Back in late February, SUSA released a report that used polling data from all 50 states to game out hypothetical Clinton-McCain and Obama-McCain general election matchups. The results showed each Democrat would win, but in quite different fashions. Unfortunately the polls had small samples and large margins of error which is problematic since many states had margins of victory less than the margin of error of the state poll (see a more in-depth analysis here).
Another study (both of these were recently highlighted at the Monkey Cage) used a variety of current polls to compile a 50-state electoral map that not only showed a more narrow Obama win, but also a Clinton loss to McCain (most current version here).
Just to make all this even more confusing, electoral-vote.com has its own matchup maps, these generated from recent polls in each state. The results are strikingly unconvincing. Clinton loses to McCain 304-203 (31 tossups) and Obama loses to McCain 303-226 (9 tossups). The problem with these matchups is that many of the polls are old (data from late February) and from less reputable sources. Perhaps the lesson is simply that predicting electoral results this far in advance is a rather futile exercise. The results are neither consistent nor wholly convincing.
But that doesn't mean we should stop doing it. The pols can tell us which states are likely to be competitive. In the tossup column, Obama and Clinton have six states in common (IA, MI, NM, NH, OH and PA), which leaves FL (27 EVs) for Clinton but not Obama to contest and CO, MO, NV and VA (38 EVs) for Obama but not Clinton. Not surprisingly, this meshes well with each candidate's campaign pitch: Obama says he'll bring more states into play for Democrats; Clinton argues that we can't afford to gamble on an untested candidate and uncompetitive states. It's hard to see what McCain's strategy is at this point, but it's clear, from this data at least, which candidate he would rather face.