Joseki Heuristics
General heuristics
Joseki heuristics are guidelines for playing corner patterns. They serve to determine if a corner pattern results in an equal result, or favorable to either of the players.
Well established corner patterns (joseki) are, or at least have been, considered to be equal. Deviating from them results in either unsettled, slightly favourable or clearly biased results.
Here are the questions to ask oneself when evaluating a corner pattern, in progress or played out:
- Who gets the corner territory?
- Who gets influence to side 1 (or sticks his head out to ...)
- Who gets influence to side 2 (idem)
- Who gets influence to the centre (idem)
- Are there any defects (aji) or lack thereof (thickness) in either position?
- Last but not least, who gets the initiative?
Based on these observations, any player should be able to determine for oneself if the sequence is locally playable, equal, favourable, worth trying, whether the advantage is small or huge, if the initiative compensates the net result ...
Making these observations and doing the positional judgment, is far more important than "remembering" the joseki. Regularly checking your positional judgment and your awareness of available tactics in joseki references is of course recommended.
Advanced players move away from the local evaluation, to apply whole board thinking.
More (specific) general heuristics
- Many joseki sequences end up with one side getting territory and the other influence or power. So if you get both territory and power, or if your opponent does, then probably somebody made a joseki mistake.
- Get life, Get power (influence, strength), Get territory.
- Make a move to settle your shape as a first guess.
- Run out to live as a second guess.
- Attack, as a third guess.
- Don't get sealed in, if you can help it.
- When it is:
- 1-on-1: you may tenuki.
- 2-on-your-1: don't tenuki.
- 3-on-1: may tenuki, really try to.
- 4-on-1: always tenuki.
- 2-on-your-1: don't tenuki.
- If a move should be punished, punish it right away.
- Make good shape in the right direction.
Comments
larsen Hey, I expect that heuristics should be rather something like "You approach a hoshi stone, opponent extends. Now if you played inside opponent's formation (i.e. an approach move was not supported by your corner), crawl into the corner with keima, then extend. Otherwise you may play more loosely (star point on the side)".
Do You mean to create a collection of such commonsense guidelines, or just to keep it general?
Rich: Nice idea, a couple of comments:
1) It seems attack is more active than running to the centre, which usually occurs because one is being attacked. Shouldn't attack come second, running out third?
2) I would consider punishing a move right away irrespective to be a bad habit, IMO; it could be (and in my case, usually is) aji keshi. Of course, it depends on just how wrong the move is; for a small mistake, you might gain more from using the aji or benefitting from the extra move your opponent must play.
3) Point nine seems superfluous: surely you always want to make good shape in the right direction? It seems too much like giving "always play good moves" as a heuristic for winning games.
RobertJasiek: So far the page creates a false impression of a need for just a few heuristics. As can be seen in my books 1, 2, 3, there are hundreds of heuristics. It is useful to consider a hierarchy of heuristics so that one can ignore most of them most of the time. A top level candidate is "Do global strategic planning!". An example of a practical intermediate level heuristic is: "Carefully apply only those strategic concepts that currently are obviously relevant! For the relevant strategic concepts, apply their specific heuristics!". E.g., if the strategic concept "development direction" is currently relevant, one comes to considering application of specific heuristics such as "Consider the relation to the directions in the positional context!", and then yet more specific "Consider if great black influence towards the upper side fits the adjacent corner there!".