A 100% sample will generally be OK for single table cardinality, but it may not mean the join cardinality is accurate. Look at the plan line #2 in the first example. The HASH JOIN version has a cardinality of 8775, the HASH JOIN RIGHT SEMI version has a cardinality of 19. That's a difference of 462x. Pretty big! How many rows are really returned? In the second example the access order and path is identical, the only difference being lines 6 & 9 where one is SORT UNIQUE and the other is HASH UNIQUE. I'm wondering how much difference there really is between these two. I would think very little, but what is it? On Tue, Nov 17, 2009 at 2:28 PM, Neil Kodner <nkodner@xxxxxxxxx> wrote: > I'm using > estimate_percent=>100 and method_opt=>'FOR ALL COLUMNS SIZE 1' for these > tables -- Regards, Greg Rahn http://structureddata.org -- //www.freelists.org/webpage/oracle-l