The strongest masters open in the world, no English entries....

[Andy McCulloch]

Please read Jack's post again.

He clearly indicates that the morning, afternoon and evening tournaments are simultaneous, not the games. Agreed it is a slight misuse of the term simultaneous, but easily understood.

[Stewart Reuben]

Dear Andy,
Please then read my post. I did not suggest that 3 rated games could not be played on the same day, at different times. I wrote that not even two could be played simultaneously. It falls foul of Article 11.3a, of the Laws.
3.1 of the Rating Regulations state that there must be potentially no more than 12 hours play in one day. Thus Jack could fit in 3 games in one day, as could anybody else, provided the rate of play is not greater than 60 moves in 2 hours. So 93 games is possible in 31 day months. It could even be one tournament or match.
I wrote the rules that way so that tournaments with 1 game Friday, 3 Saturday and 2 Sunday could be FIDE Rated. Such tournaments are no longer popular.

[Christopher Kreuzer]

3.1 of the Rating Regulations state that there must be potentially no more than 12 hours play in one day. Thus Jack could fit in 3 games in one day, as could anybody else, provided the rate of play is not greater than 60 moves in 2 hours. So 93 games is possible in 31 day months. It could even be one tournament or match.

I wonder if anyone has considered trying to set a world record for most number of rated games over a period like that described by Stewart? It could, as he says, be a match. Playing three standardplay rated games a day over a whole month does sound like a recipe for madness, though...

[IM Jack Rudd]

Based on the other thread, I am strongly tempted to try to arrange a 93-game match between me and Steve Crockett.

[Stewart Reuben]

I have sometimes advised players with a high k factor, who are clearly improving, to play a large number of games in one period. It wasn't at my suggestion, but in, I think 1984, Neil McDonald played a large mumber of games in one year, presumably more thn 70. His rating went up to 2395 (maybe 2495) withut ever having a TPR as high as that. The rating period then was one year.
Let us suppose I (rated 2050) play a 90 game match in one month with an improving junior rated the same. Let us suppose he wins 65-25 (a score plucked out of my head without doing any calculation.
Now we do the calculation. His expected score is 45. Thus he gains (65-45) x40 = 800 rating points. He is now 2850. I go down to 1650.
I am even a little bit disappointed that nobody has done this, nor even played 60 in one month as far as I know.
It is a fault in the FIDE Rating System when they increased the k factor to 40 for players under 18. k=40 was after my tme on the QC.
No such fault exists for rapidplay or blitz as I introduced a rule that, if a player played more than a certain number of games in a period, his new rating is calculated based solely on the new information. Playing 70 blitz games in one day is not unusual.

Jack if you did organise such a match against anybody, I suggest you take the precaution of having a FIDE arbiter in charge.

[Alan Walton]

I was discussing with a GM in one of my excusions abroad during dinner about the k=40 problem, and he fundamentally thinks there is a flaw with the system when your new rating gets published higher than your actual tournament performance achieved

From Stewart's example the actual tournament performance is only 2227, for the person scoring 65/90; so ideally the maximum points gained from this one example should be limited to 177, this doesn't matter if you K=40, K=20, or K=15 as it would still mean you will be higher rated than your actual performance

[IM Jack Rudd]

I was discussing with a GM in one of my excusions abroad during dinner about the k=40 problem, and he fundamentally thinks there is a flaw with the system when your new rating gets published higher than your actual tournament performance achieved

Elo systems have issues there, yes. It only really starts getting problematic when a player double-overshoots his performance rating, so that his new rating is a worse estimate than the previous one. It's unlikely to happen much, even with k=40, for monthly lists, but for yearly ones it was a huge problem.

(If you really want a system that can handle arbitrarily large amounts of input data, Elo is not what you want to try to adapt - Clarke systems are much better suited for the purpose.)

[NickFaulks]

No such fault exists for rapidplay or blitz as I introduced a rule

No you didn't!

[NickFaulks]

I was discussing with a GM in one of my excusions abroad during dinner about the k=40 problem, and he fundamentally thinks there is a flaw with the system when your new rating gets published higher than your actual tournament performance achieved

That is not a huge insight, but it is absolutely right. The rapid and blitz rating calculations have an adjustment to meet this point.

The main problem now with the standard rating list is that the calculations for the October list include games played not only in September, based on the player's September rating, but also some based on his August rating. This means that k=40 can apply to more games per period than I had realised, and persists for longer.

Fortunately there is at present only one really bad example, but it is a shocker.

https://ratings.fide.com/card.phtml?event=2070901

This player has done nothing wrong, he just happened to get some fairly good results on very beneficial dates.

We have known for a long time what needs changing, but it is evidently difficult to do. edit - by which I mean difficult to change the program, not the rules.

[Roger de Coverly]

This player has done nothing wrong, he just happened to get some fairly good results on very beneficial dates.

In one summer campaign at the age of 13 or 14, he has got enough rating points to qualify him as a GM provided he can get the Norms.

We have known for a long time what needs changing, but it is evidently difficult to do.

In the days when lists were six monthly, from time to time a rule appeared that restricted the maximum gain in a single rating period.

A problem with framing such a rule is that a jump from 1300 to 1600 or 1600 to 1900 is also legitimate but has less consequences than a jump from 2300 to 2600.

Perhaps the K=40 for under 18s needs to be dropped to a lower rating, so the furthest such players can get at K=40 is the CM standard. I was thinking that 2000 might be suitable. So once a player reaches 2000, they can no longer jump directly to FM with K=40 and 300 point gain, but no further than within sight of CM with K=20. It still leaves anomalies that a player with a 1999 rating could reach FM with one leap, but a player of 2000 cannot.

[NickFaulks]

In one summer campaign at the age of 13 or 14, he has got enough rating points to qualify him as a GM provided he can get the Norms.

My guess is that his playing strength is at least 2400 now and will be 2500 by the time he gets three GM norms. The worry is that his very high rating, from which he can only drop, will actually deter him from playing.

[NickFaulks]

Perhaps the K=40 for under 18s needs to be dropped to a lower rating

There's no need to discuss changes to the system, which was determined after massive historical analysis. The problem lies not with the structure, but the implementation, ie the precise timing of the calculations and the starting ratings on which they are based. This will be made right, it's just going to take longer than I hoped.

[Roger de Coverly]

ie the precise timing of the calculations and the starting ratings on which they are based.

If an event started in July 2015, it would have used the 1st July ratings. If it didn't finish before the cut off date for submission, it would fall into the calculations for 1st September rather than 1st August. This was usually regarded as not a problem, but perhaps if it makes a difference between a K=40 and a K=20 calculation, then it does. The ECF gives itself longer to do the calculations, even being prepared to wait for an event such as Hastings to finish because it crosses a grading boundary. The Kidlington Congress at the start of February regularly pairs on the basis of July/August grades but has the January ones used for the grading calculations.

Would FIDE really change the rules so that a tournament that started and perhaps even finished in July would use the 1st August rating list because it was processed as part of the 1st September calculations? The ECF would do this, but it's not a regular feature as it applies only if a set of results is submitted too late for inclusion.

[Ian Thompson]

Dear Andy,
Please then read my post. I did not suggest that 3 rated games could not be played on the same day, at different times. I wrote that not even two could be played simultaneously. It falls foul of Article 11.3a, of the Laws.

So why confuse people by quoting Mike Truran when your point has nothing to do with Mike's response to Jack's post?

3.1 of the Rating Regulations state that there must be potentially no more than 12 hours play in one day. Thus Jack could fit in 3 games in one day, as could anybody else, provided the rate of play is not greater than 60 moves in 2 hours. So 93 games is possible in 31 day months. It could even be one tournament or match.

I don't see anything in the Rating Regulations preventing someone from playing in one tournament with 3 rounds a day played between midday and midnight and another tournament with 3 rounds a day played between midnight and midday. That would give you 186 games in a month.

Nor do I see anything preventing someone from playing in, for example, 24 tournaments a day with each tournament having 1 round per day, the games in each tournament starting on the hour at a different hour of the day from all other tournaments. That would give you 744 games in a month if you managed to complete every game within an hour, so you were able to play the next game in the next tournament. Extend this approach to more tournaments and shorter intervals between them and the theoretical number of games in a month is unlimited.

[Stewart Reuben]

1.1 of the Rating Regulations. For a game to be rated each player must have the following minimum periods in which to complete all the moves, assuming the game last 60 moves'
Where at least on of the players in the tournament has a rating 2200 or higher, each player must have a minimum of 120 minutes.
This is what applies for Jack as he is rated 2200+. It doesn't suggest that he can play more rapidly and thus go to play yet another game.
1600 or higher a minimum of 90 minutes.
U1600 60 minutes.
3.1 There must not be more than 12 hours play in one day. I just don't understand Ian Thompson's idea of playing 24 tournaments a day, which would give thinking time of 30 minutes per player for each game, provided he played 24 hours. Nor do I understand how he gets in 6 games where at least one player in the tournament is 1600+.
There used to be a rule of not more than 3 games a day. This was taken out of the rules in 2012. Thus the only limit is 12 hours per day. Ian could currently play 4 rated games in one day as he is just under 2200, provided his opponents were all also under 2200.