16 Nov 07
Originally posted by wormwoodYeah, I guess that's what I was trying to get at. As incandenza mentioned, "normal" here is 1200, and on ICC 1400 and on FICS probably 1200 for blitz... but, as we all seem to agree on, a 1200 there is far stronger than here so the ruler is different.
no, it's just the different scaling in different rating systems. 1200s on ICC are far better players than 1200's here. it's just that the 'ruler' (eg. the rating) that the players are measured against is not the same. (not to even mention they measure different things). it's like comparing BCF to FIDE without conversion.
a 1200 on FICS will stay i ...[text shortened]... le the 1200 here is likely to ask: "what's that 'book' you guys keep talking about?"
I suppose what I was getting at was that if we compare the "normal" among the different servers, the ratio of beginners and very casual players (ie no book knowledge, little or no tactical ability, etc) here vs FICS or ICC is much, much higher here then on either of those servers. Which, again, makes sense since RHP is more accessible (no server commands to learn, no firewall/port issues to address, etc).
16 Nov 07
Originally posted by wormwoodYeah, I wasn't disputing what you were saying about the cumulative vs. non-cumulative percentages. Just further pointing out that the %ile for a given rating sometimes seems surprisingly high.
that was just a coincidence because the top 100pt ranges have so few people.
The more I think about this, I think it also has to do with the fact that the other sites have a RD- (ratings deviation)-based Glicko-style setup, and this site doesn't. If I'm understanding it correctly, on here once you leave the provisional period, no new points are ever added to the system as a whole (your opponent always loses the same number you gain). So it is more difficult overall to move away from the starting rating.
16 Nov 07
Originally posted by scandiumyeah, true. and the differences don't even stop there, but there are very different groups even within a singe server. like the difference between regular blitz crowd on ICC and the 5-minute variant crowd on ICC. both groups are on the same server, have the same initial rating and rating formula, but still the difference in strength between equally rated players are huge. the 1200 5-minute player is far stronger than the 1200 regular blitz player.
I suppose what I was getting at was that if we compare the "normal" among the different servers, the ratio of beginners and very casual players (ie no book knowledge, little or no tactical ability, etc) here vs FICS or ICC is much, much higher here then on either of those servers. Which, again, makes sense since RHP is more accessible (no server commands to learn, no firewall/port issues to address, etc).
my theory is that it's because newbies didn't know how to start a 5-minute game by writing 'play 5-minute' on the command line, so the two rating systems drifted apart. and now, when there's a simple button on the interface to start a 5-minute game, the old players have so much 'weight' on the 5-minute pool that the newbies can't really affect the equilibrium. at least not very fast. maybe in time the two groups will drift back closer to each other, or another influence will drive them further apart.
16 Nov 07
Originally posted by incandenzathe provisional period works very similarly to the RD, and you get very close to your 'real rating' during it, so I don't think it makes a lot of difference. there is a difference, by I think it gets masked very soon after you play more games.
The more I think about this, I think it also has to do with the fact that the other sites have a RD- (ratings deviation)-based Glicko-style setup, and this site doesn't. If I'm understanding it correctly, on here once you leave the provisional period, no new points are ever added to the system as a whole (your opponent always loses the same number you gain). So it is more difficult overall to move away from the starting rating.
16 Nov 07
Originally posted by wormwoodHmm, I think you're probably right.
the provisional period works very similarly to the RD, and you get very close to your 'real rating' during it, so I don't think it makes a lot of difference. there is a difference, by I think it gets masked very soon after you play more games.
I guess it does just boil down to RHP being a much more casual crowd overall than ICC/FICS/etc.
16 Nov 07
Originally posted by incandenzayeah, probably. it takes more of an enthusiast (on average) to install a client interface to access ICC or FICS, and even some unix-like twiddling to access all the features, but here you can just log in to a normal website and play. it makes it a lot easier to a casual player to start playing here.
I guess it does just boil down to RHP being a much more casual crowd overall than ICC/FICS/etc.
17 Nov 07
Originally posted by incandenzaI think the you've put your finger on the answer without quite recognizing it for what it is. The solution (the main part) is that ICC grades on a curve whereas RHP doesn't. From the ICC help file:
I wonder why the distribution here is so different what from what I would consider "normal". e.g. on ICC a rating of 1600 standard would put you pretty much dead in the middle of active players, but here a 1600 rating puts you in the top 10%.
"The rating during the provisional period is the average of a set of values, one for each game played. The value for a game against an established player is the opponent's rating plus 400 for a win and minus 400 for a loss. For a game against another provisional player, the value is moved towards the previous average to lessen the impact of the unreliable result. ***Extra points are then added to the rating for the purpose of keeping the average rating of all established active players close to 1600.*** In particular, 1/5th of 1600 minus the current average is added to the rating."
In other words, ICC deliberately tinkers with its rating formula to establish a "norm" of 1600 whereas RHP sets its norm at 1200 (a distribution which is reflected by the fact that the 1200 group here is the mode average, i.e., the most players of any ratings group here are clustered within a hundred points of it either way).
Another point to consider:
"Ratings usually range from 600 to 3000 on ICC, but there is theoretically no limit at either end."
(Also from the help file.) In other words, whether due to computer engine use (explicit or not) or to something else, there are more ICC players in higher level ratings groups than here.
Both of these factors act to skew ICC ratings percentiles relative to RHP.
ICC quotes taken from:
http://www.chessclub.com/help/ratings
18 Nov 07
7 edits
Originally posted by scandiumRegarding FICS I note that as of Nov. 12th, 2007, Scandium's rating at STANDARD (time-control) chess is 1810, with 905 games played and an all-time high of 1828 (Dec. 14th, 2004); whereas here, with CORRESPONDENCE chess (and use of opening databases allowed) his current rating is 1563 -- though he's still provisional and has only completed eight games here. Of those eight, only two players in the 1100-1300 range are shown. Doesn't this sound like a bit of a small sample from which to make generalizations about the strength of 1200-level players at RHP relative to FICS (or for that matter, ICC)?
Yeah, I guess that's what I was trying to get at. As incandenza mentioned, "normal" here is 1200, and on ICC 1[6]00 and on FICS probably 1200 for blitz... but, as we all seem to agree on, a 1200 there is far stronger than here so the ruler is different.
Also, comparing a 1200 rated blitz player to a 1200 rated correspondence chess player is likely to mislead, since (as scandium's own FICS blitz rating of 1343 with 6,675 games played demonstrates) a strong standard player may have a blitz rating nearly 500 points beneath his standard rating. Furthermore, Blitz players are far more likely to be "booked up" since their play depends so much more strongly on memorized lines.
(Gotta love that edit button!)