption of normality, then the outcomes are null because they are based on data from a non-normal distribution. Now that I think about it, they might have a problem with heteroscedasticity...I need to see the formula, now I have my doubts.

Thanks Justin. However, this has made me a little skeptical about the process for how players are rated. Based on what I see, the round ratings are calculated using a recursive function. The round ssa is probably calculated using a linear regression where the player ratings are the inputs, and the scores are the outputs. Once the best-fit line is found, they predict what score a 1000-rated player should get. Since scores and ratings are not normally distributed, a regression should not be run.

This will inflate your rating if you play against higher rated players and decrease your rating if you play against lower-rated players, even if you shoot the same score under the same conditions. If Pro-only or AM-only tournaments are played, Am's ratings will spiral down over time and Pro ratings will spiral up.

Anyhow, I think I'm really a 970 player - but thanks again

In a regression, only the Y (outcome) variable needs to meet the assumption of normal distribution. It is okay that the X (input) variable is not normally distributed. Presumably, they perform a mean-centered correction on Y which fixes the normality issue, thus allowing reasonable conclusions to be drawn from those numbers. However, if they are not performing some sort of correction to meet the assumption of normality, then the outcomes are null because they are based on data from a non-normal distribution. Now that I think about it, they might have a problem with heteroscedasticity...I need to see the formula, now I have my doubts.

Do you guys know Chuck Kennedy?

http://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics

Umm.... What?:confused:

yeah, you guys need to get a private thread

i know you just take the inverse reciprocal of the starting value of your ending score... right???

i know you just take the inverse reciprocal of the starting value of your ending score... right???
No, no, no... you take the derivative of the integral of the weighted mean as it approaches 0 when x = 2.

No, no, no... you take the derivative of the integral of the weighted mean as it approaches 0 when x = 2.

Okay, now you have me confused...

I thought it was the negative derivative under the radical making it an imaginary number of the weighted number as it approaches 0 when y = x negative squared. :p

it's still 1+1=2, right?

it's still 1+1=2, right?

i used to think so... now i'm a little confused

Do you guys know Chuck Kennedy?

http://en.wikipedia.org/wiki/Lies,_damned_lies,_and_statistics

Nice Scott...I did some searching and it looks like dozens of players have already confronted Chuck about how playing with pros pulls your rating up and playing with ams pulls ratings down.

Since neither the player ratings nor the round scores are normally distributed, he simply needs to calculate the ssa by making the median round score equal to the median player rating and extrapolate to 1000 from there. I'll have to shoot him a note ;)