With the change to having official monthly ratings and elimination of most of the printed supplements, perhaps it is time to consider removing the 4 game restriction to appear in the official ratings list I run a monthly quick quad. and the new players in this event would like to see their first rating quickly.
A concern against the change is the small number of games that the rating is based on. This is perhaps not too bad when the number is 3, but what about players who only get one or two games in their first event.
I am just putting this out for discussion, not as a formal proposal
In general I’d be opposed to that idea since a provisional rating based on so few games tells you so little about their playing strength. With the MSA a player can find their rating quickly even if it isn’t “published” until they play 4 games.
The only exception I would consider is if the player’s first event was a 3 rounder such as a quad or 3 round swiss.
We run weekly quads, and we will use the MSA rating generated by a player’s first quad in order to determine which quad to place them in the following week. Most weeks we have 2 sections and the person usually still ends in the bottom quad, but sometimes we have 3 sections and the person may end out in the middle section.
The reason for requiring a minimum of four games to have a published rating are because with less than four games the accuracy of the provisional rating is far more uncertain.
In fact, I find Ernie’s suggestion odd, because with more frequent official ratings lists he wants less accurate first-time-published provisional ratings. With more frequent official rating lists, it seems it would be possible (and desirable) to make sure those ratings are more accurate, because it will take less time for a publishable rating to be published so we can use that shortened time to improve the accuracy of those initial published ratings by waiting until we have more data.
If anything, I would like to see a rating not be publishable until a new player has competed in at least two events totalling at least six ratable games, but I don’t know if the Ratings Committee would agree with that. (Under my suggestion it would still take two quads to get a published rating.)
I would also recommend that a player not get a publishable rating until that player has won or drawn at least one game. Over 1500 players received published ratings in 2006 without winning or drawing a single game.
Why would I recommend this change? Because the purpose of a rating is to be able to predict a player’s future results. Until a new player has found at least ONE person that he or she doesn’t lose to, I don’t believe we have any valid indication of that player’s strength with which to make that prediction.
Another 84 received published ratings in 2006 without losing or drawing a single game. While we don’t know the true strength of these players either, we at least can set a reasonable lower bound on them.
FIDE, incidentally, requires nine ratable games before a player’s rating appears on a FIDE Ratingt List (FRL), and a player must have at least one point in those events (and three ratable opponents) in order for a result to be counted towards that player’s FIDE rating. Moreover, FIDE does not make any information about a player’s “provisional” FIDE rating available.
How many of those ended up with ratings of 100? Maybe this is the change we ought to make instead of the previous suggestion about the 100 point floor.
Couldn’t a similar argument be made in terms of a reasonable upper bound existing for players that have yet to win or draw a game? I fail to see that these two situations aren’t analogous. Other than that, I agree that lowering the 4-game requirement would be inadvisable. If anything, the requirement should be increased, but I understand that this might cause problems with new players that want a rating asap.
971 of those 1500+ players who lost all their games wound up with a published rating of 100, without that floor who knows what their ratings would have been?
Of those who won all their games, their ratings ranged from 764 to 2402 with the median being around 1200.
I don’t consider the two situations particularly analoguous, partially because of the lower number of players who won all their games and also because there was a broad dispersal of ratings among those few who had won all their games but a large clustering at 100 from the many players who lost all their games.
I think as long as we give adequate notice of the change (such as having it take effect after the end of the current school year), the scholastic community can plan around it. What I’m not sure is if this is something the Board can implement (assuming the RC approves) or if it needs the Delegates approval. I suspect the former, but I’m not sure.
Moreover, unlike the RC’s suggestion for players rated 100, this situation is quite deterministic, players or coaches will KNOW if a player lost all his games and/or if he has played in only one event or fewer than 6 games.
When I was USCF Rating Statistician in 1964, the rule was that you needed 7 games to obtain a published rating, and the office was flooded with complaints from players who had played 4 or 5 games and had no rating. After we changed the 7 game requirement to 4, these complaints declined dramatically.
I would rather not go down to 3 which decreases accuracy, but 4 is a good number as most events are either 4 rounds or 5 rounds, and a fair number of players in 5 rounders miss a game due to a bye or forfeit win.
We have a fair indication of the upper boundary of the player’s strength, which is what is important. If a scholastic player starts out with a zero score and gets a rating of about 100-300, and there are Under 400 prizes, the player will likely have a better chance there than staying in unrated.
Likewise, an adult who starts out with zero might get a rating of 800 and will be eligible for Under 1000 or Under 1200 prizes which are generally easier to win than Unrated prizes. Yes, the 800 player might be overrated and really only 300 strength, but that’s not a big issue.
The FIDE method makes no sense, as if a player doesn’t quite earn a rating and subsequently improves strongly, it is actually helpful to that player’s later rating that the initial low rating was not “achieved.”