Yeah these guys are pretty much on the mark. Another thing to try is to do things in stages, as some stuff will be at the best price-performance point at different times than others. For example, you may only want to drop $200 on a cpu, which should get you about 3-4 iterations down from the "top of the line". I did this in nov. with a p4 2.8. A 2.6 at the time was $180, but a 3.0 was at least $275. On the other hand a 2.4 was literally $179, so 2.6 was a no-brainer.
Video is usually the toughest to do this way, because everything changes so quickly (at also very slowly). For the last several years, pretty much every game out there used very similar engines and algorithms, so it was just a matter of doing the same thing faster, but now a new generation of games is emerging and nobody knows what will work well. So a year ago one could decide on a dollar amount and do just fine. Right now, though, you could pay $500 for a 9800XT and still not be able to do all the features on up and coming games.
For some perspective I started with a p3 500 in 1999 with ATA66, 128mb, and a TNT2. I added another 128mb, a 40GB HD, a 32x CDR, 5.1 sound, and a G3 ti200. I got the ti200 a few months after the g4 series was released (Summer 02?) for about $150 (it was as low as $130 at the time). Last november (2003), I did a big upgrade to P4 2.8, 1gb Ram, new mobo, sata HD, case, CDR---Everything except video, particularly since everyone was still saying HL2 was going to change EVERYTHING. I'm still using my G3ti200 with limited success- the rest of the computer makes up for the shortcomings - and when I do get a card I can throw a little more at it.
Hope that helps.