- Blind mode tutorial
lichess.org
Donate

Rating System Broken.

#51

Losing to a kid who is rated 2094 FIDE and who will be in the top third of grandmasters if he simply maintains his current level of improvement (and whose non-provisional 3 check rating is 1887) is not my fault. I had the black pieces, and started with a -2 disadvantage before a move was played (white has an advantage of 2 points in the starting position of 3 check).

Losing to a kid who is rated 2094 FIDE and who will be in the top third of grandmasters if he simply maintains his current level of improvement (and whose non-provisional 3 check rating is 1887) is not my fault. I had the black pieces, and started with a -2 disadvantage before a move was played (white has an advantage of 2 points in the starting position of 3 check).
#52

@Aceno0b

You don't get it, do you? You expect the rating system not only to calculate your new rating based on your opponents rating and the outcome of the game. You expect it to check your opponents FIDE rating and whether he is a future GM and then determine that you deserve to lose only one rating point.

Welcome to reality.

@Aceno0b You don't get it, do you? You expect the rating system not only to calculate your new rating based on your opponents rating and the outcome of the game. You expect it to check your opponents FIDE rating and whether he is a future GM and then determine that you deserve to lose only one rating point. Welcome to reality.
#53

No, I expect it to reflect his level of skill

No, I expect it to reflect his level of skill
#54

@Katzenschinken

Oh yes, Lichess rating system is plain joke. That is exectly what we pointing here with concrete examples.

Evidences are clear, just to the people who want to read and understand. The fact that Mark Glickman is a matematician, actually hold PhD and is senior lecturer on statistics at Harvard doesn't meant his chess rating system is ideal. One man cannot be ever 100% correct in anything!

The prove for this claim is that have two revisions of his chess rating system and as well removed one procedure step in 2012 ("NOTE: As of February 22, 2012 the Glicko-2 iterative procedure ("Step 5") has been corrected and is now stable. Please see the Glicko-2 rating system document for details.")

Implementation of his system actually vary quite a bit.

@Katzenschinken Oh yes, Lichess rating system is plain joke. That is exectly what we pointing here with concrete examples. Evidences are clear, just to the people who want to read and understand. The fact that Mark Glickman is a matematician, actually hold PhD and is senior lecturer on statistics at Harvard doesn't meant his chess rating system is ideal. One man cannot be ever 100% correct in anything! The prove for this claim is that have two revisions of his chess rating system and as well removed one procedure step in 2012 ("NOTE: As of February 22, 2012 the Glicko-2 iterative procedure ("Step 5") has been corrected and is now stable. Please see the Glicko-2 rating system document for details.") Implementation of his system actually vary quite a bit.
#55

#54 A little information can be a dangerous thing.

Step 5 was theoretically unstable for enormous data sets although I haven't heard of any problems in practice.

In the mathematical subfield of numerical analysis, numerical stability is a generally desirable property of numerical algorithms. The precise definition of stability depends on the context. One is numerical linear algebra and the other is algorithms for solving ordinary and partial differential equations by discrete approximation.

In numerical linear algebra the principal concern is instabilities caused by proximity to singularities of various kinds, such as very small or nearly colliding eigenvalues. On the other hand, in numerical algorithms for differential equations the concern is the growth of round-off errors and/or small fluctuations in initial data which might cause a large deviation of final answer from the exact solution.

Some numerical algorithms may damp out the small fluctuations (errors) in the input data; others might magnify such errors. Calculations that can be proven not to magnify approximation errors are called numerically stable. One of the common tasks of numerical analysis is to try to select algorithms which are robust – that is to say, do not produce a wildly different result for very small change in the input data.

An opposite phenomenon is instability. Typically, an algorithm involves an approximative method, and in some cases one could prove that the algorithm would approach the right solution in some limit (when using actual real numbers, not floating point numbers). Even in this case, there is no guarantee that it would converge to the correct solution, because the floating-point round-off or truncation errors can be magnified, instead of damped, causing the deviation from the exact solution to grow exponentially.
https://en.wikipedia.org/wiki/Numerical_stability
https://i.imgur.com/oZ36B.jpg

#54 A little information can be a dangerous thing. Step 5 was theoretically unstable for enormous data sets although I haven't heard of any problems in practice. In the mathematical subfield of numerical analysis, numerical stability is a generally desirable property of numerical algorithms. The precise definition of stability depends on the context. One is numerical linear algebra and the other is algorithms for solving ordinary and partial differential equations by discrete approximation. In numerical linear algebra the principal concern is instabilities caused by proximity to singularities of various kinds, such as very small or nearly colliding eigenvalues. On the other hand, in numerical algorithms for differential equations the concern is the growth of round-off errors and/or small fluctuations in initial data which might cause a large deviation of final answer from the exact solution. Some numerical algorithms may damp out the small fluctuations (errors) in the input data; others might magnify such errors. Calculations that can be proven not to magnify approximation errors are called numerically stable. One of the common tasks of numerical analysis is to try to select algorithms which are robust – that is to say, do not produce a wildly different result for very small change in the input data. An opposite phenomenon is instability. Typically, an algorithm involves an approximative method, and in some cases one could prove that the algorithm would approach the right solution in some limit (when using actual real numbers, not floating point numbers). Even in this case, there is no guarantee that it would converge to the correct solution, because the floating-point round-off or truncation errors can be magnified, instead of damped, causing the deviation from the exact solution to grow exponentially. https://en.wikipedia.org/wiki/Numerical_stability https://i.imgur.com/oZ36B.jpg
#56

The rating system is incorrect anyway... I should be about 1600 and am rated 2000 here... so let's begin fixing this issue before discussing your concerns?? @Fischer_and_Chips88

The rating system is incorrect anyway... I should be about 1600 and am rated 2000 here... so let's begin fixing this issue before discussing your concerns?? @Fischer_and_Chips88
#58

@HellevatorOperator

Magnus is world No.1 because he live in the chess world only, from day he learned to play it and have developed blindfold skil to play with several (dozen?) opponents.

I doubt that can do any amateur who learned to played chess in later age. All super GMs started to learn chess from age 3 or 4 or similar as progress in these ages is quite enormous, actually programming his brain especially to play chess. I doubt any super GM live out of chess world.

@HellevatorOperator Magnus is world No.1 because he live in the chess world only, from day he learned to play it and have developed blindfold skil to play with several (dozen?) opponents. I doubt that can do any amateur who learned to played chess in later age. All super GMs started to learn chess from age 3 or 4 or similar as progress in these ages is quite enormous, actually programming his brain especially to play chess. I doubt any super GM live out of chess world.
#59

#56 The Elo system (glicko-2 for this instance) determines the relative strength of a player in a certain (typ of) game for a given pool of players. The exact value x of a player A alone does not have any meaning at all. First if you add a second player B with rating y the expectation value for the outcome from A´'s perspective of an hypothetical game between player A and B can be determined using the first formula on page 5 of this document http://www.glicko.net/glicko/glicko.pdf (just substitute my variable names with the names used in the document). [For glicko also consider the so called "ratings deviations".]

#56 The Elo system (glicko-2 for this instance) determines the relative strength of a player in a certain (typ of) game for a given pool of players. The exact value x of a player A alone does not have any meaning at all. First if you add a second player B with rating y the expectation value for the outcome from A´'s perspective of an hypothetical game between player A and B can be determined using the first formula on page 5 of this document http://www.glicko.net/glicko/glicko.pdf (just substitute my variable names with the names used in the document). [For glicko also consider the so called "ratings deviations".]

This topic has been archived and can no longer be replied to.