"grown ups" discussing a topic.
Nice retort, but kinda childish, would not you say?
Plug in the numbers, easy enough to do, and observe the "trend" the formula predicts.
As online ratings progress lower (below 1800), the predicted FIDE rating increases incrementaly higher.
Online ratings above 1800, the predicted OTB FIDE rating is lower than the online rating. The same trend is observed. The higher the online rating, the predicted OTB FIDE rating is lower by the same incremental amount.
The "formula" is unsound.
The OP established his formula based on statistical evidence. Specific data that he believes represents the whole. It is very easy to create such a formula, it can not be refuted based on the data presented.
It is a simple fact. Players with an online rating below 1800 DO NOT necessarily have a higher FIDE rating as the formula predicts.
Players with an online rating above 1800 DO NOT necessarily have a lower FIDE rating as the formula predicts.
I, for one, am glad to learn that the method of Ordinary Least Squares regression has conclusively been proven to be unsound. Stunning conclusion!
@mdinnerspace , I think you should try to publish a compendium of your thoughts in a peer-reviewed mathematical statistics journal. Your findings may reverse a century of prior empirical results. This is Nobel prize-level stuff, really.
Just found this thread again. 253 posts of an argument about the validity of least squares regression involving someone who doesn't know math?
I love it.
The "formula" must be consistent for all rating groups. The OP said it was not meant for the 700 rated player. Then who was it meant for? If it can not be applied equally for all ratings, it is useless.
"The "formula" must be consistent for all rating groups."
No. It is entirely possible for Lichess ratings to be more/less inflated for high/low-rated players. The consistency requirement you impose only exists in your head.
"The OP said it was not meant for the 700 rated player. Then who was it meant for? If it can not be applied equally for all ratings, it is useless."
If molasses can't be used to power my car, does it mean that "it is useless"? A tool does not need to solve all problems.
Edit: Sorry @mdinnerspace I forgot to tag you in this post that was meant for you.
@BigGreenShrek
Amazing, right?!?!
But I'm in too deep now, I must keep going...
The number +187 in the formula was "chosen" because it fits the median rating of 1800. The formula works. (If the median was found to be lower, a lower constant number would be used, likewise, if the median rating was higher, a higher constant number would be inserted.) The problem lies in as rating numbers decrease or increase from the median, the constant number losses it's validity.
@mdinnerspace You write: "The number +187 in the formula was "chosen" because it fits the median rating of 1800."
This is absolutely NOT how the 187 number was chosen.
The constant in the linear regression model has nothing to do with the median.
"I observe no such evidence that players with an online rating below 1800 will have a higher OTB FIDE rating and players above 1800 will have a lower OTB Fide rating."
Me either. The equation doesn't say that, either. If it did, then that would mean that we have a negative slope somewhere. We do not.
This topic has been archived and can no longer be replied to.