Thanks for bringing this point up, James.


Yes. That's correct. Ratings can increase if (a) you do very well in the most recent test, OR (b) you had a bad performance twelve months ago (which goes out of rating calculations now) but have since done ok. Generally, we tend to correlate rating improvement with the first of these reasons, which is natural but not a complete view. LMI Rating system (Puzzles) takes into account performances over the last 12 rated puzzle tests, which roughly means twelve months. William had a (relatively) lower rating so far because of his (arguably) inferior show in the Forsmarts Anniversary contest 2013, where he was ranked 37th out of 108 non-zero scorers. In the latest ratings, he got rid of this baggage. And due to relatively stable performances over the last twelve months, his rating has improved and come close to his actual recent performances. I would say, "the rating system waited for a fair number of improved performances before registering an improvement." And, since he has not been a regular participant (yet), the rating has potential to improve further. I can see some analogy to FIDE ratings (chess), where it takes time to reach your actual rating levels.


Your argument is valid since willwc has played less tests in the recent twelve month period. For a more regular player, this argument is not valid. For example, if he had performed as well as you did on Pento Parade, he would have still reached a rating of ~833 to be 17th, a jump of 7. And, if he had got more than ~1140 on Pento Parade, he would have bettered a rating of 845 also. But, in general, yes, NS on this test would be lower because of the exceptional top score.


Agree with your point. This point has been discussed while devising the ratings formula. Also look at this post from motris summarizing the discussions around this. I am not saying we cannot make changes. But "Is an improvement needed for handling such tests with outlier performances?" is the question.

Also, you can refer to this earlier post. In the example, player D resembles willwc. So we take a straightforward weighted average of his NS to get to his rating.

You can make a similar argument for Neil as he only has 4 tests. But because two of his tests are marathons, where the PS is much higher for the top puzzlers compared to other tests, his rating average is much better.

Updated ratings after Irregular Variations.

Hi! Why do you list only the first 300? How can I find myself in the rating lists?

http://logicmastersindia.com/forum/lmi/ratings/?testType=S&full=yes...

Ratings updated after Sudoku Surprise (sorry about the delay)

After some impressive performances in last 5 sudoku tests, Kota moves to the top position.

Hello,
There has been a lot of sudoku tests these times, but few puzzle tests ( only Pentoparade in october and Forsmarts in december ).
Can we know when the next puzzle tests are planned ?

Sudoku Ratings are updated after Converse,

AND

Puzzle Ratings are updated after Puzzle Marathon.

Sudoku Ratings are updated after Unlucky 13.
Only three points separate the top 3 - jaku111, Kota and deu.

Also, for authors/testers, I think NumTests should include the tests they authored/tested.
For calculations, it should be ignored.

Thanks for the suggestions. You may also refer to the post - http://logicmastersindia.com/forum/forums/thread-view.asp?tid=119&s... for more details on the rating calculations.

So,before the ratings get updated after Round 2 of PR , I wish to state here an observation and a rare coincidence . My current LMI Puzzle ranking(133rd) and LMI sudoku ranking(38th) are exactly the same as my position in WPC(133rd) and WSC(38th) 2016 in Slovakia respectively !! :)

Couldn't have been more coincidental. Thanks for sharing.

Deb

Ohhh wow! I often tried to check mine too, but never got close :-)

When will the ratings be updated incorporating the SM Round 1 results?

The ratings shall be updated very soon.

What means "very soon"?

Apologies for a long delay in ratings update.

The Puzzle and Sudoku Ratings are now up to date after Evergreens PR test for puzzles and Odd Even/ Substitution SM test for sudoku.

The top 3 in Puzzle ratings are Endo Ken, Hideaki Jo and Kota Morinishi. For India, the top 3 has Prasanna, Rohan and Amit.

The top 3 in Sudoku ratings are Jan Zverina, Kota Morinishi and Seung Jae Kwak. For India, the top 3 has Rohan, Prasanna and Rishi.

The Sudoku Ratings are now up to date after Maths SM test.

Kota is now #1, Nikola is now #12 and Tantan is #17.

The LMI Puzzle Ratings are now up to date - after the Casual and Word PR test.

Neil moves up 6 places to #6, Walker moves up 4 places to #5 and Bram moves up 2 places to #8.

Can we have the link to the ratings placed somewhere visible from the main page, just like the top bar has a link to this thread?

On the main page, you can find the ratings on the right hand side under the heading 'Monthly Test Ratings'

Whoops, by main page I meant the main page of the forum, because it's almost exclusively the page I visit. But that's better than nothing.

Main page bottom right has the links.

The LMI Sudoku Ratings are now up to date after SM6.

Tiit is now #4, Wendy is now #16 and Kota remains #1 (gains 2 points).