type of join in sql

[BigP]

can some body shed some light on how does the optimizer decides to choose the kind of join i.e. nested loop, sort merge or hash join . In one of queries if i tweak the join condition it changes the type of join and start using index , otherwise it doesn't . I thought it depends on statistics .. if optimize finds that are more qualifying records in inner table then it will prefer to go for sort merge and will do full scan of inner table , but if it thinks there are less records in inner table it will user nested loop . am I correct ?   TIA , bp