One Group Wins
It is a controversial statement to say "The side having only one group wins".
A single large group can be efficient. Four groups have to surround eight eyes. One group surrounds two eyes.
A related statement is "Big dragons never die". A Dragon tends to be a very narrow group however with not much territory (if any more than two eye points) because it has been chased across the board.
Some fuseki experiments like the Great Wall can lead to single groups. Playing in the style of Takemiya can create big moyos or even Cosmic Go.
In old-style Chinese scoring, there is a group tax of two points assessed for each group. However this statement is usually debated without reference to the rule set.
Charles Matthews A contribution towards this debate. In a seminar I attended, Matthew Macfadyen pointed out this dilemma: if you see two groups of your opponent's that can be cut apart, or forced to connect up, what do you do?
The answer:
- in the middle game it should be worth more to separate them into weak groups.
How much more is the topic QARTS is supposed to tackle. Anyway, many points even if nothing is killed.
On the other hand,
- in the endgame you probably can't kill any weak group, and attacks are limited in value
so you then prefer to steal the eye space from one group in sente, forcing the opponent to play a gote connection (which is probably just a neutral point).
The cross-over between the two attitudes could be a sensitive indicator in oyose positions.
Bill: I don't think it's the difference between the phases of the game that matters as much as the question of the weakness of both groups. If both groups are weak then the connection is not going to be a dame. Then splitting will often be better, endgame or no. And if one group is strong, splitting may not gain much, even in the early middle game.
See also Whole Board Connection Theory for an example where one group loses.