Kurotto Kingdom (10th May - 30th May) has started Discuss
SM 2022 R4 - Math & Irregular (27th May - 1st Jun) has started Discuss
LMI Players' Rating System270 posts • Page 4 of 11 • 1 2 3 4 5 6 7 8 9 10 11
@ 2011-05-16 6:13 AM (#4424 - in reply to #4392) (#4424) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-05-16 6:13 AM

debmohanty - 2011-05-13 7:31 PM

Also we discussed few posts back in this thread that we'll add an option for players who would want to consider the results to be rated.
It will be implemented starting MAYnipulation

This is how it will look like


Do we still remove 0-scores after that?

This didn't seem to work. Either the check box was too small or the purpose of it was not clear.
Most of the players having zero score still have the check box selected.
@ 2011-05-16 11:06 AM (#4427 - in reply to #4424) (#4427) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-16 11:06 AM

debmohanty - 2011-05-16 6:13 AM

This didn't seem to work. Either the check box was too small or the purpose of it was not clear.
Most of the players having zero score still have the check box selected.
So we'll exclude all zero scores from ratings...
@ 2011-05-17 4:48 PM (#4437 - in reply to #4427) (#4437) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-05-17 4:48 PM

rakesh_rai - 2011-05-16 11:06 AM

debmohanty - 2011-05-16 6:13 AM

This didn't seem to work. Either the check box was too small or the purpose of it was not clear.
Most of the players having zero score still have the check box selected.
So we'll exclude all zero scores from ratings...

To put in perspective, here are the numbers -
Out of 116 players who started the test, exactly 7 marked that their results shouldn't be considered for ratings. (1 of them got non-zero score)
Of the remaining 109 players, 32 got zero scores.
@ 2011-05-17 6:34 PM (#4438 - in reply to #1357) (#4438) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-17 6:34 PM

Updated LMI Puzzle Ratings after MAYnipulation (May 2011 LMI puzzle test), and LMI Sudoku Ratings after the April 2011 LMI sudoku test are now available.

The ratings are based on the new logic shared earlier. Four players find a place in the Top 10 in both lists - motris, deu, nikola and misko.

Overall 487 players (from 45 countries) are included in the sudoku ratings and 425 (from 44 countries) in the puzzle ratings.
@ 2011-05-18 4:20 AM (#4441 - in reply to #1357) (#4441) Top

kiwijam



Posts: 178
100202020
Country : New Zealand

kiwijam posted @ 2011-05-18 4:20 AM

Rakesh, good work on the ratings list.
I can see I have a rating number now, and like most of us I'd like to improve it! But I don't know the numbers that were used to calculate it (like my last 12 scores), or the numbers for other puzzlers with similar ratings (these are the people I want to beat in the monthly test to move higher).
Is it possible to have a link to a full table also (like the 'pink' table of examples you published above)?
@ 2011-05-18 12:00 PM (#4442 - in reply to #4441) (#4442) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-18 12:00 PM

kiwijam - 2011-05-18 4:20 AM

I can see I have a rating number now, and like most of us I'd like to improve it! But I don't know the numbers that were used to calculate it (like my last 12 scores), or the numbers for other puzzlers with similar ratings (these are the people I want to beat in the monthly test to move higher).
If you click on your name, you go to your profile page that contains scores in all tests that you have taken part. These are the numbers based on which your ratings are calculated. Note that your scores in TVC are not included for ratings. You can see the history of scores for any player this way.

Is it possible to have a link to a full table also (like the 'pink' table of examples you published above)?
We have shared the computation logic already. But we feel that a full table of numbers used during ratings calculation can be cumbersome and complex, so it may not be needed online.

But, since you have asked, your normalized scores in the three tests are 601, 670 and 723. And you have played only 3 tests so far. Although the weighted average rating (based on these tests) is 665, you are penalized by a small factor to arrive at your final rating which is 651 currently. Here is the 'pink' table for you:



(kiwijam.png)



Attachments
----------------
Attachments kiwijam.png (10KB - 1 downloads)
@ 2011-05-18 12:06 PM (#4443 - in reply to #4442) (#4443) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-05-18 12:06 PM

May be we should enhance the Profile page to display the pink table and related computations. It needs some amount of work. So no promises right now.
@ 2011-05-21 9:11 PM (#4467 - in reply to #4443) (#4467) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-05-21 9:11 PM

I would love to see "my pink table" in my profile page! And I think I speak for many puzzlers here. :-)

Have fun, Stefan
@ 2011-05-24 2:11 PM (#4496 - in reply to #4467) (#4496) Top

Administrator



20001000100100100100
Country : India

Administrator posted @ 2011-05-24 2:11 PM

euklid - 2011-05-21 9:11 PM

I would love to see "my pink table" in my profile page! And I think I speak for many puzzlers here. :-)

Have fun, Stefan
The profile page has the pink table now.
@ 2011-05-25 1:06 AM (#4507 - in reply to #4496) (#4507) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-05-25 1:06 AM

My Puzzle Rating is not correct. The weighted average of my NS should be 707, my rating is 724, though.

Are you sure that you have implemented the weighted average of the NS correctly? I arrive at approx. 724 when I compute a non-weighted(!) average of my results, including the EvergreensI-result which should have weight zero by now.

In the pink table I would love to see the EvergreensI data (NS,...) also. It has weight 0 of course but the data is available and interesting nevertheless.

Thanks,
Stefan
@ 2011-05-25 10:36 AM (#4508 - in reply to #4507) (#4508) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-25 10:36 AM

euklid - 2011-05-25 1:06 AM

My Puzzle Rating is not correct. The weighted average of my NS should be 707, my rating is 724, though.

Are you sure that you have implemented the weighted average of the NS correctly? I arrive at approx. 724 when I compute a non-weighted(!) average of my results, including the EvergreensI-result which should have weight zero by now.
Thanks for pointing out this discrepancy. There are no problems with the implementation; however, we did find that your Twist score was completely missed in the ratings calculation. The ratings will be corrected today.

In the pink table I would love to see the EvergreensI data (NS,...) also. It has weight 0 of course but the data is available and interesting nevertheless.
OK. We'll include the ratings data (PS/RS/NS) for zero weight tests as well.

Edited by rakesh_rai 2011-05-25 10:37 AM
@ 2011-05-25 12:12 PM (#4509 - in reply to #4508) (#4509) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-25 12:12 PM

rakesh_rai - 2011-05-25 10:36 AM

In the pink table I would love to see the EvergreensI data (NS,...) also. It has weight 0 of course but the data is available and interesting nevertheless.
OK. We'll include the ratings data (PS/RS/NS) for zero weight tests as well.
This is done. Now you can see your normalized scores for even those tests which are not included in current ratings (like EG1).
@ 2011-05-25 8:10 PM (#4514 - in reply to #4508) (#4514) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-25 8:10 PM

rakesh_rai - 2011-05-25 10:36 AM

euklid - 2011-05-25 1:06 AM

My Puzzle Rating is not correct. The weighted average of my NS should be 707, my rating is 724, though.

Are you sure that you have implemented the weighted average of the NS correctly? I arrive at approx. 724 when I compute a non-weighted(!) average of my results, including the EvergreensI-result which should have weight zero by now.
Thanks for pointing out this discrepancy. There are no problems with the implementation; however, we did find that your Twist score was completely missed in the ratings calculation. The ratings will be corrected today.
The Puzzle Ratings have now been updated. Also, the sudoku ratings have been updated to rectify one error in Sampler Platter scores.
@ 2011-05-26 1:52 AM (#4516 - in reply to #4514) (#4516) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-05-26 1:52 AM

One more request of minor importance. Is it possible to mark those players' names in the result tables who have opted not to be included in the LMI rating?

MaYnipulation has 83 participants according to result table but 82 participants according to the LMI rating. If the one player were marked somehow (e.g. asterix at his name), everybody could calculate his LMI rating for himself and "verify" the correctness of the official LMI rating. :-)
@ 2011-05-26 9:50 PM (#4520 - in reply to #4516) (#4520) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-05-26 9:50 PM

That is a reasonable request - we should do it in next test.
Thanks for all the suggestions, however minor they may be, they certainly are missing.

Please don't hesitate to suggest any other things!
@ 2011-05-28 12:26 PM (#4536 - in reply to #4516) (#4536) Top

Administrator



20001000100100100100
Country : India

Administrator posted @ 2011-05-28 12:26 PM

euklid - 2011-05-26 1:52 AM

One more request of minor importance. Is it possible to mark those players' names in the result tables who have opted not to be included in the LMI rating?

MaYnipulation has 83 participants according to result table but 82 participants according to the LMI rating. If the one player were marked somehow (e.g. asterix at his name), everybody could calculate his LMI rating for himself and "verify" the correctness of the official LMI rating. :-)

This is done. You should see * next to the name for players who opt not to include their score in ratings.
@ 2011-05-31 8:53 PM (#4643 - in reply to #1357) (#4643) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-05-31 8:53 PM

The sudoku ratings are updated after Something is Missing.

motris, deu, misko, nikola and jaku111 are the Top 5.

This was the 13th LMI sudoku test and, hence, Mastermind Twins are excluded from the ratings now.

Edited by rakesh_rai 2011-05-31 8:53 PM
@ 2011-06-02 11:39 AM (#4662 - in reply to #4643) (#4662) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-06-02 11:39 AM

My Sudoku rating is 623 now, which is incorrect. It should be 681. Please check.

Stefan
@ 2011-06-02 12:09 PM (#4663 - in reply to #4662) (#4663) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-06-02 12:09 PM

This looks like a big computation mistake somewhere in the system, hope I'm wrong.
@ 2011-06-02 6:31 PM (#4665 - in reply to #4662) (#4665) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-06-02 6:31 PM

euklid - 2011-06-02 11:39 AM

My Sudoku rating is 623 now, which is incorrect. It should be 681. Please check.
The rating system, for some reason, likes you.

Yes. This was another error which had gone unnoticed somehow. Thanks for spotting it. Those who scored equal to the test median score are affected. For example, you scored 145 in the last sudoku test which happened to be the median score as well. And the system was giving PS as 50 instead of 500. The correct ratings should be up soon.
@ 2011-06-03 7:29 PM (#4671 - in reply to #4665) (#4671) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-06-03 7:29 PM

Thanks for finding the mistake and correcting it. Still there remains a small(?) mistake since I have a sudoku rating of 688 now but should have 681.

My Normalized Scores (NS) given at my profile are correct, i.e. they are the same that I did compute myself.

But then, the rating must be:
(492+837+696+643*.8+721*.8+711*.6)/5.2=681

Since rounding errors are not sufficient to explain the difference of 7 points, there still must be a mistake.

Good luck for finding the error,
Stefan

P.S.: My prorated Score (PS) of the SomethingIsMissing test was always given as 500 at my profile. This proves that you have calculated the PS two times. One time to display it at my profile (PS=500) and one time to calculate my rating (there you used PS=50). Computing two times is always dangerous (if you adapt the rating in the future you will always have to do it twice...) but it might have some practical reasons. A similar mistake based on double computation must still be currently in place.
@ 2011-06-03 8:45 PM (#4675 - in reply to #4671) (#4675) Top

debmohanty




1000500100100100202020
Country : India

debmohanty posted @ 2011-06-03 8:45 PM

We are having an issue here - we've 2 different systems
1) Rakesh uses excel to compute all ratings and uploads after every test
2) What you see in profile page is directly computed from the score sheet

There are minor differences in computation in these 2 systems because of which we are seeing this difference.

As a first step, we are working on making sure that 1) and 2) are same.
As a future step, we are planning to get rid of 1) so that we've all the computation at one place. This will take a while because of the complexity involved.
@ 2011-06-11 1:42 PM (#4803 - in reply to #1357) (#4803) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-06-11 1:42 PM

The LMI Puzzle Ratings have been updated after Fillomino Fillia.

The Top 5: Motris, Deu, Uvo, MelloMelon and Nikola

Edited by rakesh_rai 2011-06-21 9:28 PM
@ 2011-06-21 9:28 PM (#4929 - in reply to #1357) (#4929) Top

rakesh_rai




Posts: 774
500100100202020
Country : India

rakesh_rai posted @ 2011-06-21 9:28 PM

The LMI Sudoku Ratings have been updated after L O G I D O K U.

The Top 5: Motris, Nikola, Jaku111, Misko and Deu.
@ 2011-06-23 9:54 PM (#4949 - in reply to #4929) (#4949) Top

euklid



Posts: 28
20
Country : Austria

euklid posted @ 2011-06-23 9:54 PM

Just in case that it is a still unknown bug:

My rank score (RS) for "Something is missing" has suddenly changed from 469 to 573. That would mean that the number of participants has jumped from 96 to 119,5 (my rank of #52 has not changed).
My rank score (RS) for "X'mas Special" has suddenly changed from 734 to 770. That would mean that the number of participants has jumped from 139 to 161 (my rank of #38 has not changed).

My prorated scores (PS) have not changed.

Stefan
LMI Players' Rating System270 posts • Page 4 of 11 • 1 2 3 4 5 6 7 8 9 10 11
Jump to forum :
Search this forum
Printer friendly version