Palle Girgensohn <girgen(at)pingpong(dot)net> writes:
> How do you mean they look far off?
> Seq Scan on group_data (cost=0.00..47544.43 rows=114164 width=43) (actual
> time=114.015..1334.479 rows=4 loops=1)
114164 estimated vs 4 actual rows is pretty far off. Perhaps something
skewed about the data distribution?
> If you want, I can send you the data.
If it's not too huge, sure, send it to me off-list.
regards, tom lane