AGA Ratings submission format

msgreg · #1

I have a question about rating file format, specifically komi.

I've been thinking about rating games at our club meetings to submit to AGA. The following information is on the AGA web site.

http://www.usgo.org/qualifications-rated-games

Quote:

Each game record comprises White's ID, followed by Black's ID, who won (B for Black, W for White), the number of handicap stones, and the komi (recorded as integer equal to the number of prisoners Black would give White).

Why is Komi an integer when AGA rules say komi should be 7.5 for 19x19 games? What is the proper value to put in the ratings file for submission?

MS_Sydney · #2

That is indeed odd.

xed_over · #3

When counting the score, I find it easier to give white 8 stones (for a 7.5 komi) as prisoners. It just makes counting easier. No math.

Since I don't have a half a stone, what's the difference if I say 7.5 or 8, really?

hyperpape · #4

The problem is that you can also express it as 7 points, White wins ties.

yoyoma · #5

Code:

20194   17638   B   0   7
18270   17636   W   0   0
99999   3979   B   2   0

Black won an even game with 7.5 komi
White won a game with 0.5 komi
Black won a 2 stone handicap game, 0.5 komi

xed_over · #6

did I round the wrong direction? I replied too quickly maybe and didn't read it carefully

msgreg · #7

Well, that's exactly why I'm asking. Other Internet quotes on scoring indicate a variety.

Quote:

Under the Ing rules, komi is 8 points and black wins ties.

Quote:

Oct 22, 2012 – (Black gives a 5 point komi with White winning in case of jigo tie).

So it seems it could go in either direction. Does it matter whether you specify Ing rules in the submission? (I.e. when you specify Ing, does the submission assume black wins ties, and if you specify AGA it assumes white wins ties?)

This seems bizarrely ambiguous for go. Why not allow for non-integer Komi, and/or an indicator for who wins ties? Do those things affect the rating or rank? Or specify that the format applies to integer komi and X wins ties (whether it's black or white doesn't matter if the format itself always assumes a default value).

pwaldron · #8

The current ratings format is a historical thing from more than twenty (thirty?) years ago when Japanese rules dominated. The ratings format's komi field indicates the (integer) number of prisoners that black gives to white at the beginning of the game. An extra-half point for white is automatically assumed so that white wins ties.

If you're playing with a rule set where black wins ties then it's the same as reducing the integer komi by one and having white win ties.

msgreg · #9

Quote:

An extra-half point for white is automatically assumed so that white wins ties.

Perfect. I've emailed aga webmaster to suggest the correction.

One other ambiguity is how to treat 1 handicap "stone" vs 0.5 komi. Is there a difference between 1 handicap 0 komi vs 0 handicap 0 komi. The ambiguity is that 1 handicap stone is treated as 0.5 komi. So is there any different treatment between the to cases?

pwaldron · **#10**

msgreg wrote:

Quote:

An extra-half point for white is automatically assumed so that white wins ties.

Perfect. I've emailed aga webmaster to suggest the correction.

One other ambiguity is how to treat 1 handicap "stone" vs 0.5 komi. Is there a difference between 1 handicap 0 komi vs 0 handicap 0 komi. The ambiguity is that 1 handicap stone is treated as 0.5 komi. So is there any different treatment between the to cases?

There's no difference between 1 stone and 0 stones.

vash3g · **#11**

bearer of bad news, there is currently no aga webmaster for the updates. if the ratings officer wants to change something on the site he can email the account and the next person who volunteers can help make the updates for those pages. just a friendly heads up.

AGA Ratings submission format

Who is online