A redirect of standard output into arbitrary file name always works. You are doing it correctly as man wget says, using -O
wget http://www.kernel.org/pub/linux/kernel/README -O foo
--2013-01-13 18:59:44-- http://www.kernel.org/pub/linux/kernel/README
Resolving www.kernel.org... 149.20.4.69, 149.20.20.133
Connecting to www.kernel.org|149.20.4.69|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 12056 (12K) [text/plain]
Saving to: `foo'
100%[======================================================================================================================================>] 12,056 --.-K/s in 0.003s
2013-01-13 18:59:45 (4.39 MB/s) - `foo' saved [12056/12056]
Indeed, you must be getting an HTML in your file (usually can be checked with man file).
[EDIT]
In your case client is receiving 302 Found (you can check it with curl -v URL).
The following curl does the trick by respecting the 3xx:
$ curl -L http://sourceforge.net/projects/sofastatistics/files/latest/download?source=files -o foo.deb
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
0 463 0 0 0 0 0 0 --:--:-- 0:00:01 --:--:-- 0
0 0 0 0 0 0 0 0 --:--:-- 0:00:02 --:--:-- 0
100 2035k 100 2035k 0 0 390k 0 0:00:05 0:00:05 --:--:-- 1541k
$ file foo.deb
foo.deb: gzip compressed data, was "sofastats-1.3.1.tar", last modified: Thu Jan 10 00:30:44 2013, max compression
There should be similar option for wget to tolerate HTTP redirects.