How to estimate your FIDE rating (conversion formula inside)

The formula is only as accurate as your Lichess ratings, specifically both Blitz and Classical ratings, @howchessYT . It's not a stupid formula when you use it on an inaccurate profile. Have you tried it on a profile with dozens of blitz and classical games played, especially recently? It's fairly close at times, and definitely within reason for an estimate.

It looks like there is heated debate here :D

I didn't bother reading the debate, I just tried the formula for fun and I'd say the result is pretty accurate for me.

So kudos for nice formula :D

It isn't bad I suppose

You guys haven't figured it out yet.
The OP's extensive research shows members blitz ratings here are +78 points (median average) above their stated FIDE OTB rating and the classical rating is +169 above the stated FIDE rating. All very reasonable. Generally speaking online ratings are slightly higher than FIDE OTB ratings and this was proven out by the OP's research.

The OP then proposes a formula to predict a FIDE OTB board rating for those who do not have one, based on their online blitz and classical ratings.

The formula predicts a HIGHER FIDE OTB rating than the players online blitz rating. (This applies to all players below 2100. The estimated rating progressively increases as the online rating decreases.)

Sound scientific? Reasonable? Worthy of note ???
Hardly

@mdinnerspace I will be taking a break from lichess for a while, but I wanted to take one last shot at this argument.
To summarize:

You said that including a constant in a function makes it useless for relating variables. That has been shown false in the example of the correlation between Celsius and Fahrenheit. Also, you propose an approximate equation in the last post that includes a constant,
(FIDE=Blitz-78) so I think that this issue is settled.

You also said that the data was unreliable because it came from voluntary statements of FIDE rating. As previously stated, measures have been taken to ensure that the ridiculous results have diminished effect. Also, you seem to recognize in your last post that the strong correlation presented by the graphs also lessens the effect of outliers. So I think that that is settled, too.

Now for your final objection, the one that you have been alluding to in some form for the entire discussion. Your objection is that the equation violates the data (and we agree that the data is correct) in that the equation predicts an OTB rating higher than the online blitz rating for players below 2100, while the data shows that lichess blitz ratings are, on average, around 78 points higher than OTB ratings for all players. This objection is also false. The data agrees with the equation in saying that lower rated players DO have better OTB ratings than online rating. Look at the graph given multiple times by the OP. It does not support your assertions.

I rest my case.

The data is base on these lower rated players claims of their FIDE rating. I agree that the "claims" reported are accurately given by the OP, but not that they are correct.

Where is the verifiable evidence? What is required is an official FIDE list that makes a comparison to known Lichess ratings. Basing the graph on "voluntary" information given on an online profile is hardly scientific.

From my many years of 1st hand experience it is quite uncommon for new OTB players, after 12 games, to achieve a FIDE or USCF rating that was higher than their online blitz rating. Players achieve online skills that are specific to blitz play and in general they are able to increase their rating by experience.
They start a new at OTB play. Only from experience and many games will they bring their OTB rating up to an established rating.

An established rating after many games, may in fact become very close to their online blitz rating. (But in general, up and down the spectrum online ratings are higher). The entire point of the OP's exercise is to predict a FIDE rating for those who do not have one, for those who have not played OTB by making a prediction of an expected rating based on online play.

2nd. A "constant" (187) can not be added to lower resulting numbers of 2 variables and added in the same way to higher resulting numbers of two variables. The result will always be "skewed"in a linear direction both upwards and downwards.

Using the formula a Blitz 1678 and classical 1769 results in a 1674 FIDE. (4 points difference)
a Blitz of 578 and a classical of 679 results in a FIDE 731 (an increase of 153 points) Lower rateds progressively are estimated to have a higher FIDE rating than their blitz rating.

Now observe the estimate for a blitz of 2878 and a classical of 2969 which gives a FIDE rating of 2705 (a full 173 points below their blitz rating). The higher the ratings, the FIDE estimate becomes progressively lower.

The discrepancy lies in the FACT that "187" is being added to a very low number, bumping that number up proportionately larger; while the "187" is being added to a very large number, having a far less proportionate effect with the resulting estimate to be lower.

The OP creates a graph based on the assumption ... lower rated blitz players have higher FIDE ratings and higher rated blitz players have lower FIDE ratings. No verifiable evidence has been presented to substantiate this claim, unless you count "voluntary info given in online profiles".

I suggest that the highly rated blitz players most often are giving accurate FIDE ratings, which in general are slightly lower than their blitz ratings. Think there is a possibility low rated players are slightly "exaggerating" their FIDE rating to be higher than their online blitz rating? Who would dare fib like this ?

In my case, my blitz rating was a full 400 points below my USCF OTB rating. Tho opposite is often found. Sure a "median" is easily found by making averages. But what do these averages tell us? Unless it can be shown the majority fall within a specific range, =/- 100 points by example, formulas that make estimates are of no practicable value.

To address the comparison made to a Celsiius - Farienheit conversion formula:
0 degress C = 32.0 degress F
As C gets colder (numbers going down) F also gives lower numbers showing colder temperatures.
As C get hotter (numbers going up) F also gives higher higher numbers showing higher temperatures.
A consistent, similar progression is seen in either direction, unlike the OP's formula which progressively gives lower numbers on one side of the scale and progressively higher numbers on the other.

The conversion formula has one variable that of C
F = C x 9/5 +32
The constant 32 is derived from the starting point 0 C.
The two formulas are not remotely similar. The formula for FIDE = (.38 x n) + (.48 x n+1) + 187 has 2 variables and 3 constants, hardly a reliable formula.

@mdinnerspace Actually that last formula you put down can be reduced to .8n+188 with one variable and one constant.

yes. simpler the better. It is the OP's formula
n= blitz rating
n+1 = classical rating (which he states is higher)
Possibly an average of the two... then .8n +188 may work better.

It doesn't work better in the observed sample, at least. I don't know why it would work better out of sample.