Initial rating 1200 too high

New users are given a rating of 1200 for Rapid when they join.

But this is well above the average of about 1050 so 90% of new users (who will also be new to chess are likely to drop down below 1000 after a few games.

The users profile then lists highest ever rating for that user as 1200 which is well above their actual skill level and well above their highest score achieved through actual playing and not indicative of their skill level.

Now I cannot get an idea of even my own real highest ever rating as the graph only shows last 30 days and my 'highest' score is listed 1200 (my initial joining score), which is well above my actual skill level.

Suggest new users should join on 800 (or some other value round about the 30th percentile.

I believe when you create an account you can choose a starting rating. 1200 is default.

^Fair enough.

I did not know it was choose-able

Maybe suggest the default should be a bit lower.

^I think the graph just shows your rating trend over the last 30 days.

Would be useful if you could see the graph for 1 year or 5 years etc.

Heinkel111, this is not correct. The average rating is 400 ELO higher not lower. I have checked this.

 ^Not sure what you mean DRA.

I am talking about the 'highest' rating not being meaningful since it often refers to the initial rating of 1200 since players will normally drop well below this after their first few games.

Yes, many drop below this because the average initial rating is higher than 1200: if everybody has not enough ELO points at the beginning, then your ELO will drop below this initial points due to a competition for not enough points in the system, due to the conservation of ELO points in the entire system. If you compare the ELO distribution with normal players than you will find that Chess.com gives you 400 ELO points less than a normal FIDE ELO would be. This can be shown scientifically.

1200 is provisional and will immediately fall drastically when the user starts playing some games.  As long as it doesn't cause confusion (if I see an exact 1200 ELO I would assume it's a newbie) I don't think it matters what the default value is

Edit: you also need some (arbitrary) provisional ELO before playing your first game so that your revised rating after the game can be calculated.  The more games you play, the more accurate this number becomes.  Even to start that first game you need to be paired with someone, and perhaps the fastest way to reach an accurate value is to start at 1200.  If you started at 100 and were playing <1000 rated newbies for the first few hours you might find the play too easy and the climb too slow (similarly if you were started off against a ~2000).

^^thanks sm

^^^thanks DRA. OK I now understand that conservation of points is an issue. So when a new person joins they should be given the median (or mean?) rating for Daily and Rapid and they can then sink or swim and system points will be conserved.

That seems to be different to what wanmekowan ^^^^^^^^ was saying about choosing your initial rating.

Also it does not really explain why I started on 1200. 1200 was not the median or mean, it was well above both. Also it is a suspiciously round number for a median or mean.

Does every new joiner get 1200 for rapid?

Why does it not vary to match the median/mean rating at the time of joining to support conservation of points?

If the join-up rating is too high that will place inflationary pressure on everyone else's rating over time as there is basically a bunch of free points getting dumped in the system every time somebody new joins.

No it is a BIG BIG PROBLEM in the ELO system! ELO must be comparable and on Chess.Com it is 400-600 ELO points too low. Millions of chess players will totally misinterpret their ELO rating. 

The ELO falls due to artificial ELO scarcity caused by the 1200 starting value and a 1600 average skill level o the chess hobby and professional community. Must be fixed!

 ^thanks DRA.

I did not realise that chess.com and ELO ratings were significantly out of sync.

Would be interesting to have a conversion heuristic so you could get an estimated ELO rating from a given chess.com rating for any given category of play.

Probably a subject for another thread.

I could be wrong, but I think the new default is 800. The options just provide skill levels, beginner through expert, starting with 400 and going up to 1800.

On v2 of the site, the default was 1200.

No, this would extremely worsen everything even more. Do you actually understand what I am writing?

^I think there are 2 separate problems here DRA

1. The join rating for each category should maybe be the median/mean rating for each category  at the time of joining -  in order to support total system points conservation (this problem is the subject of this thread).

2. The chess.com rating is out of sync with ELO ratings (maybe a subject for another thread)

The site uses Glicko and the algorithm has a built in mechanism to quickly converge on a more accurate rating for a pool of players and time control.

Ratings across different rating pools, may or may not be similar. The ratings between chess.com, FIDE, the USCF, etc, do not need to match and in fact, with different pools of players, and overall rating algorithms,  they are unlikely to. There may be some ability to correlate between pools and systems though.

The site uses a self-reporting mechanism for starting strength and has on the mobile app for quite a while and it has been the default method on the website since v3 became the default.

Yes, but Glicko is just a refinement of ELO system by integrating the deviation of the rating. With other words, that does not make a difference. The ELO system is false due to the conservation of ELO points in the system, no matter if Harvard's Glicko, FIDE; UCSF, doe not matter at all. 

Put very simple, so that everyone can understand this: there are not enough ELO point in the system so that all players are competing for only half available points than normal.

As a result, everybody has 400-500 ELO point below his or her real-world chess skill level !!!

I call BS on this one.  And I reckon you're not a doctor either.  

This again shows how little you are willing to understand. You didn't understand a word again, right? You also do not know how to give an argument, BS is not an argument per se, even if a doctor tells you, you would not want to understand. You should learn how to exchange arguments dude.

Ratings measure past performaces and are based on the pool of players and ratings played against. If a rating system starts at a higher arbitrary number, then most of the ratings in that pool will be higher.

Rating pools are also not finite, or zero sum. The ratings gained by one player are often not the same as the other player's loss. So, saying players are playing for limited amount of available points is incorrect.

When taken over many games, with a wide range of opponents (i.e. a large pool), ratings will tend to level out for a particular player, once enough games have been played; barring that player getting new skills and increasing their performance. That is regardless of new players entering or old players leaving the pool.

While ratings are an estimate of overall strength, you have to realize they only give statistically expected outcomes, over many games, between rating differences.

Thanks, these are better arguments and a better exchange than before but it is still not right. 

Roughly speaking, the number of points gained equals the number of points that are lost. 

First, in most games, it is exactly the same number and only in very few games it is different and equals out, due to the Glicko deviation correction in both directions, up and down, thus it equals out.  Like an error bar showing the standard deviation in both directions, it equals out at the average.

If a string player wins against a weak player he or she will only get few points, and if the weak player wins he or she will get more points. Equally happens vice versa but in the opposite direction, which leads to a conservation of ELO points in the system, it stays roughly the same and even drops due to the reason that all players get better over time, right? But the system does not account for this too. Hence, by evidence, many hundreds of ELO points are missing. 

Just compare the FIDE System to the Chess.COm system and you will understand what I mean: 

More than 500 ELO points are missing between FIDE rating and Chess.com rating. 

I checked this with many methods, at least 500 ELO points are missing using independent experimental testing, but don't call me a nerd