negative points in handicap games

[hyperpape]

I can no longer remember whether this pro said it to me or to a friend, but after that kyu player had put down 9 stones in a simul, he said "Not so many. If you want to win, put down 9. If you want to learn, put down 5".

By the same logic, playing someone a little better than you without handicap can be excellent practice.

[brodie]

otenki, if you ever want to just practice some fuseki in free games, just to kind of mess around and discuss, we should play on kgs sometime. (of course i would also be fine with playing more serious ranked games with discussion, or free even games, etc). I used to have that book, but did not make enough use of it, and would love to practice some more fuseki. my rank right on kgs fluctuates between 9 and 11k on kgs, but my formal training (joseki, fuseki) is pretty rusty. pm me if you want some practice! it would motivate me to spend more time practicing myself.

[RobertJasiek]

The proper value, up to 9 stones, is around 14*H - 7, where H is the rank difference. But the point value of handicap stones decreases as the number of stones increases.

The value does not depend linearly on the number of handicap stones. Up to a certain (still unknown minimal) number of handicap stones sufficient to kill any opposing stone, there is some effect of increasing value of extra handicap stones because they can cooperate better with each other. As I have tested in dozens of games against roughly equally strong 5 dan players, a Japanese style 9 stones handicap equals about 130 points komi under territory scoring. That is 14.4 points per stone, i.e. more than your formula predicts.

For even games without handicap, currently there are three evidences that probably 7 is the correct komi:
- professional game statistics (however, they should also play thousands of 8 or 9 komi games to be sure)
- the conceptual (but still unproven) idea that komi must be half the first move's value
- my corner killing and influence idea (see Joseki 2 Strategy, chapter 4.4.1), which assesses the average local move value of an early corner move to be slightly above 14 in every test example pattern