Belgian Go Federation rating system


Rating principles / Summary

A rating is an evaluation of a playing strength calculated according to Go games results.
The traditional ranking system that every Go players around the world use is made up with "Kyu" and "Dan" ranks.
A difference in rank of 1 Kyu or 1 Dan represents (more or less) a difference of 1 extra stone of handicap on a 19x19 Go board.

Complete beginners are somewhere between 40 and 30 Kyu.
Beginners are between 30 and 20 Kyu.
A majority of Go club players are between 20 Kyu and 10 Kyu.
Strong players are between 9 Kyu and 1 Kyu.
Best players are classified in "dan" levels (between 1 Dan and and 7 Dan (strongest).
(just for the record: the best among the best players are professionals, who are ranked according to a specific "pro dan" scale, from 1 dan pro to 9 dan pro).
For calculating purpose, the Belgian Go Federation ratings system use "points" representing their equivalent Kyu/Dan values (100 points per rank so it is somehow a ranking with a precision of 2 decimals).
The Belgian ratings follows to a large extent the European ratings (European Go Database), but aims to let ratings, especially of beginners, evolve more quickly (e.g., based on club games, smaller board games, using larger weights for games, ...)

Ratings formula - general principle

The Belgian Go Federation rating system is fundamentally an ELO rating system, in the sense that for every won game, the player who wins gains points and goes up in the rating list and the player who loses goes down in the rating list.

Of course the stronger the opponent is, more points there is to gain in case of win and less to lose in case of loss. And vice-versa, the weaker the opponent is, less points there is to gain in case of win and more to lose in case of loss.

Aside from this main feature, the system include a factor which purpose is mostly to stabilize the fluctuation of the ratings as we go higher in the scale. There is also free points given for every game played – whatever there is a win or loss. The value of the bonus is very small at high levels and increase in value as we go down in the rating list. This allows to push every player up adequately to represent improvement.

Points can be translated into the Dan/Kyu ranks (divide the number of points by 100, a positive value is dan, a negative value is kyu).

The Belgian rating system has evolved over time. E.g., before May 2001, a different system was in use not based on winning probabilities. The current system doesn't keep any history from before that time. Since June 2001, a new formula is used, based on the EGF ratings model at that time. Since then, several modifications have been made, e.g., to avoid punishing white in handicap games against improving black players (in 2004), to allow games on smaller board sizes, etc. We may not have followed all EGF modifications made over that time.

Link between rating and ranking

Ratings between 100(d-1) and 100d (d>=1) correspond to a rank of d dan.
E.g. an average 1-dan player has a rating of 50, an average 3-dan player has a rating of 250.
Ratings between -100k and -100(k-1) (k>=1) correspond to a rank of k kyu.
E.g. an average 10-kyu players has a rating of -950.

As soon as a player obtains a rating above the maximum rating for his grade, he is promoted. As soon as a player drops 100 below the minimum rating for his grade, he is demoted. E.g. a player increasing from 95 to 101 points is promoted to 2 dan. A 5 kyu player dropping from -490 to -510 remains 5 kyu. Only if his rating decreases further to -601 or less, he becomes 6 kyu.

1. Ratings are always computed for a complete event. A 1 dan player with rating 99 winning first 2 games and losing then 3 games in a tournament, ending up at 90, remains 1 dan even if he would have a 2 dan rating after the first day of the tournament.

2. Corrections to grades are possible if new information is added. E.g. suppose a player is -510. He plays a tournament on May 1st and on May 10th. Suppose the organizer of the first tournament reports the results late. As a result of the tournament of May 10th, the player increases to -490, becoming 5 kyu. If later the information of the first tournament is added, both tournaments are recalculated. If this player appears to have lost 50 rating points at May 1st, and ends up at -525 after May 10th, he remains 6kyu.

I think my rating is inaccurate. What can I do ?

  1. Ratings remain an estimation of a player strength according to results. In order to be able to update a player rating, official game results are needed. Therefore, to get an accurate rating official games have to be played, either tournament games or club games whose results are recorded and sent to the ratings.
  2. Players who have improved (by playing on line f.i.) and didn't play official games for some time can have a strength significantly different from the one of the rating list.
  3. If your rating strongly differs from your actual strength some manual update might be needed (see below).

What is a manual update and how should it be requested ?

  • Your club president can estimate your rating and send a request to update your rating to the ratings commission if clearly you have become much stronger than your official rating. You can also ask your club president if you think you are in this situation.
  • Up to 5 Kyu your request should be more or less "automatically" granted.
    For players stronger than 5 kyu, the Rating Commission will decide cautiously on the case if there is strong evidence of the claimed strength.
  • The rating commission may also take the initiative to manually update the rating of a player if results suggest that this is appropriate
  • The rating commission has a number of rules of thumb for deciding on manual changes, but reserves the right to judge in the light of all evidence available, possibly deviating from its habits.

Can I register with a different rank for a tournament than my official rating ?

In principle, you should always register for a tournament with your official rank.

For Kyu players, when registering for a Belgian tournament you can provide a different rating than your official, but only a stronger rating and only if the tournament organizer agrees. This would allow you to be paired according to the rating estimation you provide. Whatever, we suggest you strongly to be cautious with your level declaration - and to provide a higher level estimation only if you are confident that you have at least the claimed strength.

  • If you lose all your games, there will be no evidence that you are stronger than your official rating. It is therefore better to be sufficiently conservative in your rank claim.
  • If your official rank is between -600 and -1, the difference between your claimed rank and your official rank may be at most 1 stone (but do not register as a dan player). Between -600 and -1300, the difference may be 2 stones.

annex 1: parameters influencing the number of points at stake in a rated game

The number of points at stake in a rated game depends on various criteria.

As we explained, first of all, on your strength: the stronger you are, the smaller of points you can get or lose by winning / losing your game. But beside this, the characteristics of the game impacts also the number of points at stake. We give a "weight" in function of various criteria:

1. The size of the goban :

  • For a 19*19 game, the weight is (of course) 1.
  • For a 13*13 game, the weight is 0,5
  • For a 9*9 game, the weight is 0.33
  • For other board sizes, the weight is a linear interpolation between the closest board sizes given above.

2. A game played - or not - in a tournament

  • For a tournament game, the weight is 1,2.
  • For a game played (more informally) in a club, the weight is 0,8.

3. Handicap stones.

For games played with a handicap, the weight is:

  • for the player playing with black: 1 - (#h*0,1)
  • for the player playing with white: (1 - (#h*0,1))^2 where x^2 = x*x (since 2004)

4. Game played on-line or face-to-face

The weight is 0,8 for on-line games (including correspondence games) (1 for face-to-face games of course)

5. Thinking time allowance.

The weight is 1 for 75 minutes sudden death, or 60 minutes + 5 minutes for at most 20 stones, or slower.

The faster the game, the less it counts proportionally. F.i., a 45-minute sudden death counts for 45/75 = 60% Minimum time allowance is 40 minutes sudden death or 30 minutes with at most 20 moves in 5 min. In other words, (too) quick games are not rated.

These criteria are of course cumulative.

Example: the weight for the back player playing a 19*19 tournament on-line game with 3 stones handicap with 60 minutes sudden death thinking time is: 1,0 (19*19) * 1,2 (tournament) * 0,75 (3 stones handicap)* 0,8 (on line)* 0,8 (60 min) = 0,5

annex 2: rating formula in detail

• StrengthDiff = Strength - OpponentStrength • AdjustedStrengthDiff = StrengthDiff + (100 * Handicap) • ExpectedResult = 1 / (exp(-AdjustedStrengthDiff / Afactor) + 1) • HandicapFactor = 1 - abs(Handicap / 10) • PlaceFactor = if Tournament then 1.2 else 0.8 • AchievedResult = if win then 1.0 else 0.0 • ChangeInRating = ConFactor * (AchievedResult - ExpectedResult) * HandicapFactor * PlaceFactor


The difference in strength is adjusted by increasing the strength of the person receiving handicap. The handicap is the number of stones given, reduced by 0.5 since no komi is given (identical to EurpRatings).
The Afactor (table I) is a measure of the probability that a player can win from a stronger player. This factor is larger for kyu players than for dan players. This factor is the same as 'a' in the EuroRatings and was not present in the old formula.
The expected result is a value between 0.0 (loss) and 1.0 (win). It can be calculated from the adjusted strength difference and the afactor. Playing against an even opponent gives probability 0.5.
Handicap games do not count as much as even games. This is expressed by the handicapfactor. The handicapfactor for black evolves linearly between 1.0 (no handicap) until 0.1 (9 stones). For white, the handicapfactor is either the same (before May 2004) or the square of blacks handicapfactor (since May 2004) to ensure white doesn't suffer disproportionally from teaching games.
Normal games are played on a board size of 19x19. Games on different (smaller) board sizes have a weight evolving linearly from counting fully for beginners (20 kyu) to not counting for 1 kyu (and don't count higher). Since 2021, 13x13 games have a weight 0.5 and 9x9 games have a weight 0.3.
Tournament games count more than club games. This is expressed by the PlaceFactor. Tournament games are assumed to satisfy "category A" conditions of the EGF, i.e. minimum 1h plus byoyomi, or 75 minuts absolute thinking time.
Tournament games with lower time limits get a discount factor.
The ConFactor (table I) expresses the value of winning/losing a game, or the speed the rating changes. This is larger for weaker players and smaller the stronger players. (identical to EuroRatings)
The change in rating is computed by multiplying the mentioned factor.

Table I: The dependence of parameters ConFactor and AFactor on the rating. We use linear interpolation between the points given in the table. For ratings below -1950 and above 650, linear extrapolation is used.

Rating ConFactor Afactor
-1950 116 200
-1850 110 195
-1750 105 190
-1650 100 185
-1550 95 180
-1450 90 175
-1350 85 170
-1250 80 165
-1150 75 160
-1050 70 155
-950 65 150
-850 60 145
-750 55 140
-650 51 135
-550 47 130
-450 43 125
-350 39 120
-250 35 115
-150 31 110
-50 27 105
50 24 100
150 21 95
250 18 90
350 15 85
450 13 80
550 11 75
650 10 70

Table II: How many points a player (of level 20 kyu,..., 6 dan) gains by beating an opponent that is two stones weaker (-200.0), ... , Equally strong (0.0), ... , two stones stronger (+200.0) in an even tournament game

20 kyu37.444.752.660.969.678.386.694.5101.8
15 kyu26.
10 kyu16.321.026.532.639.045.451.557.061.7
5 kyu8.711.916.020.725.830.935.639.742.9
1 dan3.45.37.710.914.417.921.123.525.4
6 dan0.

Suppose a player A is 20kyu(-1950).
He wins against an opponent B which is -1950 too. The expected result is 0.5 since both players have the same strength. The ConFactor is 116.
If this game is a club game, he increases 116*0.5*0.8=46.4 points to -1903.6. His opponent drops to 1996.4.
Player A is now about one grade stronger than B. (92.8 rating points) So the next day, B gets 1 handicap from A (i.e. black without komi). This has a value of 5 points or 50 rating points. So the adjusted rating difference is 92.8-50=42.8. The AFactor is about 200. The expected result is 1/(1+exp(-42.8/Afactor))=0.45 for B and 0.55 for A. A wins again. A's Confactor is about 113, B's Confactor about 119. So A increases 113*(1-0.55)*0.8*0.95=38.6 points to -1864.9. He an 19 kyu now.
A is increasing rapidly since he is talented and reports his games promptly for processing in the rating list. A few weeks later he is -1450.
He now plays a 4 stone game against a 8 kyu (-750) player C. Handicap games do not count so much. A 1 stone game gives a factor 0.95, a 4 stone game gives a factor 0.65. His Confactor is 90, strength difference 700, adjusted strength difference 350. He has a chance of 0.12 of winning, but he wins this game unexpectedly, which earns him 90*0.8* (1-0.12)*0.65=41.2 points. Player C only has a ConFactor of 55. So he loses only 55*0.8*(0.88-0)=25.2 points. A few years later, player A has reached the level of 1 dan (50 ratingpoints).
Winning a game against an opponent of the same strength in a tournament has a value of 24*0.5*1.2=14.4 points now.


The above explanation is somewhat simplified and may at some point not correspond fully with the reality of the rating system. The "official" formula is the code as it has been implemented with its many additional rules to cover special cases and exceptions, rather than the high-level explanations of it.