Proposal: New Tournament Divisions

[Eric Strange]

I like your proposal but I must disagree with one aspect.

2201+, 1901-2200 (U2200), 1701-1900 (U1900), and 0-1700 (U1700)

Given that I have been watching ratings very very close to make sure they are calculating correctly and seeing how the new system is going to effect current ratings I believe your numbers should be adjusted. I have noticed an extreme decrease in ratings!!!

i believe that a master level player may eventually be a 2100-2300 and grand master be 2400 For instance Shane Mccosker was in the top 20 and 2300+ and has dropped to 2200 and is 30th. Alex lost over 100 points, and many other players are losing points. I would assume a 100 point average decrease for top rated players. very soon I believe 2200+ will be a very small division. Given your numbers I think this would work perfectly.

RATINGS START OUT AT 1600 FOR PROVISIONAL AND NOBODY CAN DROP UNDER 1000!!! SO JUST ADJUST EVERYTHING THIS WAY.

1000-1600, 1601-1850, 1851-2100, 2101+ <---- THIS WAY ALSO GIVES AN EVEN 250 POINTS FOR THE 2 MIDDLE DIVISIONS. I DO BELIEVE THAT PLAYERS SKATING CLOSE TO EITHER DIVISION WITH SUFFICIENT "BACKUP" SHOULD BE ALLOWED TO PLAY UP.

let me know what you think John. I hope that you trust my opinion with how ratings will pan out and that these numbers I gave will work perfectly. I spend at least 2 hours a week studying how elo is affecting players current ratings.

[liam stephens]

The COMPRESSION SYNDROME*

I am by no means an expert in Rating Systems, but I have never seen an explanation of how to deal with what I call the Compression Syndrome.

This may best be illustrated by a typical example with Kindergarten level maths.
Swiss Tournament, 3 Move, 10 rounds, a single ballot (2 games) per round.
1. Scored by rounds, 2 points for a won round, 1 point for a drawn round.
Let’s assume Player A scores 8 won rounds 2 drawn rounds for a total of 18 points out of a possible 20
Therefore percentage score is 90%
2. Score by games, 2 points for a won game, 1 point for a drawn game.
Player A wins 8 rounds with a win and a draw in each round so, 8 X 3 = 24 points
and draws the last 2 rounds, 4 draws = 4 points
Total 28 points out of a possible 40
Therefore percentage score is 70%
So there is a massive 20% difference in scoring by games as compared by rounds.
The effect of this on the ratings calculations is to compress them into a much narrower range or bandwith when the tournament is scored under the points per game system.
As you are aware that is how the majority of our tournaments are run.
Only in the masters section of the US Nationals (3 move) is the scoring done by rounds.
(I am not counting Go as you Please tournaments here. For example in the British and Irish Freestyle we have single game rounds with colours allocated, much in the same way as they have in Chess tournaments)
The scoring by points per game could well be the reason for the drop in players ratings that is referred to above.

Historically the Elo System proved inadequate for Chess ( some might even go so far as to claim it is unsound ?) in that over time the ratings became inflated and ever more self wish fulfilling. It was for that reason that alternatives such as the Glicko system were introduced. Glicko is supposed to address the inadequacies of Elo but I very much doubt that it deals with the Compression Syndrome, as that is a feature unique to the balloted openings with scoring by games instead of rounds.

In my view, unless and until this anomaly is addressed the Ratings will continue to be a topsy turvy mish mash affair.

*A nice title for a Matt Damon movie.

[Eric Strange]

Luckily there is no argument for game vs round. My system only allows scoring one way. 1 point = loss and draw, 2 points = 2 draws, 3 points = win and draw, 4 points = 2 wins and that is per round.

I ran into a tournament in cross tables that only shows 1-2 points and I could not calculate it because if I typed 1 or 2 points it would calculate it as 1 draw and a loss or 2 draws in the round which would not have done the players satisfactory.

As I said in a post a few weeks ago. I am only accepting 2 forms of cross tables and that is, 2 games per round, 1-4 points per round.

I will accept accumulative points, so if I win 2 rounds I will have 8 points (I can do the simple math in my head.

But my system will only take 1-4 points per round.

Liam,

Chess adopted Glicko which is an even worse system than Elo. It addresses NONE of the issues all it does is give players who play in a tournament every few years an advantage over the players who play in tournaments monthly. poppycock

Elo is the most widely used on any gaming site you go to. and to be honest ratings are better compressed than spread out as much as david butlers system. It if everyone's ratings are relatively close then there will be more competition. I like competition

[rich beckwith]

John & Eric,

Good discussion and I will make some remarks. What Eric says about declining ratings values is interesting and emphasizes the need that cutoffs developed in the 1990s may need an update.

John, I’m not clear if you are proposing a change in # rounds/games. You hinted at 12-people divisions with possible idea of a round-robin (which could be done in 11 two-game rounds = 22 games with 11/12 people). This is fine if you have 36 or 48 players with exactly (or very close to) 12 people in each pre-determined ratings group, but you will likely have other scenarios over time, so Swiss may still be necessary. I agree round-robin is fairer if the numbers work out just right (with no withdrawals). Also 22 games would be a decrease from this year’s proposed 26 games, which was already a decrease from 32 games back when we played 8 four-games rounds (and is still the usual preferred scenario for 3-move Nationals). I’m not claiming what the optimal number of games should be here – just noting differences.

I think a bigger issue that needs addressed first is how many divisions are appropriate. In recent GAYP Nationals , we really didn’t have enough people in a division for the Swiss system to work well. Recall we were attempting to do Swiss for 12 rounds with fresh pairings (plus one repair round) with only 16 or 18 people in a division. This resulted in high finishers getting a bye in the last round (sometimes with all but one or two players getting a bye), and the program was having trouble making optimal late-round pairings (although the pairings were legal), which received a number of criticisms. The 2008 3-move National (8 rounds) was also low-attended and shared many of these same Swiss issues, so the problem is not unique to our 13-round GAYP format. In order to play an optimal Swiss for 11-13 rounds, you really should have at least 20-24 people in a division. Many years ago, we had this many people (or more) in three divisions. However (and unfortunately), I don’t think three divisions is sustainable or else the pairing anomalies we have seen in recent years will continue. However, reducing to two divisions may be unpopular as it breaks with tradition, and there would be one less division title to strive for. There are also possible mixed formats – like start in one group and breaking out into divisions mid-tournament after Swiss has sorted people out. (Seeding could be used to sort players faster.) But such systems would also break from tradition and could get messy. Also, I believe players really want to know what division they will voluntarily be competing in (and the number of days and games) before traveling to the tournament. Given the pros and cons of all possible ideas, perhaps the best 2-3 options should be formulated and put forward for a pre-tournament ACF member vote.

While I am at it, would Alex’s “ballot scoring” system be a better alternative for Majors/Minors 3-move Nationals than the current system of game-scoring and 8-point rounds? [Each three-move ballot is independently scored 2-0 or 1-1, times two ballot played, so if you have three draws and a loss, you get one point, whereas you used to get no points with round-scoring, which was voted out in 2005 for Majors/Minors.] I’m curious how the Majors/Minors feel about 8-point bonanzas now that we have run the experiment a few times.

Another point…I think players should be well aware of format long before traveling to Medina, and significant format changes preferably require more than a simple exec vote. Technically, rule proposals must be accepted three months before the tournament takes place. However, I would welcome much-needed complete proposals at the pre-tournament meeting for possible vote & subsequent execution at future Nationals, pending Exec approval. Alternatively, a more comprehensive vote-by-ACF-bulletin scenario could be done at a later time if more proposal time is needed, although you might have people casting votes who don’t even plan on coming to Nationals.

In short, your input is appreciated – I think this would be a good opportunity to revisit format (# rounds, # of games) and ratings cutoffs all in one package.

[Eric Strange]

OKAY WHOA,

Number of games per round is OFF the table.

I don't think you all understand how difficult it is to program what we did for the ratings to calculate.

ANY CHANGE TO POINTS PER ROUND OR AMOUNT OF GAMES PER ROUND WILL MAKE THE TOURNAMENT INCALCULABLE IN MY SYSTEM. THIS IS NOT AN OPTION IF YOU WANT TO HAVE THE TOURNAMENT CALCULATED.