Go back
RHP ratings...

RHP ratings...

Only Chess

K
Key West Wannabe

North Carolina

Joined
14 Sep 04
Moves
12874
Clock
28 Nov 04
Vote Up
Vote Down

I have a question for you long-time subscribers about RHP ratings. When I first got here, I played a few games during my provisional period against the p1200 guys and I won a bunch of games. My rating shot up to p1800 something. Then I got several games from 1800 players and I got skunked quite a few times. So I dropped back down into the 1300's. After getting wins against them I am back into the 1600's and gaining...but I am STILL finishing that bunch of games with 1300 players and in some of the 1500 games I am playing, I am getting a a mixed bag. The ratings seem to lag behind your real rating by about a month or so, since that is how long it takes to get a lot of games here finished.

After that rant, my question is when does your rating stop fluctuating and you finally end up where you rightfully belong?

R

Finland

Joined
30 Dec 02
Moves
5164
Clock
28 Nov 04
Vote Up
Vote Down

(I'm not a long time subscribe and I don't have experience about my own rating stablizing, so you didn't ask me but)

First, the mathematical model behind the rating calculation probably is not completely accurate, so your "true rating" (i.e. the limit average rating if you continue playing on your current level) might not be unique but depend on the rating distribution of your opponents.

Second, your playing level probably is not constant but fluctuates at least a little over time (even if you didn't improve), for instance because sometimes you have more time to put on games (many other reasons too, I think)

Third, even if the rating model was accurate and your playing level constant, there would be some random fluctuation, natural fluctuation because results of games are random and even greater fluctuation because even if your rating would be close to the "real value" the opponents' ratings may have large measurement errors (e.g. because they have played only small amount of games, because they are improving fast, because they have just lost hundreds of rating points due to time-outs on an inactive period)-

So, the rating point where a player "rightfully belongs" probably doesn't even exist, and even if it did, a player's rating would still fluctuate, possibly a lot.

I hope someone understand at least something about what I tried to say. If anyone is interested (probably not), I might like to discuss the subject more. (And no, I'm no real expert on this subject, I may have got something wrong)

SN
Pimp of the elves

San Diego, CA

Joined
17 Mar 04
Moves
65429
Clock
29 Nov 04
Vote Up
Vote Down

Yea! What he said!

Actually, very well put! Those where my thoughts on the subject exactly.

TSD
The 3rd Coming

London

Joined
23 Aug 04
Moves
25775
Clock
30 Nov 04
Vote Up
Vote Down

Originally posted by Rabella
[bI hope someone understand at least something about what I tried to say. If anyone is interested (probably not), I might like to discuss the subject more. (And no, I'm no real expert on this subject, I may have got something wrong) [/b]
Have written a few simulations in the past on similar dynamic systems, thought it might be fun to do one on the RHP rating system.

Then use it to test questions like. If a player's true rating is 1500 and starts at 1400 what is the average number of games required to be within 10 points of the true rating etc. Or just demonstrate the random fluctuations described above.

Unfortuately don't have the time but think it would illustrate well that the further a player is from their true rating quicker and bigger the corrects would be - hence imbalances due to a strong player being temporarily low and winning against you doesn't really matter in the medium run etc.

I also hope someone understand at least some of this.

e

Joined
24 Aug 04
Moves
3183
Clock
30 Nov 04
3 edits
Vote Up
Vote Down

Originally posted by The Swine Down Hope
Unfortuately don't have the time but think it would illustrate well that the further a player is from their true rating quicker and bigger the corrects would be - hence imbalances due to a strong player being temporarily low an ...[text shortened]... winning against you doesn't really matter in the medium run etc.
Doesn't this assume the player challenges (or is challenged) by those who are at his or her true rating level. Imagine an 1800 player, who is currently p1200 and only plays against those rated at 1100. It will take many games for that person to reach 1800, although they probably will.

This is an extreme example, a real example, probably the general case is, when you're 1300 you play other 1300 players. When you are 1400 you play other 1400 players, leading to the constant rating slopes, not exponential curves, you see in most player profiles.

Which brings me to my next point, and I hope answers the poster's question.

Your rating reflects your relative rating in a given player pool, for instance RHP. It cannot be directly related to a FIDE or Yahoo! rating. It is also a type of moving average. Therefore, like any average, as you finish games (i.e. collect more data points), the average approaches a better approximation of what is trying to be measured.
  • In most places the rating is considered "true" after 20 games. This seems pretty reasonable to me, although KWCorona's case seems to be an exception.

    Also, because it is an average, which includes past data, the number reported will usually not reflect the current true value. (This is a hand-waving the true problem.)

    Actually the whole rating system is very logical and a lot of thought
    went into it. It's based on calculating the likelyhood of a player winning
    the game in question.

    This paper covers the whole topic pretty well.
    http://math.bu.edu/people/mg/papers/chance.ps

    * Think of a coin toss. If one tosses a coin just once, one might
    conclude there is a 100% chance of heads. This is obviously
    totally wrong and way too high. The wrong conclusion was reached
    because there were not enough coin tosses (i.e. data points). If
    one increases the tosses to 4 then the answer reported could be
    75%, also wrong. If one throws the coin even more, say 1000 times,
    then one will then have a better idea of what the odds of heads
    really are.

  • Cookies help us deliver our Services. By using our Services or clicking I agree, you agree to our use of cookies. Learn More.