Ratings are just ideas I can beat people high and i can also lose to people lower.
Chess rating system
Sort:
KiNgApE wrote:
Ratings are just ideas I can beat people high and i can also lose to people lower.
Then either they get lucky or you get unlucky. You may beat a person 150 points above you/lose to someone 150 points below you, but you wont be able to consistently beat someone 300 points above you.
I believe the queen is worth 9 points, but other than that, viswanathan is correct.
G
Why is this topic-question 22 pages long??!!?!?!?! 
(here come the flamers to whine at me that by making this post, i'm just making it longer, blah blah blah)
o0obruceleeo0o wrote:
So the reason Kasparov is better than me is because his rating is higher than mine?
yes
is there a point at which a really highly rated player can lose points beating a really low rated player?
Devout_Monk wrote:
Why is this topic-question 22 pages long??!!?!?!?!
Geez, Devout_Monk, you're such a hypocrit! First you tell us that the thread is too long, but have obviously decided to ignore the fact that your post will inevitably increase both its length and its longevity!
Secondly, although not related to thread length, ??!!?!?!?! is NOT a recognised punctuation symbol.
I think my work here is done. 
The mistake was transposing the ? and the ! in the second and third positions.
shrike1 wrote:
is there a point at which a really highly rated player can lose points beating a really low rated player?
No. At least not in a system such as that used by the USCF. If graphed out using ratings difference as the horizontal axis and points gained as the vertical, the curve approaches zero asymptotically. At a 600 point difference, the higher rated player would gain only one point with a win. At higher differences, it would be a fraction of a point, possibly not enough to affect the published value.
Is there a point at which a very highly rated player could lose points beating a player with a much lower rating?
No -- the awarding of rating points approaches zero asymptomatically the larger the disparity between the winner and the loser -- it never goes negative.
"...asymptomatically..." ???
Asymptotically (which I misspelled in my earlier and just edited post). From the root, asymptote.
Bur_Oak wrote:
"...asymptomatically..." ???
Asymptotically (which I misspelled in my earlier and just edited post). From the root, asymptote.
Oops -- the combination of speed and spellcheck steered my wrong.
drphon98 wrote:
o0obruceleeo0o wrote:
So the reason Kasparov is better than me is because his rating is higher than mine?
yes
No! Kasparov's rating is evidence he is better than you. It is how we know he is better than you. It is not why he is better than you.
saidh wrote:
A better predictor of a players strength would be the ratio of his rating to the average rating of his opponents. I see a lot of people who build their ratings by piling up wins on beginners so the allmighty rating develop is only a shell of the ability you're "supposed" to have at their height.
This doesn't work if the glicko system is used correctly. The more you do this, the less you earn, until you are earning only fractional points. And if you lose to any of those low rated players, you can lose a significant chunk of points. The risk/rewards eventually balance out and your stuck, unless you start playing tougher opponents.
It only really works if you provide a minimum one point per win, which I think is only done for the live chess here. Hopefully that will go away when they update to the new version of live chess.
But the ratio idea doesn't work. If you have a 1200 with an average opponent of 1200 and a 1600 with an average opponent of 1600, your ratio idea will say they are equal. How does that make any sense?
And yes, you can gain points against people 300-400 points below you. And at that point you have a 1 in 10 chance of losing, and you will eventually lose the points you gained.
The ratio idea also punishes people for playing games against higher rated players, which is the best way to learn, and encourages cherry-picking. Nothing is wrong with the system that is in place now.
What do you even need it for then? Just look at their average opponent score. It's right there in their stats.
ichabod801 wrote:
What do you even need it for then? Just look at their average opponent score. It's right there in their stats.
Agreed. All of the conclusions you describe being able to draw based on the ratio can be drawn from the average opponent rating which is a much more intuitive stat for to new users to wrap their heads around.
Ratings are fundamentally a statistic, meaning a measure of one parameter of a population, thus saying very little about the precise outcome of a single encounter other than the probabilities associated with that encounter. A 400 point rating advantage yields roughly a 1-in-10 probability of the lower player winning a single encounter (by Elo's reckoning). If I take 100,000 such encounters amongst thousands of players who have each played numerous games against numerous opponents, that roughly 1-in-10 probability will show up (at least within some prescribed interval).
As we start to move away from that usage (a population statistic with probability inferences), we begin to dilute the meaning and value of ratings. But, hey ... I still wish I were rated 2800 (but only if I could back it up).