Right now I'm predicting a 51-49 GOP Senate and a 219-216 GOP House. That could of course change at any time, but that's where I think we are at present.

A few observations: this poll included Friday through Tuesday. Weekend polls are notorious for overestimating Democratic support. There are several main arguments as to why this happens, but there is no doubt it occurs.

Polls of "likely voters" tend to be more accurate in predicting outcomes than polls of "registered voters," and far more predictive than those of "all adults." However, it is important to note the distinctions between various polling firms' methods of determining "likely voters."

Most simply allow the respondent to place himself in the category by asserting his strong intent to vote. This is the weakest qualification of "likely voters," because many people will swear to high heaven they will be there on Election Day, yet still fail to show up.

On the other end of the spectrum, polls can screen strongly for the likely voter by asking questions about past voting habits. Those who voted in the past are more "likely" to continue that behavior. The more consecutive elections in which a respondents has voted, the more likely they are to vote in the next.

This method does eliminate new voters, and those newly enthused, from the sample altogether, so it fails to measure this subset at all.

We could better judge the methodology of polls if they disclosed how many phone interviews they conducted to get the "600 likely voters" in the released figures. If they polled 1000 "adults" and came up with that many LV, they didn't screen very hard. If they conducted 2400 interviews with "adults," the number would be more credible, but pollsters don't do that many interviews in a five-day state poll, because it costs far more than a smaller total sample.

Also, wasn't ARG in the "pantheon of heroes" in 2002 and 2004 like Cook, Sabato, Zogby, AP-Ipsos, cBS/NYT that predicted a Kerry rout?