Flattened handicaps

[hyperpape]

It is a definition that that's what the handicap means, but it's not a definition that the only way to achieve that meaning is to play properly handicapped games. So I ask again: are you sure about that? Have you proven it?

[Mef]

It is a definition that that's what the handicap means, but it's not a definition that the only way to achieve that meaning is to play properly handicapped games. So I ask again: are you sure about that? Have you proven it?

It's worth interjecting, at the very least AGA, EGF, KGS, and the old IGS system* attempt to do this if their hand is forced (and that is the data they are fed). Looking at how much the parameters vary between each system (expected winning percentage from a 1 stone difference might be 70-85%, two stone difference anywhere from 85-97%) suggests how much of a challenge in might be to accurately do this well over a large group of players (admittedly, it may equally point to how hard it is to accurately rank go players in the first place!).


*The new IGS system may try to do this, but I haven't attempted to back-calculate how their parameters are derived. At first glance to me they appear to be more back-of-the-envelope guesses to make a simple, more or less linear system that is not worried about winning percentages.

[Laman]

Proven what... isn't that just true? ("that"=In an accurate rating system, a properly handicapped game gives a 50% chance to each player, and this is all that the ranks mean.)

this is not entirely obvious. for example, if you have three players, from strongest to weakest, A, B, C, and A has 50% chance of beating B with x handicaps and B has 50% chance of beating C with y handicaps, proper handicap system requires A to have 50% chance of beating C with x+y handicaps, which is not a property of the rating system, but of the handicaps themselves. and it is not clear whether this is true.

and as rating systems are not primarily and only build around handicaps, it is reasonable to suppose that handicaps fits them, but you can't take it as granted

[skydyr]

Proven what... isn't that just true? ("that"=In an accurate rating system, a properly handicapped game gives a 50% chance to each player, and this is all that the ranks mean.)

this is not entirely obvious. for example, if you have three players, from strongest to weakest, A, B, C, and A has 50% chance of beating B with x handicaps and B has 50% chance of beating C with y handicaps, proper handicap system requires A to have 50% chance of beating C with x+y handicaps, which is not a property of the rating system, but of the handicaps themselves. and it is not clear whether this is true.

and as rating systems are not primarily and only build around handicaps, it is reasonable to suppose that handicaps fits them, but you can't take it as granted

The other thing with this is that there is a presumption that all handicaps scale in difficulty in a linear fashion, which is not necessarily true. It may be based on my style of play, but I often do much better as black in handicap games where there is a stone on tengen (5 for example) as opposed to one without (6). I think it's also pretty commonly accepted that going from 3 to 4 stones is a big jump, and that no komi is not as big as half of a two stone handicap.

[hyperpape]

For those reasons, it was even too much of me to say that it's definitionally true that a handicap means 50-50% winning with one rank difference. At best it's definitionally true that that's one of the goals of a handicap system.