Dude, you still don't understand what adding 187 is doing, or why I did that.
When someone is trying to explain something and you don't understand, it's OK to ask for clarification.
What's not OK is to plow ahead in ignorance while calling other people's work "useless"
At least, you should have the humility to say that the formula has no use FOR YOU.
@Sarg0n the constant is always the y-intercept for an equation in y= firm. Same thing; he didn't invent it. But using a constant does not nullify the equation whatsoever.
Its pretty clear that the basis for rejecting the formula comes from ignorance and not from a specific, valid concern of the formula itself. All parts of the formula has a basis for bring there, and a constant doesn't invalidate a formula. An equation, such as the conversion of temperatures, shows that to be the case.
The OP's premise is that there is a correlation between online ratings and FIDE OTB ratings. He presents a formula attempting to establish the premise.
.38 times x (a variable) + .48 times y (a variable) plus z which is a constant of 187 = A ...
in an attempt to establish a relationship exists between x and y to that of A.
Adding any constant nullifies the formula as representing an equation. 6th grade mathematics teaches inequalities.
That makes absolutely no sense! Constants nullify nothing. Constants are used all of the time in all sorts of equations, like (see my last post) temperature conversion. Using three variables makes no difference. Constants are completely acceptable.
Also, the formula is not an attempt to establish the premise. The formulas quantifies the premise established by the data in the graph.
@mdinnerspace
In all honesty, what the heck do you have against this formula?
Adding constants in conversions is absolutely acceptable and is seen all over the place. Temperature conversion is an excellent example given by another here. Is that equation wrong?
You may not accept the results of an equation but you don't invalidate the equation because you don't think the results are correct. Sometimes results surprise us.
Let's take the example of a player rated 278 and 369.
.38 x 278 = 105
.48. x 369 = 129
105 + 129 + 187 = a estimated FIDE rating of 421 !!
It truly is amazing most everybody can not see that this constant of 187, that is applied to all ratings, skews the predicted result. Lower ratings, when the numbers are entered into the formula, produce a proportionately higher predicted result as compared to higher ratings. The "formula" is invalid. Nobody is arguing the math. The numbers speak for themselves. As an equation, the formula violates basic mathematical principles of attempting to establish a correlation between A and B.
The formula is derived from the available data. It makes no sense to ask the formula to make predictions outside the range of observed data (statistically speaking: outside the distribution's "support"). Any linear approximation will produce funky results when you plug in extreme values.
This is what the previous poster was getting at with his Fahrenheit example. Obviously, you did not understand why that example was relevant.
If you look at the scatter plot, you'll notice that almost noone below 1200 on lichess reports their FIDE rating.
The crazy examples you give are not useful, because they are extreme. It is entirely expected that the formula will behave weirdly at that spot. There is NOTHING surprising or invalidating there. If you knew anything about statistics, you would have realized that.
im just gonna show you something:
couple of national ratings which attempt to convert fide ratings:
use two formulas... one for higher players and one for lower players. there is a reason for that.
FIDE Rating list of May 1st, 2014 (Players Active since 2010)
For Players CFC 2200+ (193 players): FIDE = CFC - 103.2
For Players CFC U2200 (423 players): FIDE = CFC - 22.3
USCF = 720 + 0.625*FIDE if FIDE < 2000
USCF = -350 + 1.16*FIDE if FIDE >= 2000
This topic has been archived and can no longer be replied to.