Chess Tempo

Username:
Password:
/ Register

User Details

Username:
Blitz Rating:
Standard Rating:
Logout
December 01, 2008, 11:09:16 pm *
Welcome, Guest. Please login or register.
News: SMF - Just Installed!
 
Pages: [1]
Print
Author Topic: Points given and taken  (Read 723 times)
shajbat
Newbie
*
Posts: 10


View Profile
« on: August 04, 2008, 07:07:24 pm »

Hi Richard,

I have been away for 14 days.  I have noticed that the points given and taken away for problems have drastically changed due to only 14 days of inactivity.  Surly, a tactician’s ability deteriorates with time (usually years and months) but 14 days is not such a long time for such a drastic point manipulation.

Thx

Shajbat



Logged
richard
Administrator
Hero Member
*****
Posts: 987



View Profile
« Reply #1 on: August 04, 2008, 07:32:08 pm »

Hi Shajbat,

What you are experiencing is probably a combination of increased RD  and the impact of the new problems in the problem set. There was a significant update to the problem set made yesterday which added around 4k new problems and removed about 5k old problems that no longer passed updated problem quality tests.  Until the new problems have a chance to settle down you will likely notice some volatility (although more for the problems ratings than yours hopefully).

Regards,
Richard.
Logged
drahacikfm
Sr. Member
****
Posts: 416


View Profile
« Reply #2 on: August 04, 2008, 07:49:39 pm »

I do think there is a pretty big contrast between the rating changes you get when you are very active, and the rating changes you get after an absence.  But I think the problem is at the active end, not at the "absence" end.  What I mean is that maybe the minimum allowed RD of 30 is a little too low.  (It is 30, isn't it?)  Maybe the minimum should be 35 or 40.  Because when you have an RD of 30, it's like banging your head against the wall trying to change your rating... the changes are so small, correct or fail.
« Last Edit: August 04, 2008, 08:54:43 pm by drahacikfm » Logged

FIDE Master Drahacik
shajbat
Newbie
*
Posts: 10


View Profile
« Reply #3 on: August 04, 2008, 07:52:07 pm »

Thx for your reply Richard.


It has been a bloodbath today.  I take of care of a 2200 rated problem and I only get 1.5 points but I miss a 1900 rated problem and I lose 9 points.  I just checked other threads here.  It seems other tacticians are facing the same issue.  Hopefully, the dust settles soon.

Shajbat
Logged
shajbat
Newbie
*
Posts: 10


View Profile
« Reply #4 on: August 04, 2008, 08:03:53 pm »

drahacikfm,

You raise a very good point.  Before I left for vacation the average point decrements and increments were very minimal.  Perhaps that had to do with the fact that I was doing a large number of problems a day.  It seemed to me that initially one had to get a high rating in order to maintain that rating through small manipulation of points (positive or negative).
Now it seems that the problem is of a different sort.  I solve 2200+ rated problems and I only get a small increment and I fail a 1900 rated problem (which does not seem to be a 1900 rated problem but rather more difficult) and I lose a whopping 9 points). 
Anyway, I am all for a revision of RD here.


Shajbat
Logged
drahacikfm
Sr. Member
****
Posts: 416


View Profile
« Reply #5 on: August 04, 2008, 08:58:05 pm »

Shajbat:

I bet the 2200 problem you got right was a provisional problem.  That means it had very few attempts.  For problems with no attempts yet, you get only 20% of the usual rating change.  For problems with one attempt, you get 40%.  And so on, 3 attempts 60%, 4 attempts 80% and for more, the usual 100% changes.

The 1900 problem was probably a normal problem with 100% changes.
Logged

FIDE Master Drahacik
richard
Administrator
Hero Member
*****
Posts: 987



View Profile
« Reply #6 on: August 05, 2008, 12:16:34 am »

I've just uploaded a change which shifts the minimum RD from 30 up to 35.  This will allow regular users to see a slightly larger rating change per problem done. Of course this can be a double edged sword, you'll get slightly more for correct problems but also slight more punishment for incorrect problems.

Regards,
Richard.
Logged
fredthecat
Newbie
*
Posts: 4


Frank and Fred the Cat sitting.


View Profile
« Reply #7 on: September 05, 2008, 12:19:00 am »

What are you referring to when you use the initials RD?  Sorry all, I'm kinda a newbie...
Logged
drahacikfm
Sr. Member
****
Posts: 416


View Profile
« Reply #8 on: September 05, 2008, 01:18:56 am »

RD is the "rating deviation" in the Glicko rating system, used here on Chess Tempo.  In the FAQ there is a link to the complete description of Glicko, which is pretty complicated.  You don't only have a rating, you have the RD number too which you can see on the Stats page.

Basically if you do a lot of problems, your RD goes lower.  If you don't do any problems for days or weeks, your RD goes higher with time.  The rating change you get when you do a problem is more when your RD is higher and less when your RD is lower.  If your RD is less than 80, you are "active".

The idea is that if you went away for a long time, you might have gotten stronger, or you might have gotten weaker, so your rating is not very accurate and you should get bigger changes in your rating when you come back.
Logged

FIDE Master Drahacik
oded ross
Newbie
*
Posts: 23


View Profile
« Reply #9 on: November 23, 2008, 08:53:30 pm »

I've just uploaded a change which shifts the minimum RD from 30 up to 35.  This will allow regular users to see a slightly larger rating change per problem done. Of course this can be a double edged sword, you'll get slightly more for correct problems but also slight more punishment for incorrect problems.

Regards,
Richard.

Out of curiosity, what is the lowest RD recorded since that change? I'm not sure about the lowest I've seen for any user, either 35.10 or 35.13.
How many problems should one try in a certain timeframe to get his RD down to 35.00? or 35.10?

Another question: is there an upper bound limit to the RD? Some users who haven't tried either Blitz or Standard in over a year have an RD exceeding 500.
At what point the RD is high enough that it is treated just like a new user?
Logged
richard
Administrator
Hero Member
*****
Posts: 987



View Profile
« Reply #10 on: November 24, 2008, 12:42:00 am »

The lower end of the rating is clipped at 35 and the upper end at 350 (users who drift over 350 are automatically clipped down to 350 next time they play, but their >350 RD will show up in the user list until then).  You can see users at 35, but you'd have to see them at exactly the same time they had completed their last problem as what shows up in the list/stats page is the RD decayed to the time at which you view it, not the last time they played.

I don't recall the decay rate off the top of my head (and it is not the same as the standard glicko to allow for the fact that users can do hundreds of problems a day , something not possible in OTB play). 

Regards,
Richard.


Logged
oded ross
Newbie
*
Posts: 23


View Profile
« Reply #11 on: November 24, 2008, 12:52:22 am »

The lower end of the rating is clipped at 35 and the upper end at 350 (users who drift over 350 are automatically clipped down to 350 next time they play, but their >350 RD will show up in the user list until then). 
You can see users at 35, but you'd have to see them at exactly the same time they had completed their last problem as what shows up in the list/stats page is the RD decayed to the time at which you view it, not the last time they played.
Are new users also treated as having an RD=350 ?
As for the last paragraph, I already knew that, just was curious to see a required number (for an hour or a day) because I checked stats of users who weren't done yet after a few hundreds(!) of problems and yet to see an RD<35.10 (I tend to look for bizarre stats in other field of interest too...  Cheesy ).
Logged
richard
Administrator
Hero Member
*****
Posts: 987



View Profile
« Reply #12 on: November 24, 2008, 01:04:50 am »

Yes new users do start at 350.  I'll be updating the FAQ soonish so I'll try and get a statement in about the actual decay rate (the glicko code scares me a bit, so I'd rather not be reading it right before bed lest it give me nightmares :-)  ).

Regards,
Richard.
Logged
Pages: [1]
Print
Jump to: