English Chess Forum

Effective immediately global ratings for all humans (non players too) have been set to 10,000 Draws will not affect ratings. Ratings will rise or fall based on the formula 250-number of moves of the loser. The combined gain & loss for a game will always be zero. There is no ratings floor or ceiling.

Your cooperation is most appreciated

Steven DuCharme wrote: ↑

Sun May 19, 2019 7:00 pm

Effective immediately global ratings for all humans (non players too) have been set to 10,000 Draws will not affect ratings. Ratings will rise or fall based on the formula 250-number of moves of the loser. The combined gain & loss for a game will always be zero. There is no ratings floor or ceiling.

Your cooperation is most appreciated

So if I beat Carlsen in 100 moves I improve my rating by 150 points but if I beat the lad I've just taught to play with fool's mate I get 248 points.
Interesting system!

Why not base it on the combined age of the losing players Grand-parents!

The system discourages resignation and weak play. So there!

I like it.

Always thought the grading system should in classes. A-Z

You start of as an 'A' and to advance you must beat 20 players in your class.
Draws or losses do not count (you never drop down a class) and you cannot jump classes say by
beating players a few classes above you. That counts as one win in your 20 till you get to the next class.

When you get to 'Z' you get a Z number.
if you beat a fellow 'Z' player it increases your Z number by one.

It would increase more aggressive chess (no fear of losing) and players would
stay in the game longer to get they classes up.

Tournaments could be 'classed' D's E's or F's etc knowing there was no sandbagger in there
because eventually they will go up a class..

Just a thought/

I am planning on introducing the following in the Autumn for my Juniors:

Every Junior player starts the year with a Junior rating which is the highest out of:
1. The highest of their FIDE ratings
2. Their highest ECF grading converted to a rating (ECF x 7.5 + 700)
3. An online rating of over 700 (capped at 1000)
4. An initial value of 700

Their rating is then adjusted in response to chess results at chess club and updated weekly.

Calculation method
Your rating goes up by between 1 and 40 points for winning (20 if the opponent has the same rating as you, 1 if they are 190 or more points below you, 40 if they are 200 or more points above you). Your opponent's rating falls by the same amount.

If you draw then you can gain up to 19 point (if your opponent is 190 or more points above you) or lose 19 points (if they are 190 or more points below you). If they have almost the same rating as you then it is unchanged.

Suppose A has a rating of 950, and B a rating of 830.
• If A beats B he gains 8 points to 958 and B loses 8 points to 822
• If A draws with B he loses 12 points to 938 and B gains 12 points to 842
• If A loses to B he loses 32 points to 918 and B gains 32 points to 862

Neill Cooper wrote: ↑

Mon May 20, 2019 9:35 pm

I am planning on introducing the following in the Autumn for my Juniors:

I find this ICC type system very practical. My only comment is that the rating interval of 200 points seems very low.

NickFaulks wrote: ↑

Mon May 20, 2019 10:15 pm

Neill Cooper wrote: ↑

Mon May 20, 2019 9:35 pm

I am planning on introducing the following in the Autumn for my Juniors:

I find this ICC type system very practical. My only comment is that the rating interval of 200 points seems very low.

Thanks Nick, good point. What would you suggest would be a better value?

You have calibrated the system so that the expected score per game is (difference in ratings)/200 + .5. The Elo system is calibrated to about (difference in ratings)/k + .5, where k is somewhere between 700 and 800 depending where you sit on the FIDE/ECF conversion debate. I'd suggest that your "20" becomes say 75 or you divide all initial ratings by, say, 4. If you do nothing weker players will look as though they are improving at the expense of stronger players.

Brian Valentine wrote: ↑

Tue May 21, 2019 9:10 am

You have calibrated the system so that the expected score per game is (difference in ratings)/200 + .5.

No, Neil is using (difference in ratings)/400 + .5. I would prefer (difference in ratings)/800 + .5, so the maximum rating difference would be 400.

Nick, Your formula is correct and I agree with your proposed change.