How to estimate your FIDE rating (conversion formula inside)

Except Sarg0n...
The formula predicts a HIGHER Fide rating than the online rating. This is the issue. Online ratings "fit in a cloud" about 100 points above a Fide rating. All fine and good, as the data given reveals +78 and +169 points for blitz and classical respectively.

Insert those numbers into the Formula!
Say 1078 and 1169 and you'd expect a prediction of 1000 right?

(187) + .48 x 1078 (517) + .38 x 1169 (444) = 1148

A predicted Fide rating that is 70 points HIGHER than the online blitz rating. (Instead of 78 points lower, missing the prediction by 148 points). The OP made the claim his formula is accurate within a few points.

Case closed. The formula is not consistent with the original premise by which his data gives as:

* A typical (median) user's FIDE rating tends to be 78 points lower than her Lichess Blitz rating
* A typical (median) user's FIDE rating tends to be 169 points lower than her Lichess Classical rating
OP - taken from the 1st post.

I know how to find out Your Fide rating. You have to play online, train train etc. and then go OTB to play Fide tournaments, until Your K will be 10, then play some more tournaments with K = 10, and then You will know Your Fide rating pretty precisely. Simple :)

Those 10% below 1200 also don‘t wag the big dog.

"Case closed. The formula is not consistent with the original premise by which his data gives as:

* A typical (median) user's FIDE rating tends to be 78 points lower than her Lichess Blitz rating
* A typical (median) user's FIDE rating tends to be 169 points lower than her Lichess Classical rating" @mdinnerspace

Do you know what a median is

10% ? Review lichess stats.
The formula predicts a HIGHER Fide rating than the expected result for ALL ratings up to 2200, where the trend takes the opposite course and begins to predict a lower rating.

Again, the man demonstrates that he doesn't understand the difference between the median (Fide - Blitz) gap, and the expected value (i.e., predicted outcome) in a linear model with two regressors.

It is ridiculous to believe that the two should coincide perfectly. That *argument* (*not* the person) is simply ignorant.

@mdinnerspace , FIDE ratings have a floor of 1000 while Lichess has a rating floor of 800. Thus the lowest rated players usually have a FIDE rating higher than their Lichess rating.

Besides, FIDE uses ELO rating calculation formula, which produces a more "compact" range than the Glicko system used in Lichess, which results in the opposite effect (Lichess rating being higher) over the other end of the scale.

en.wikipedia.org/wiki/Elo_rating_system
en.wikipedia.org/wiki/Glicko_rating_system

The formula predicts that within the 1400 range is a mid-term were they tend to get balanced: 187 + (.48 x 1400) + (.38 x 1400) = 1391
And for higher ratings it will predict lower FIDE ratings.

If you can follow my advice and explore the data yourself, I'm sure it will be fun.

@EvilChess

You are making a basic mistake in your example, one that is easily done.
You are using .38 x 1400 and .48 x 1400.
This is Incorrect. There is a difference of 91 points between blitz and classical ratings.
If we are to accept the "median" as being accurate, which we must do according to researched data of ratings here at lichess; the formula MUST predict 1478 x .48 and 1569 x .38 +187 = 1400 which it does not.

as the median is:

* A typical (median) user's FIDE rating tends to be 78 points lower than her Lichess Blitz rating
* A typical (median) user's FIDE rating tends to be 169 points lower than her Lichess Classical rating

Quite possibly this same simple error was made by the OP in developing his formula? Using the same number for both blitz and classical?

@mdinnerspace You write:

"If we are to accept the "median" as being accurate, which we must do according to researched data of ratings here at lichess."

We must do no such thing.

There is absolutely nothing that requires that this gap be constant across the sample, or that the formula conform to that gap throughout the sample. In fact, @EvilChess just gave you two very specific reason that explain why that gap might not be constant over all the sample.

THE MEDIAN GAP IS NOT A GOOD BENCHMARK. The fact that the formula does not make the same prediction as the median gap is not a problem, because the gap is a simplistic measure of the difference between Lichess and Fide.

Different predictions using the formula and the median gap DO NOT FALSIFY THE USEFULNESS OF THE FORMULA.

We have explained this many times to you, but you still fail to understand.

oh..Ok. I get it now. The median is not consistent. It can wildly fluctuate between ratings. For 1200 rated players it has one value and for 1400 a completely different one. Then for 1600 a 3rd and 1800 it's something completely else.
I was just wondering upon which data the formula was derived from? But now I understand. Since the median could be anything for any rating, the formula can also make any prediction, higher or lower and sometimes equal to online ratings while remaining an accurate conversion.
Anyway, so long as the formula makes an accurate prediction within a few points (as stated by the OP) of a 1st time players OTB Fide rating based on their online blitz rating we are all happy!