RHP Ratings versus OTB....I GOT An Answer

People are always asking how they compare and the standard answer is they don't, buuuuuuuuut, I believe I have a simple solution.
USCF ratings show the percentile breakdown of the different rating ranges, for instance, 2200+ is the top 1% of all players, 1800+ the top 8%, and a rating of 1000 is right about at the 50 percentile rank.
Soooooooo, kiddies, check your fide/bcf/whatever rating and see what percentile rank that puts you in amongst all the players. You should be at about the same on RHP. That may give you an idea of the rating differance in point value.

Originally posted by General Putzer
People are always asking how they compare and the standard answer is they don't, buuuuuuuuut, I believe I have a simple solution.
USCF ratings show the percentile breakdown of the different rating ranges, for instance, 2200+ is the top 1% of all players, 1800+ the top 8%, and a rating of 1000 is right about at the 50 percentile rank.
Sooo ...[text shortened]... be at about the same on RHP. That may give you an idea of the rating differance in point value.

USCF 1000 may be the 50th percentile of all chess players (I have no idea how you'd even attempt to find this out without bias) but clearly the membership of RHP is a biased sample by it's very nature (being that extremely casual player [know the rules and very little else] are less likely to find and sign up to this site).

You seem to have missed my point, Xanthos. If a person plays in the top 10 percentile in say, USCF tournaments, then he should be in ABOUT the top 10% of the players on RHP. A uscf rating of 1700 is around the top 10% rank, and a quick glance at the player tables show the top 10% of RHP players is around the 1550 mark.
Get it?

Originally posted by General Putzer
You seem to have missed my point, Xanthos. If a person plays in the top 10 percentile in say, USCF tournaments, then he should be in ABOUT the top 10% of the players on RHP. A uscf rating of 1700 is around the top 10% rank, and a quick glance at the player tables show the top 10% of RHP players is around the 1550 mark.
Get it?

You're assuming that the population of USCF tournament players and active RHP members are equally distributed with no bias between them.

EDIT: Also in your original post you said "all players" whereas your second post you said "tournament players". These are very different populations, a quick check reveals you mean the second (that is the definition which gives 90th percentile around USCF 1700).

All players in USCF are by definition tournament players, that's how you get a rating. I'm suprised you thought it worth bringing up.

Anyway, as far as a bias in playing strength between the two groups, that certainly is a consideration, but we're talking about a very large population sample here, almost 11000 players now on RHP. If the sample size were very small, I'd be concerned, but eleven thousand?
I think that's large enough to make any difference in overall strength between the two groups unlikely, and if there is, it's so small as to be statistically irrelevant.

Originally posted by General Putzer
Anyway, as far as a bias in playing strength between the two groups, that certainly is a consideration, but we're talking about a very large population sample here, almost 11000 players now on RHP. If the sample size were very small, I'd be concerned, but eleven thousand?
I think that's large enough to make any difference in overall strength betw ...[text shortened]... the two groups unlikely, and if there is, it's so small as to be statistically irrelevant.

It may be enough to get rid of random sampling errors. But it doesn't help with systematic sampling bias. Neither group is randomly selected. I don't think there's any reason to believe the two distributions should be the same.

Originally posted by General Putzer
All players in USCF are by definition tournament players, that's how you get a rating. I'm suprised you thought it worth bringing up.

Anyway, as far as a bias in playing strength between the two groups, that certainly is a consideration, but we're talking about a very large population sample here, almost 11000 players now on RHP. If the sample s ...[text shortened]... he two groups unlikely, and if there is, it's so small as to be statistically irrelevant.

Larger populations do not reduce possible bias differences. What larger populations will do is give a better normal distribution of ability (not necessarily rating as the formula may change the overall shape) this will reduce random bias values. However if an overall bias exists one populations ability normal curve may be shifted left or right by a certain amount.

In fact the larger your two populations the better you can determine whether there is a actual bias difference or not.

Example: Imagine both the USCF and RHP populations play 100 games against each other. RHP scores say 55 out of 100 in these games. If both populations have normal distributions of ability can we say with 90% confidence that the average ability of RHP is higher than that of USCF?

Now increase the number of games to 1000 (with RHP scoring 550).

Originally posted by mtthw
It may be enough to get rid of random sampling errors. But it doesn't help with systematic sampling bias. Neither group is randomly selected. I don't think there's any reason to believe the two distributions should be the same.

I agree. probably the USCF set is pretty much cut around 1200?, weaker players not taking much part in tournaments. in comparison RHP has a LOT of casual players. what was the rhp-median, 1280? -take those people and compare their games against USCF median-player's games. I have no knowledge of USCF-median, but I'd put my money on 1500-1600.

OK good points Xanthos, so lets take the supposition that there is a significant difference in playing strength between FIDE/USCF/BCF whatever, and RHP players.
If thats the case, then a player who is,say 1800 USCF (top 8th percentile) should score lower than that after many games, and only be (for example) among the top 15% of RHP players. (Or maybe he'd score higher if USCF has a stronger average strength)
Is that what we in fact are seeing? Any OTB players out there with examples of their relative playing strength (percentile ranking)versus what they have on RHP?
I don't have enough games here to make the judgement, but my initial experience seems to be that my relative USCF percentile placement is close to what I'm expecting on RHP>

Originally posted by wormwood
I agree. probably the USCF set is pretty much cut around 1200?, weaker players not taking much part in tournaments. in comparison RHP has a LOT of casual players. what was the rhp-median, 1280? -take those people and compare their games against USCF median-player's games. I have no knowledge of USCF-median, but I'd put my money on 1500-1600.

Uhhhhhh....wormwood........we're talking relative placement percentages, rating points are irrelevant, and in fact that's what we're trying to get away from using as a basis of comparison. Please pay attention..

There are many more objections that can be raised, but for the entertainment value, I submit:

My current RHP rating puts me at the 96th percentile
My current USCF rating puts me at the 83rd percentile (70th if you exclude scholastice players).

Clearly, I find stronger competition OTB than I do at RHP. On the other hand, the approximately 400 RHP members than are rated above me is a larger number than the number of USCF members in my state that are rated higher.

Originally posted by General Putzer
Uhhhhhh....wormwood........we're talking relative placement percentages, rating points are irrelevant, and in fact that's what we're trying to get away from using as a basis of comparison. Please pay attention..

my point was the two distributions you are comparing, are fundamentally different, the same thing I believe xanthos has been trying to explain. and because of that difference, you cannot compare them like you do. the other one is lacking parts that the other one has.


"why did gaussian distribution cross the road?"
"-I don't know. why?"
"because the rayleigh distribution was on the other side."

Originally posted by wormwood
I agree. probably the USCF set is pretty much cut around 1200?, weaker players not taking much part in tournaments. in comparison RHP has a LOT of casual players.

You got that right. Lots of casual players throwing percentage comparisons way off. It's a lot easier playing at rhp than paying tons of money to register and play in uscf tournaments.

Realizing it’s a subjective evaluation, how do RHP players compare to OTB players of similar ratings? If you play a 1500 OTB, how do you think they compare to a 1500 here?

Originally posted by masscat
Realizing it’s a subjective evaluation, how do RHP players compare to OTB players of similar ratings? If you play a 1500 OTB, how do you think they compare to a 1500 here?

Based on my limited observations, I would say that most USCF 1300 players are stronger than most RHP 1300 players, that most USCF 1700 players are comparable to RHP 1700 players, and that most USCF 2100 players would be accused of cheating if they played on RHP.