From: "Stouten, Pieter [R&D/0467]" PHARMACIA COM> Date: 11 dec 2002 Subject: Elo ratings and grades Hello Shogi fans, I would recommend that anybody interested in discussing rating and grades-related issues first read the description of the system at . Item 2 has a ratings-grades correspondence table and items 2 and 12 together explain the promotion mechanism. --- bogin YAHOO CO JP> wrote: > Given the fact that Mr. Schnider recently did so well in > winning the recent International Shogi Forum and was > promoted to 5dan, why is he still listed as still only > a 3dan on the most recent rating list? Is there some sort > of time period that much take place before a promotion is > recognized? > Gert's results at the ISF were taken into account in calculating his new rating, and apparently he did not meet the FESA/USSF criteria for promotion. Typically, when a new player with an established grade (e.g., from Japan) enters the Pan-Atlantic Elo list, we accept that grade. However, we have currently no general mechanism to change a player's grade if he already has an established grade with the Pan-Atlantic Elo system. I am sure this will be a topic of discussion within the Pan-Atlantic Elo committee (Larry Kaufman, Hans Secelle, Asle Olufsen and me) and within FESA. > Also, in the same tournament, Mr. Kolomiyets also did very > well. And, according to the postings I read here his play > was of a very high level. > Subjective assessments of playing levels don't count, only results do. > Then, I am confused how he could be rated over 2100 and > yet only be considered a 2dan. > explains it. His rating is above the 4-dan lower bound (LB, 2080). That means that typically he would automatically be awarded a 3-dan grade. However, since he has not yet played 18 rated games, he only gets an automatic promotion to 2-dan. Once he has played 18 rated games and is above the 4-dan LB, he will automatically get his 3-dan promotion. In order to obtain a 4-dan promotion, he would need be above the 4-dan LB for 16 games or above the 4-dan midpoint (MP, 2160) for 8 games or above the 4-dan upper bound (UB, 2240) for 1 game. Once awarded, grades cannot be retracted, so we require sustained performance, which is why Kolomiyets has not been promoted to 4-dan yet. > Of course, you could say that he hasn't played enough > games yet but there are 4 people rated higher than he and > yet they have played even fewer games. And, they are all > 3dan or above. > Do you mean 4 people were rated higher or graded higher? Anyway, maybe these 4 people did not get their grades through the Pan-Atlantic ratings system. > So, even though there may be a disparity between different > players of the same rank, I find it hard to believe that > is on the same level as a 1500 rated 2dan or even an 1800 > rated 2dan. Isn't over 600 points a little big of a > difference for players of the same rank? > There may be two reasons why a player has a rating well below his grade: 1) the grade is based on a different system than ours and might never have been awarded by us, 2) a player performs significantly worse than during his peak. I am a good example of the latter case: I deserved my 3-dan grade when I got it (at least I prefer to think so ), but my current rating is near the 2-dan MP. The fact that some players perform well below their nominal grade should not be an argument to promote others to grades above that. That leads to an inflationary system. In Europe we had a system where players got promoted on the basis of the grades of their opponents (rather than their ratings). That led to inflation of grades, which is why we replaced it by an Elo ratings-based grading system. > If dan/kyuu is being determined solely on the basis of > one's numerical rating then would it be better to rate > all of the players above a certain thresh hold at a > certain rank? > Would you also then demote people if they fell below that rank? Anyway, in reality we have a system like you propose. We only require that players are established (i.e., have played 18 or more rated games) and we allow 3 ways of promotion (X games above LB, X/2 above MP or 1 above UB). The philosophy behind this is that if a player peaks once above the UB of a grade, one can safely assume he has that grade, while a player who just hovers above the LB has to sustain that performance for X games to prove that he is really of that grade. Hope this explains the system a bit. Ciao, Pieter