Quick Questions

non-linearity of ranks in the AGA ?? [#236]

Back to forum

New reply

velobici: non-linearity of ranks in the AGA ?? (2006-01-03 16:38) [#805]

Went to a go oriented new year's party on 1 January 2006. there we saw an AGA 7 dan beat an AGA 5 dan while giving a four stone handicap. this AGA 5 dan is a solid 5 dan having been at this rank for over 18 months and participating in many tournaments. the AGA 7d is also well known and active.

so it seems that above 5 dan, a dan rank is not necessarily equivalent to a single stone in strength.

one caveat: this 5 dan is VERY GOOD playing against a handicap. he has much less experience, receiving handicap stones.

thoughts ?

X Re: non-linearity of ranks in the AGA ?? (2006-01-03 18:16) [#807]

Bob McGuigan: First, linearity of rank, if it exists at all, is a global phenomenon. Two particular individuals might well have playing terms different from what would be expected from their ranks. Second, nothing can be inferred from the results of one game. If I were the 5d I'd be likely to feel a lot of pressure and play too conservatively, or be over confident and make a lot of overplays. One proverb regarding handicap go is that each inappropriate loss of sente or wasted move equals a loss of one stone worth of handicap. It's not hard to believe that at a party, where some drinking may have been going on, four mistakes of that sort might be made.

reply Not alot of data (2006-01-03 17:15) [#806]

Not sure how much you can conclude from one game. How serious was it? You say it was at a party - was alcholol involved?

velobici: Party conditions (2006-01-03 20:27) [#808]

Not a lot of drinking, a beer or two. People were playing rather seriously...not tournament serious, but certainly not messing around. As far as I know this is the second time that this has occurred. (same two players, same handicap.)

MrShin: Rating or ranking? (2006-01-04 00:35) [#809]

Last I checked, AGA stops it's ranking at 7 dan. Do you mean 7 dan based on AGA rating (which can go above +7) or AGA rank (which although doesn't exist, you can't enter a tourney higher than 7 dan)? If you mean rating, maybe the 7 dan hasn't played a rated game in quite awhile. Just a thought.

velobici: Re: Rating or ranking? (2006-01-06 00:18) [#824]

I mean rating rather than rank. My bad for using the wrong word in the title.

velobici: Open tournament section games...no handicaps (2006-01-06 00:21) [#825]

It occurs to me that these players, and all other players of their rating, play in the Open section of tournaments....the section without handicaps. The lack of handicaps may slow the spread of ratings. Surely, one's rating changes faster when a 7.x beats a 5.y while giving 4 stones as opposed to a 7.x beating a 5.y when playing on even.

reply It occurs to me.. (2006-01-06 01:13) [#826]

It occurs to me that one game proves nothing. Comparing games played online, or in clubs, let alone at parties to serious rated games is problematic - drawing conclusions based on one such game is absurd.

ChrisHayashida: ((no subject)) (2006-02-24 07:48) [#1104]

Darn, and I thought that this meant that it was statistically impossible for me to lose in an even game against someone rated lower than me... :)

DrStraw: Anything can happen in one game (2006-02-24 12:08) [#1106]

I am an AGA 5d (and yes, it is solid if you count 17 years as solid) and I once played an AGA 2d a 7 stone handicap games on KGS (based on his new ? rank). I won! I later lost to the same person on 3 stones. All this proves is that anything can happen in one game.

I cannot remember the probabilities in the AGA algorithm but I suspect a one rank difference means that W has around a 67% chance of winning. If results are normally distributed (perhaps not an unreasonable assumption) then at a two stone handicap (the effective handicap in your situation) B would expect to win maybe 90% of the time (I'd have to look up the actual value).

aceofspades: Normal distribution of results (2006-02-25 08:11) [#1109]

DrStraw, I know this is somewhat offtopic but according to [ext] http://en.wikipedia.org/wiki/ELO_rating_system, chess players' performances do not fit well with normal distributions, particularly in lower ability ranges. On the other hand, [ext] http://en.wikipedia.org/wiki/Go_ranks_and_ratings#Go_rating_with_ELO says most systems assume normally distributed results.

Back to forum

New reply

[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library