Chess Grading Information

download.gif (2097 bytes)

by T.C.Gavriel

(1998/99 BCF grade 191; Fide rating 2080; ICC rating fluctuates daily between 2250-2450)

2nd revision of this paper: 30/1/1999


Index

Introduction and personal views
British Grading System
FIDE/ ELO, USCF  rating systems
Internet chess ratings


Introduction

What is a grade?

Chess is a very complex game, and it is interesting to have some measure of mastery. If one takes the game seriously, then gradings can become important to have this measure, and to know one's relative strength is in the context of thousands of other chess players.   The rating systems attempt to assign a value indicating a strength of a player. Players of the game, can then have a guide to indicate how strong they are, and can obtain motivation / demotivation! for further improvement.

Titles also sometimes accompany ratings. For example, to earn the "FIDE master" title, one must:-

Should one be driven by grades?

There is a danger of being a slave to the rating system, and stop playing chess for fun, but rather play chess for grading. For example, having grandmaster draws to protect one's rating or playing very dull openings to reduce the possibility of defeat.

Examples include playing the London system as White (or the boring c3 sicillian), simply because the results one gets out of it are good. Opponents may become bored of chess as a result, and possibly give up playing.

How attitudes can change as a result of grading

Attitudes are sometimes change for better or worse as a consequence of grades. For example:-

Can we do better than a chess grade ?

Ultimately it would be nice to have a better guide to ones particular strengths and weaknesses. For example, a breakdown by opening of how strong one is! Technology today is quite advanced, and Chessbase 7 for example can look at a sample of one's games, and produce statistics for each opening, and for each opponent! In this way, one can also deduce indirectly which type of positions one is also strong at or weak at.

However maybe in the future, there will also be statistics available for various aspects of one play- for example play against an isolated d pawn, or play against doubled pawns. Such highly specific evaluation would be useful for the player who wishes to improve, and would thus go well beyond simply putting one's strength in the context of a population of other chess players. One could for example know that one has a grade of 2400 in the sicillian dragon defence!

(An interesting chess web site which provides a chess test, and gives a grading in different chess areas is Chess World)

Problems with grading systems

Sample size and unluckiness/ luckless

There are various problems with ratings which is evidence to suggest that one should not take gradings too seriously as a measure of our strength.

Unlucky runs

One for can example have an unlucky run of results where the opponents have for example played a particular opening where one is weak. Or the opponents could have been of a style of play which one is uncomfortable against.  The greater the game sample the grading is based on, the more likely the grading is to be accurate. This is because hopefully in the context of all the games played, the unlucky events are balanced by lucky events, and the final grade is a useful indicator of one's chess strength. The FIDE rating system for example should be based on a game sample of 30+ games to give reasonable confidence of a player's strength.

Name ambiguity

Another problem (especially with the BCF grading system) is name ambiguity. Often people with very common names/ names which clash with other players, have their grading results muddled with those players. This can lead to a significant percentage of inaccurate grades being published on the grading list. It is particular unfortunate (well for one of the players anyway) when 2 players who share the same name, have a large grading margin between them!

N.B. If one becomes a BCF direct member, one can get a breakdown of the tournaments/ league games, and the relevant grading implications. This could be a useful verification tool to help ensure one's results have not been confused with someone else.

Grading lag

In the UK, in order to set a game as rated, it has to be agreed by the BCF that the game will be qualify for a rating. All rating results have to be sent to the BCF, who will then process these results, and they will be incorporated in the next rating list. A recent change has been introduced by the BCF to have the rating list published every half year instead of every year. This can provide a more up to date assessment of player's strengths, and thus less uncertainty when playing an opponent. It would be an even more ideal situation however, to be able to connect to the BCF's database, and see what one's current rating is. However this prospect may not be realised in the next decade!

Juniors in particular are known to improve dramatically, and the old way of accounting for this, was to give them the benefit of +10 in their grade. This however can lead to extremely underrated juniors, especially those with access to Chessbase that have the time because of no work commitments to delve into all the lines of critical sharp openings, such as the sicillian defence.

Grading conversion issues

The BCF-Elo calculator on BCC on-line allow a rough translation of the BCF grading system to USCF or FIDE rating. Many countries simply use the FIDE/ELO rating system and therefore there is no translation issue.


British Grading System

The BCF is responsible for creating the British Grading system and publishing the rating lists. British grades go up to a maximum of 260 BCF. Around this grade are usually the top British Grandmasters.

The following table shows the implications of winning, drawing and losing against a BCF rated opponent:-

Case Your grade Opponents Grade Game outcome Difference in grade >50 Outcome
A X, e.g. 100 Y, e.g. 120 You win No You get grade Y+50, i.e. 170
B X, e.g. 100 Y, e.g. 120 You lose No You get grade Y-50, i.e. 70
C X, e.g. 100 Y, e.g. 120 Draw No You get grade Y, i.e. 120
D X, e.g. 100 Y, e.g. 180 You win Yes (Y greater than X) You get the maximum +90 reward (grades are assumed to be 40 points apart), i.e. 190
E X, e.g. 100 Y, e.g. 180 You lose Yes (Y greater than X) You get -10, i.e. 90
F X, e.g. 180 Y, e.g. 100 You win Yes (X greater than Y) You get +10, i.e. 190
G X, e.g. 180 Y, e.g. 100 You lose Yes (X greater than Y) You get the maximum -90 penalty (grades are assumed to be 40 points apart), i.e. 90
H X, e.g. 180 Y, e.g. 100 You Draw Yes (X greater than Y) You get your grade - 40, i.e. 140
I X, e.g. 100 Y, e.g. 180 You Draw Yes (Y greater than X) You get your grade +40, i.e. 140

The above cases are reflected in the BCF performance calculator on BCC on-line.

In summary if you win, you generally get the opponents grade + 50. If you lose you generally get your opponents grade - 50. If its a draw, you usually get the opponents grade.

The complications come when the difference in grades is greater than 50 points (one of the players is a relative bunny to the other player).


FIDE/ ELO/ USCF rating systems

The FIDE title system is founded on the Elo rating. This is the method of rating chess players used for all international tournaments. It has also been adopted by many national bodies. It originated from a scale previously used by the United States Chess Federation and based on the assumption that a rating of 2000 would be equivalent to scoring 50% in a US Open championship and that no player's rating would be negative. (The calculator on BCC on-line uses the rough translation of the Elo rating + 100 to arrive at the USCF rating).

The following table shows the rating implications for strength of player:-

FIDE Rating Type of player (generalisation!)
Above 2700 World champions
2500-2700 International grandmaster
2300-2500 International master
2000-2300 Average to very strong club player

A sample of 30 games is sufficient from a statistician's perspective to give greater confidence to the rating assigned to a player.


Internet chess ratings

Taken from ICC ratings help:

***** RATINGS *****

The Basics:

You can have 5 different ratings on the ICC. There are ratings for bullet, blitz, and standard, which are different speeds of regular chess. There are ratings for bughouse and wild, which are for chess variants. See "help definitions" for an explanation of what these mean. Type "finger" to see your current ratings.

To get a rating in one of these five categories, you only need to play a rated game. Do "set rated on" and then do "seek" to ask for a game. You can ask for any time control or chess variant you wish. Type "help seek" for more information. Unrated games will not count for your rating, but you are welcome to play unrated games.

Ratings usually range from 800 to 2800 on ICC, but there is theoretically no limit at either end. Beginners usually get ratings from 800 to 1200. People with ratings over 2200 are considered to be "masters". Grandmasters playing on ICC usually have ratings from 2400 to 2800. You can type "best" to see the highest-rated players on ICC. You can type "rank" to see where you stand among ICC members.

Ratings on ICC are similar to the USCF and FIDE, but are totally separate. Do "help survey" for a statistical comparison.

A player's rating is "provisional" if he/she has played less than 20 games. A rating is "established" if it is based on 20 or more games. A different formula is used to calculate ratings for established and provisional players. See "help provisional" for some information about provisional ratings.

Everything you wanted to know about rating formulas, but were afraid to ask:

The rating during the provisional period is the average of a set of values, one for each game played. The value for a game against an established player is the opponent's rating plus 400 for a win and minus 400 for a loss. For a game against another provisional player, the value is moved towards the previous average to lessen the impact of the unreliable result. Players with no rating are treated as having rating 1650 in this case. Extra points are then added to the rating for the purpose of keeping the average rating of all established active players close to 1650. In particular, 1/5th of 1650 minus the current average is added to the rating.

To explain the established period requires the use of a formula. Suppose your rating is r1, and the opponent's is r2. Let w be 1 if you win, .5 if you draw, and 0 if you lose. After a game, your new rating will be:

1
r1 + K (w - (-----------------))
(r2-r1)/400
1 + 10

I still need to explain the variable K. This is the largest change your rating can experience as a result of the game. The value K=32 is always used for established player versus established player. (The USCF has a system in which this K-factor diminishes for more highly rated players.) If you're playing a provisional player, the factor K is scaled by n/20, where n is the number of games your opponent has played. So, as in the provisional case, if you play an opponent who has never played, your rating can't change.

This formula has the property that if both players are established then the sum of the rating changes is zero. It turns out that if the rating difference is more than 719 points, then if the strong player wins, there is no change in either rating.

Note that during the provisional period, BEATING a player whose rating is more than 400 points below yours will DECREASE your rating. This is a consequence of the averaging process. It's useful too, because it prevents the technique
of getting an inflated provisional rating after one game, and then beating 19 weak players to get an established rating that is too high.


Rating Survey by Tmeister

Total responses so far: 670

People with ICC and USCF ratings: 592        ICC average rating =  1800
Less than 200 point difference:   397 (67%)  USCF average rating = 1718
People with ICC above USCF:       385         The average difference: 82
People with ICC below USCF:       203         The median difference:  72

People with ICC and FIDE ratings: 107       ICC average rating =  2295
Less than 200 point difference:   83 (77%)  FIDE average rating = 2233
People with ICC above FIDE:       71         The average difference: 62
People with ICC below FIDE:       36         The median difference:  68

Tmeister's Rating Survey!  Thanks to all who participated by sending data!  All numbers and text were compiled and written by Tmeister. Minor editing and a couple comments added by POTZY.  April 2, 1996.

Here is most of the useful information I found by studying the relationship between ICC ratings and national ratings. First of all, I could only compare ICC ratings to ratings from USCF, FIDE, WBCA, and USCF Quick ratings. There was so little data for the other national ratings that I could not conclude anything really meaningful. For most of the comparisons, I used only players whose ICC ratings had been active for at least one of the times I tested their rating. Another thing to note with that is that the ICC ratings used for each player was an average of their rating tested three times over a two month period, with no less than ten days between tests. this way I hoped to eliminate the fact that ICC ratings vary a lot. For those tests with not very many data points, I included those players who were inactive, but under no time did I use a player who was provisional during one of the times I tested the rating.

First, the comparison of ratings between USCF and ICC-BLITZ. The first item is ICC Rating, a grouping of all players with ICC ratings in that range. data points are the total number of active players in that group, and trimmed is the % of data values trimmed from the top and bottom, to eliminate others. the mean is the average value of the difference between each player's USCF rating and ICC rating (a positive value means the player's USCF rating is larger than the ICC rating).

Comparison of USCF to ICC Blitz Rating

ICC rating Players trim  mean   StDev
OVERALL      195   10%   98.6    124.7
2400+          6    0% -103.9     99.6
2200+         28   10%  -39.8     61.0
2200-2400     22   10%  -24.5    66.2
2000-2200     20   10%  112.0    85.8
1800-2000     37   10%   99.9   120.5
1600-1800     35   10%  209.7    81.3
1400-1600     36   10%  107.2   137.0
1200-1400     23   10%   60.6   160.3
1000-1200     16   10%  116.2   104.9

What this means is, if your ICC rating is between 1800 and 2000, you can expect to have a USCF rating of approximately 100 points higher than your ICC (Blitz) rating.  Looking at the data, overall, ICC blitz players are underrated (as compared to USCF rating) by almost 100 points. Those with very high ICC ratings are overrated, however, those with ratings below 2200 are on average underrated. It was very interesting to find the mean of 1600-1800 group.  Also interesting that this group has a much lower standard deviation than the nearby groups. One theory I can come up with is that there are many players of this strength playing on ICC and that competition is very fierce.

Comparison of USCF to ICC Standard.

The introduction to this part is the same as the introduction to the  previous part, comparison of USCF to ICC Blitz.

ICC Rating  Players trim   mean   StDev
OVERALL       95    10%  -82.4    124.1
2400+          6     0% -170.4   119.9
2200+         12    10% -168.8    105.3
2200-2400      6     0% -185.2    183.5
2000-2200     14    10%  -13.3   110.4

1800-2000     31    10%  -64.3   118.6 USCF=ICC (-64)

1600-1800     15    10%   50.1     88.3 ICC+50=USCF
1400-1600     13    10% -199.9   132.3
1300-1400      3     0%  -65.2    150.6

What this means is, is your ICC Standard rating is between 1800 and 2000, you would expect to have an USCF rating around 64 points below your ICC rating. An interesting thing to note is that for the group 1600-1800, their expected USCF rating is above their ICC rating. (and like in the last comparison, this is the group with the smallest standard deviation.)  This group, like the subsequent ones, have a rather small amount of usable data points, which means the data is not
very reliable.

Comparison of FIDE to ICC Blitz

ICC rating  Players trim  mean    StDev
OVERALL       38    10%   30.5     124.1
2400+          9    10% -180.3      66.7
2200-2400     13    10%    6.2     110.9
2000-2200     10    10%  109.6      74.8
1900-2000      5     0%  251.2      74.1

Very interesting to note that for those with ICC rating of over 2400 they can expect a FIDE rating of almost 200 below the ICC rating. and those with ICC rating of under 2200 can expect a FIDE rating of more than 100 points above. This amounts to a rather meaningless comparison of the two ratings. It seems that you have to be skilled at playing on ICC in order to have a high ICC rating, and a good FIDE rating will not matter as much.

POTZY:  No FIDE ratings under 2000 exist.  Only 35% of FIDE ratings are in the range 2000-2199.  Therefore it would be expected that the people with ICC ratings 1900-2200 shown above would have much higher FIDE ratings.  This is indeed true in the chart above.  The data in the chart don't really say much, except that almost all FIDE rated players who
responded to the survey have FIDE ratings in the range 2200-2400.

Comparison of WBCA to ICC Blitz

ICC rating Players trim  mean    StDev
OVERALL      32    10%   52.2     120.3
2200+        12    10%   42.4      86.4
1800-2200    12    10%   45.8     106.4
1100-1800     8     0%   98.0     226.5

POTZY: WBCA ratings are roughly 50 points higher than ICC Blitz ratings.

USCF Quick to ICC Blitz

ICC rating Players trim    mean    StDev
OVERALL       22    10%   114.8     140.9
1800-2500     10    10%    63.8      61.8
1100-1800     12    10%   163.6     182.2

All players in these groups should expect USCF Quick Rating higher than their ICC blitz rating.

#of respondents for each ratings type:

USCF 248
FIDE  43
WBCA  35
USCFQ 22
CFC   16
Australia 8
Iceland   7
Sweden    6
Germany   6
Quebec    4
Netherl.  3
Finland   2
Israel    2
IECG (internet email chess group), Wales, Norway, greece, Catalan, Spain,
Switzerland, Russia, Denmark: 1

All statistics and comments courtesy of Tmeister.  Please ask permission from Tmeister before reproducing any of the rating study.  Thanks

This page belongs to:
Barnet Chess Club
Home