ECF ratings.

[Anthony Ibbitson]

Hi all - I was wondering whether someone might be able to answer my question with regards to the ECF chess ratings.

I've been fairly inactive on the chess scene recently and was wondering when my chess rating is likely to drop off. I'm currently rated 1754 K. My ECF grading code is 230785G and my membership number is 001480.

I haven't played any games during 2024, I played 2 games during 2023, 15 games during 2022 and I played 12 games during 2021.

I am going to try and play some tournaments this year (including Blackpool) but if I don't play anymore chess, when would I become unrated? I'm just trying to work out how soon I would need to play in order to avoid becoming unrated.

[Mick Norris]

Unrated: Any K or P rating where the player has no rated result in the previous 36 calendar months

ECF rating help

[Anthony Ibbitson]

So as long as I've played a rated game within the last 36 months, I should remain on the rating list?

[MartinCarpenter]

Although they never quite totally give up on your old rating even then. Basically they keep the records for 'ever' (until/unless advised of the obvious), just trust the data less and less as time goes on.

Your old rating is still the best initial guess that anyone has of your strength if/when you do return from a long time out.

[Christopher Kreuzer]

On the same topic, could a kind soul help explain how my initial rating after a period of time out from the game (mostly dictated by the pandemic) was calculated?

https://www.ecfrating.org.uk/v2/new/pla ... de=155285F

The questions I am looking for answers to are:

(1) At what point did I miss the chance to play with my old grade (175E in July 2020 and then converted to 2013 in Jan 2021)?
(2) Is my initial rating heavily dependent on my initial result (which wasn't great, but about what I could have expected)?
(3) If, going forward, I play consistently at the level of my grade in July 2020, how long will it take for my rating to rise back to that level?

I have tried to work this out, but am not confident I understand how the current system handles new grades and changes over time.

[Roger de Coverly]

The questions I am looking for answers to are:

You need to work through the "rating audit"
https://www.ecfrating.org.uk/v2/new/rob ... 1&domain=S

You aren't treated as a completely new player, rather you are deemed to have drawn with someone of your old rating as part of the "P" rating calculations.

Somewhere in the documentation on the rating site is a description of the rules which apply to make someone unrated.

[David Williams]

Having been initially baffled when this happened for me I think I worked it out.

When you re-start you have a provsional grade until you've played ten games. The thing that took me a while to work out was that a dummy game, a draw against someone of your old grade, is included. Presumably to prevent a 100% or 0% record, which is ungradeable. In your case your five actual and one dummy result of 50% against opponents grade 1840 on average gives you a provisional grade of 1840.

If you're still at 1840 when you reach ten games and then draw with a 2013 you will gain 4.6 points. As your grade goes up you will gain fewer points each time. A draw with a 2013 will get you 2.8 points when you're 1913. By the time you get to 2000 you'll only gain 0.4.

It does occur to me that if your next four games were at your old standard, 50% against average 2013, that would push the average grade of your opponents and therefore your own grade above 1900 when it ceases to be provisional.

[Christopher Kreuzer]

Thank you both. One follow-up question:

Would the old system have brought me back to the 2013 level (one can hope!) faster than in the new system, or is the answer there that it depends on the number of games played? In the old system, I was very active for a number of years (over 100 games in some years), and that effectively meant that very little of my grade carried forward from year to year. I expect to be less active now than I was then, so it will take a while for changes to show up. So should I expect at a rate of 15 or so games a year, for the rating to take a while to get back to the old level?

Obviously if I had performed well in my first tournament back, it would be a completely different scenario. Should the resulting rating history depend so heavily on the initial performance and take so long to 'correct'?

(I am aware that if I keep playing at this level, then I will stay at this level! )

[David Williams]

I'm not claiming any sort of expertise, but if player A draws ten games with 1800s he gets a grade of 1800. If he then draws ten games with 2000s his grade goes up to about 1850. If player B draws ten with 2000s and then ten with 1800s his grade will be 1950. Doesn't seem right.

Under the old system they would both have been 1900, and I thought the criticism of that was that player A's more recent results suggest he is now the stronger player.

[Christopher Kreuzer]

Thank you, David. That is the clearest exposition I have seen so far of what I feared was the case. Can anyone confirm David is right here?

[Brian Valentine]

Thank you, David. That is the clearest exposition I have seen so far of what I feared was the case. Can anyone confirm David is right here?

David is broadly right.

It might need some refining but if this expanation takes you down some rabbit hole then ignore it. It is how I got my head round the 700/games rule. The projection and any rate of change depends on the activity in any month where one plays. In essence concentrated activity has more effect than the same performance and total number of games over a longer period.

As a first approximation to the Elo formula one can get R1=R0(1-.029n)+ .029nP, where R0 and R1 are the starting and ending ratings respectively. n is the games in the month and P is the average of opponent's ratings +(wins-losses)*10.

This formula can be jogged forward using estimated opponent strength and scoring.

[Mick Norris]

You need to work through the "rating audit"
https://www.ecfrating.org.uk/v2/new/rob ... 1&domain=S

I've not seen that before

[Mike Gunn]

Thank you both. One follow-up question:

Would the old system have brought me back to the 2013 level (one can hope!) faster than in the new system, or is the answer there that it depends on the number of games played? In the old system, I was very active for a number of years (over 100 games in some years), and that effectively meant that very little of my grade carried forward from year to year. I expect to be less active now than I was then, so it will take a while for changes to show up. So should I expect at a rate of 15 or so games a year, for the rating to take a while to get back to the old level?
...

It depends what the gap is between your current rating and your "true rating". Here is a table I have just prepared which does the standard Elo calculation for a player with a current rating of 1600 and a true rating of 1800. It assumes that the player plays 3 games per month for a period of 36 months and draws with an 1800 player in every game. It assumes a K value of 20. The columns are: month number, rating at start of month, probablility of winning against an 1800 player, rating at end of month. I chose 3 games per month because of course under the old Clarke system your rating would have been (the equivalent of) 1800 after 12 months but under the Elo system it just reaches 1729. In my opinion the Clarke system is better than Elo for most of the chess we play in the UK because of the time factor inherent in the Elo approach.
1 1600. 0.240 1616.
2 1616. 0.257 1630.
3 1630. 0.273 1644.
4 1644. 0.289 1656.
5 1656. 0.304 1668.
6 1668. 0.319 1679.
7 1679. 0.333 1689.
8 1689. 0.346 1698.
9 1698. 0.358 1707.
10 1707. 0.369 1715.
11 1715. 0.380 1722.
12 1722. 0.390 1729.
13 1729. 0.399 1735.
14 1735. 0.407 1740.
15 1740. 0.415 1745.
16 1745. 0.422 1750.
17 1750. 0.429 1754.
18 1754. 0.435 1758.
19 1758. 0.440 1762.
20 1762. 0.445 1765.
21 1765. 0.450 1768.
22 1768. 0.454 1771.
23 1771. 0.458 1773.
24 1773. 0.462 1776.
25 1776. 0.465 1778.
26 1778. 0.468 1780.
27 1780. 0.471 1781.
28 1781. 0.473 1783.
29 1783. 0.476 1784.
30 1784. 0.478 1786.
31 1786. 0.480 1787.
32 1787. 0.481 1788.
33 1788. 0.483 1789.
34 1789. 0.484 1790.
35 1790. 0.486 1791.
36 1791. 0.487 1792.

Just in case I have got something wrong in my program I append the coding that does these calcs:
(initial values: opp_rating=1800, old_rating=1600, n_per_month=3, score=0.5
do i=1,n_months
diff=opp_rating-old_rating
P=1./(1.+10.**(diff/400))
new_rating=old_rating+n_per_month*(score-P)*20
write(*,fmt='(i3,2x,f7.0,2x,f5.3,2x,f7.0)')i,old_rating,P,new_rating
old_rating=new_rating
end do

[David Williams]

If you 'update' the old system so that your grade is updated monthly, your grade is based on your most recent 30 games. So 30 more games scoring 50% against 2013's makes you a 2013, irrespective of any previous grade. Under the new system, if you draw with someone graded 200 points higher than you your grade goes up by 5.2. If you do that 30 times in a month your grade goes up by 156. But if you only play 3 games a month that sort of increase takes a lot longer, because you will gain fewer points for each draw as your grade goes up. Your initial grade seems to haunt you virtually indefinitely. (I think.)