Race Picks
Chris Chang's page has been slimmed down, but it covers the best and worst races for non-advanced games fairly well, so I will merely cover other stuff. More is coming, but not for a few months.
Pre-warp and average tech games against the computer can be won most reliably by a combination of democracy, lithovore, and artifact homeworld (with low-g, ground combat, and spying as canonical negatives). There is no other race which allows a sustainable 60 research points from the first turn, and no other race can research Guardian-killing technology with as little effort as this one requires. While production is a bottleneck on development, the best strategy with this race involves no colonization whatsoever: research is only interrupted by construction of essential buildings (including biospheres). The only fleet this race needs to build is the minimal Guardian fleet, which is less than 1400 total production. Any true combat can then be handled by Loknar's ship. While this race has blitz attack potential very early on, a homeworld assault before getting AF mass drivers is neither cost-effective nor likely to work well.
A good non-telepathic blitz race is a combination of lithovore, omniscient, rich artifact homeworld, and +20 ship attack (with canonical negatives as above - unless you plan on eating planets with transports). It is very good at initial research and has good production, but a fast attack must be made, or it will be outdistanced by growth races. However, this race is very effective at exterminating lax human opponents because of its intimate knowledge of enemy fleets. Lasers and mass drivers are weapons of choice, and they can be implemented fairly quickly. This race seems to have trouble attacking the above race and the sublith race on Chris' page, simply as a consequence of substandard research.
One issue not often covered in detail is the advanced tech game, and this is because it is more difficult to treat as a result of the randomness of the initial conditions. This seems to be one of the few situations in which the creative race attribute is really worth the 8 picks. Without creative, your initial research choices are fairly arbitrary, and without automated factory, resesarch lab or robo-miners, you're screwed without good spying or technology trading. There is one exception to this rule, and it lies in a telepathic blitz. In a small universe game against the computer, you can often send your initial ships out to combat immediately, and conquer everyone in a short amount of time. If you are looking for a more reliable kill, you can refit them to carry MIRV nukes, throw in an empty frigate, and buy the refit with scrapped building money ASAP. Creative in addition to telepathic makes the game an almost certain win, and I have won a two-player game in three turns using a creative telepathic transdimensional combination by fanning out my pre-built battleships to the computer's colonies. The refit-to-MIRV conquest usually takes less than 10 turns if properly implemented in a two-player game, and less than 25 turns in an eight-player game.
Dedicated spying races often work in the advanced game (much more than they do in non-advanced), but they are by no means certain. It is generally advisable to have a researching population despite having spies doing your research for you, unless you plan to immediately blitz with MIRV nukes and possibly transports - obviating the whole reason behind spying in the first place. If you impose the artificial constraint of no research at all and no blitz, the game will usually take a while, especially in prewarp, when sometimes you can't build ships until late in the game.
Everyone should try the "check it out" race some time (the first bad race on Chris Chang's page). I killed the Guardian on turn 293 in pre-warp. Can anyone beat that?